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.
37474 Commits
32 Branches
555MB
Tree: 5053a9a1ff
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '5053a9a1ff'
${ noResults }
FFmpeg/libavcodec/vc1dec.c

6175 lines
240KB
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 Libav.

  8.  *

  9.  * Libav 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.  * Libav 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 Libav; 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 "avcodec.h"

  31. #include "error_resilience.h"

  32. #include "mpegvideo.h"

  33. #include "h263.h"

  34. #include "h264chroma.h"

  35. #include "vc1.h"

  36. #include "vc1data.h"

  37. #include "vc1acdata.h"

  38. #include "msmpeg4data.h"

  39. #include "unary.h"

  40. #include "mathops.h"

  41. #include "vdpau_internal.h"

  42. 

  43. #undef NDEBUG

  44. #include <assert.h>

  45. 

  46. #define MB_INTRA_VLC_BITS 9

  47. #define DC_VLC_BITS 9

  48. 

  49. 

  50. // offset tables for interlaced picture MVDATA decoding

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

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

  53. 

  54. /***********************************************************************/

  55. /**

  56.  * @name VC-1 Bitplane decoding

  57.  * @see 8.7, p56

  58.  * @{

  59.  */

  60. 

  61. /**

  62.  * Imode types

  63.  * @{

  64.  */

  65. enum Imode {

  66.     IMODE_RAW,

  67.     IMODE_NORM2,

  68.     IMODE_DIFF2,

  69.     IMODE_NORM6,

  70.     IMODE_DIFF6,

  71.     IMODE_ROWSKIP,

  72.     IMODE_COLSKIP

  73. };

  74. /** @} */ //imode defines

  75. 

  76. static void init_block_index(VC1Context *v)

  77. {

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

  79.     ff_init_block_index(s);

  80.     if (v->field_mode && !(v->second_field ^ v->tff)) {

  81.         s->dest[0] += s->current_picture_ptr->f.linesize[0];

  82.         s->dest[1] += s->current_picture_ptr->f.linesize[1];

  83.         s->dest[2] += s->current_picture_ptr->f.linesize[2];

  84.     }

  85. }

  86. 

  87. /** @} */ //Bitplane group

  88. 

  89. static void vc1_put_signed_blocks_clamped(VC1Context *v)

  90. {

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

  92.     int topleft_mb_pos, top_mb_pos;

  93.     int stride_y, fieldtx = 0;

  94.     int v_dist;

  95. 

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

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

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

  99.      * present as well.

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

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

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

  103.     if (!s->first_slice_line) {

  104.         if (s->mb_x) {

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

  106.             if (v->fcm == ILACE_FRAME)

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

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

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

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

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

  112.                                              stride_y);

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

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

  115.                                              stride_y);

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

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

  118.                                              stride_y);

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

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

  121.                                              stride_y);

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

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

  124.                                              s->uvlinesize);

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

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

  127.                                              s->uvlinesize);

  128.         }

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

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

  131.             if (v->fcm == ILACE_FRAME)

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

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

  134.             v_dist     = fieldtx ? 15 : 8;

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

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

  137.                                              stride_y);

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

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

  140.                                              stride_y);

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

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

  143.                                              stride_y);

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

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

  146.                                              stride_y);

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

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

  149.                                              s->uvlinesize);

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

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

  152.                                              s->uvlinesize);

  153.         }

  154.     }

  155. 

  156. #define inc_blk_idx(idx) do { \

  157.         idx++; \

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

  159.             idx = 0; \

  160.     } while (0)

  161. 

  162.     inc_blk_idx(v->topleft_blk_idx);

  163.     inc_blk_idx(v->top_blk_idx);

  164.     inc_blk_idx(v->left_blk_idx);

  165.     inc_blk_idx(v->cur_blk_idx);

  166. }

  167. 

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

  169. {

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

  171.     int j;

  172.     if (!s->first_slice_line) {

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

  174.         if (s->mb_x)

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

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

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

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

  179.             if (s->mb_x)

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

  181.         }

  182.     }

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

  184. 

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

  186.         if (s->mb_x) {

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

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

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

  190.         }

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

  192.     }

  193. }

  194. 

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

  196. {

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

  198.     int j;

  199. 

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

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

  202.     if (!s->first_slice_line) {

  203.         if (s->mb_x) {

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

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

  206. 

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

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

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

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

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

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

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

  214.                     }

  215.                 }

  216.             }

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

  218.         }

  219. 

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

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

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

  223. 

  224.                 if (s->mb_x)

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

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

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

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

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

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

  231.                     }

  232.                 }

  233.             }

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

  235.         }

  236. 

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

  238.             if (s->mb_x) {

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

  240.                     v->vc1dsp.vc1_h_loop_filter16(s->dest[0] - 16 * s->linesize - 16, 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 >= 2) {

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

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

  245.                     }

  246.                 }

  247.             }

  248. 

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

  250.                 if (s->mb_x)

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

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

  253.                 if (s->mb_x) {

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

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

  256.                     }

  257.                 }

  258.             }

  259.         }

  260.     }

  261. }

  262. 

  263. static void vc1_smooth_overlap_filter_iblk(VC1Context *v)

  264. {

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

  266.     int mb_pos;

  267. 

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

  269.         return;

  270. 

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

  272. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  292.             }

  293.         }

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

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

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

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

  298. 

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

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

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

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

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

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

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

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

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

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

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

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

  311.                 }

  312.             }

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

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

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

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

  317.         }

  318.     }

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

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

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

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

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

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

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

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

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

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

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

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

  331.             }

  332.         }

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

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

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

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

  337.     }

  338. }

  339. 

  340. /** Do motion compensation over 1 macroblock

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

  342.  */

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

  344. {

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

  346.     H264ChromaContext *h264chroma = &v->h264chroma;

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

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

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

  350.     int i;

  351.     const uint8_t *luty, *lutuv;

  352.     int use_ic;

  353. 

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

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

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

  357.         return;

  358. 

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

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

  361. 

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

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

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

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

  366.             s->current_picture.motion_val[1][s->block_index[i] + v->blocks_off][1] = my;

  367.         }

  368.     }

  369. 

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

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

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

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

  374. 

  375.     if (v->field_mode &&

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

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

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

  379.     }

  380. 

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

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

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

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

  385.     }

  386.     if (!dir) {

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

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

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

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

  391.             luty  = v->curr_luty [v->ref_field_type[dir]];

  392.             lutuv = v->curr_lutuv[v->ref_field_type[dir]];

  393.             use_ic = v->curr_use_ic;

  394.         } else {

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

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

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

  398.             luty  = v->last_luty [v->ref_field_type[dir]];

  399.             lutuv = v->last_lutuv[v->ref_field_type[dir]];

  400.             use_ic = v->last_use_ic;

  401.         }

  402.     } else {

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

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

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

  406.         luty  = v->next_luty [v->ref_field_type[dir]];

  407.         lutuv = v->next_lutuv[v->ref_field_type[dir]];

  408.         use_ic = v->next_use_ic;

  409.     }

  410. 

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

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

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

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

  415. 

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

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

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

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

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

  421.     } else {

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

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

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

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

  426.     }

  427. 

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

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

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

  431. 

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

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

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

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

  436.     }

  437. 

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

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

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

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

  442.     }

  443. 

  444.     if (v->rangeredfrm || use_ic

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

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

  447.         || (unsigned)(src_y - 1)        > v_edge_pos    - (my&3) - 16 - 3) {

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

  449. 

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

  451.         s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize,

  452.                                  17 + s->mspel * 2, 17 + s->mspel * 2,

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

  454.                                  s->h_edge_pos, v_edge_pos);

  455.         srcY = s->edge_emu_buffer;

  456.         s->vdsp.emulated_edge_mc(uvbuf     , srcU, s->uvlinesize, 8 + 1, 8 + 1,

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

  458.         s->vdsp.emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8 + 1, 8 + 1,

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

  460.         srcU = uvbuf;

  461.         srcV = uvbuf + 16;

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

  463.         if (v->rangeredfrm) {

  464.             int i, j;

  465.             uint8_t *src, *src2;

  466. 

  467.             src = srcY;

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

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

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

  471.                 src += s->linesize;

  472.             }

  473.             src  = srcU;

  474.             src2 = srcV;

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

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

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

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

  479.                 }

  480.                 src  += s->uvlinesize;

  481.                 src2 += s->uvlinesize;

  482.             }

  483.         }

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

  485.         if (use_ic) {

  486.             int i, j;

  487.             uint8_t *src, *src2;

  488. 

  489.             src = srcY;

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

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

  492.                     src[i] = luty[src[i]];

  493.                 src += s->linesize;

  494.             }

  495.             src  = srcU;

  496.             src2 = srcV;

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

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

  499.                     src[i]  = lutuv[src[i]];

  500.                     src2[i] = lutuv[src2[i]];

  501.                 }

  502.                 src  += s->uvlinesize;

  503.                 src2 += s->uvlinesize;

  504.             }

  505.         }

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

  507.     }

  508. 

  509.     if (s->mspel) {

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

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

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

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

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

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

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

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

  518.         if (!v->rnd)

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

  520.         else

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

  522.     }

  523. 

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

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

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

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

  528.     if (!v->rnd) {

  529.         h264chroma->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);

  530.         h264chroma->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);

  531.     } else {

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

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

  534.     }

  535. }

  536. 

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

  538. {

  539.     if (a < b) {

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

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

  542.     } else {

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

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

  545.     }

  546. }

  547. 

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

  549.  */

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

  551. {

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

  553.     uint8_t *srcY;

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

  555.     int off;

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

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

  558.     const uint8_t *luty;

  559.     int use_ic;

  560. 

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

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

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

  564.         return;

  565. 

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

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

  568. 

  569.     if (!dir) {

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

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

  572.             luty = v->curr_luty[v->ref_field_type[dir]];

  573.             use_ic = v->curr_use_ic;

  574.         } else {

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

  576.             luty = v->last_luty[v->ref_field_type[dir]];

  577.             use_ic = v->last_use_ic;

  578.         }

  579.     } else {

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

  581.         luty = v->next_luty[v->ref_field_type[dir]];

  582.         use_ic = v->next_use_ic;

  583.     }

  584. 

  585.     if (v->field_mode) {

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

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

  588.     }

  589. 

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

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

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

  593.         int tx, ty;

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

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

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

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

  598.             opp_count  += f;

  599.             same_count += 1 - f;

  600.         }

  601.         f = opp_count > same_count;

  602.         switch (f ? opp_count : same_count) {

  603.         case 4:

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

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

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

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

  608.             break;

  609.         case 3:

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

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

  612.             break;

  613.         case 2:

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

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

  616.             break;

  617.         }

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

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

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

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

  622.     }

  623. 

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

  625.         int qx, qy;

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

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

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

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

  630.             s->current_picture.motion_val[1][s->block_index[n] + v->blocks_off][1] = my;

  631.         }

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

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

  634. 

  635.         if (qx < -17)

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

  637.         else if (qx > width)

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

  639.         if (qy < -18)

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

  641.         else if (qy > height + 1)

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

  643.     }

  644. 

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

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

  647.     else

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

  649. 

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

  651.     if (!fieldmv)

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

  653.     else

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

  655. 

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

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

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

  659.     } else {

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

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

  662.             if (src_y & 1)

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

  664.             else

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

  666.         } else {

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

  668.         }

  669.     }

  670. 

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

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

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

  674. 

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

  676.         v_edge_pos--;

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

  678.         src_y--;

  679.     if (v->rangeredfrm || use_ic

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

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

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

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

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

  685.         s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize,

  686.                                  9 + s->mspel * 2, (9 + s->mspel * 2) << fieldmv,

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

  688.                                  s->h_edge_pos, v_edge_pos);

  689.         srcY = s->edge_emu_buffer;

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

  691.         if (v->rangeredfrm) {

  692.             int i, j;

  693.             uint8_t *src;

  694. 

  695.             src = srcY;

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

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

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

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

  700.             }

  701.         }

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

  703.         if (use_ic) {

  704.             int i, j;

  705.             uint8_t *src;

  706. 

  707.             src = srcY;

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

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

  710.                     src[i] = luty[src[i]];

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

  712.             }

  713.         }

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

  715.     }

  716. 

  717.     if (s->mspel) {

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

  719.         if (avg)

  720.             v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize << fieldmv, v->rnd);

  721.         else

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

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

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

  725.         if (!v->rnd)

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

  727.         else

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

  729.     }

  730. }

  731. 

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

  733. {

  734.     int idx, i;

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

  736. 

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

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

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

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

  741.     if (!idx) {

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

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

  744.         return 4;

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

  746.         switch (idx) {

  747.         case 0x1:

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

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

  750.             return 3;

  751.         case 0x2:

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

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

  754.             return 3;

  755.         case 0x4:

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

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

  758.             return 3;

  759.         case 0x8:

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

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

  762.             return 3;

  763.         }

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

  765.         int t1 = 0, t2 = 0;

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

  767.             if (!a[i]) {

  768.                 t1 = i;

  769.                 break;

  770.             }

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

  772.             if (!a[i]) {

  773.                 t2 = i;

  774.                 break;

  775.             }

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

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

  778.         return 2;

  779.     } else {

  780.         return 0;

  781.     }

  782.     return -1;

  783. }

  784. 

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

  786.  */

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

  788. {

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

  790.     H264ChromaContext *h264chroma = &v->h264chroma;

  791.     uint8_t *srcU, *srcV;

  792.     int uvmx, uvmy, uvsrc_x, uvsrc_y;

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

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

  795.     int valid_count;

  796.     int chroma_ref_type = v->cur_field_type;

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

  798.     const uint8_t *lutuv;

  799.     int use_ic;

  800. 

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

  802.         return;

  803.     if (s->flags & CODEC_FLAG_GRAY)

  804.         return;

  805. 

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

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

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

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

  810.         if (v->field_mode)

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

  812.     }

  813. 

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

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

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

  817.         chroma_ref_type = v->reffield;

  818.         if (!valid_count) {

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

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

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

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

  823.         }

  824.     } else {

  825.         int dominant = 0;

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

  827.             dominant = 1;

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

  829.         if (dominant)

  830.             chroma_ref_type = !v->cur_field_type;

  831.     }

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

  833.         return;

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

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

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

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

  838. 

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

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

  841. 

  842.     if (v->fastuvmc) {

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

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

  845.     }

  846.     // Field conversion bias

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

  848.         uvmy += 2 - 4 * chroma_ref_type;

  849. 

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

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

  852. 

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

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

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

  856.     } else {

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

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

  859.     }

  860. 

  861.     if (!dir) {

  862.         if (v->field_mode && (v->cur_field_type != chroma_ref_type) && v->second_field) {

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

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

  865.             lutuv = v->curr_lutuv[chroma_ref_type];

  866.             use_ic = v->curr_use_ic;

  867.         } else {

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

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

  870.             lutuv = v->last_lutuv[chroma_ref_type];

  871.             use_ic = v->last_use_ic;

  872.         }

  873.     } else {

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

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

  876.         lutuv = v->next_lutuv[chroma_ref_type];

  877.         use_ic = v->next_use_ic;

  878.     }

  879. 

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

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

  882. 

  883.     if (v->field_mode) {

  884.         if (chroma_ref_type) {

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

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

  887.         }

  888.     }

  889. 

  890.     if (v->rangeredfrm || use_ic

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

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

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

  894.         s->vdsp.emulated_edge_mc(s->edge_emu_buffer     , srcU, s->uvlinesize,

  895.                                  8 + 1, 8 + 1, uvsrc_x, uvsrc_y,

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

  897.         s->vdsp.emulated_edge_mc(s->edge_emu_buffer + 16, srcV, s->uvlinesize,

  898.                                  8 + 1, 8 + 1, uvsrc_x, uvsrc_y,

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

  900.         srcU = s->edge_emu_buffer;

  901.         srcV = s->edge_emu_buffer + 16;

  902. 

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

  904.         if (v->rangeredfrm) {

  905.             int i, j;

  906.             uint8_t *src, *src2;

  907. 

  908.             src  = srcU;

  909.             src2 = srcV;

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

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

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

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

  914.                 }

  915.                 src  += s->uvlinesize;

  916.                 src2 += s->uvlinesize;

  917.             }

  918.         }

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

  920.         if (use_ic) {

  921.             int i, j;

  922.             uint8_t *src, *src2;

  923. 

  924.             src  = srcU;

  925.             src2 = srcV;

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

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

  928.                     src[i]  = lutuv[src[i]];

  929.                     src2[i] = lutuv[src2[i]];

  930.                 }

  931.                 src  += s->uvlinesize;

  932.                 src2 += s->uvlinesize;

  933.             }

  934.         }

  935.     }

  936. 

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

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

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

  940.     if (!v->rnd) {

  941.         h264chroma->put_h264_chroma_pixels_tab[0](s->dest[1], srcU, s->uvlinesize, 8, uvmx, uvmy);

  942.         h264chroma->put_h264_chroma_pixels_tab[0](s->dest[2], srcV, s->uvlinesize, 8, uvmx, uvmy);

  943.     } else {

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

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

  946.     }

  947. }

  948. 

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

  950.  */

  951. static void vc1_mc_4mv_chroma4(VC1Context *v)

  952. {

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

  954.     H264ChromaContext *h264chroma = &v->h264chroma;

  955.     uint8_t *srcU, *srcV;

  956.     int uvsrc_x, uvsrc_y;

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

  958.     int i, off, tx, ty;

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

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

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

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

  963.     int use_ic = v->last_use_ic;

  964. 

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

  966.         return;

  967.     if (s->flags & CODEC_FLAG_GRAY)

  968.         return;

  969. 

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

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

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

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

  974.         if (fieldmv)

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

  976.         else

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

  978.     }

  979. 

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

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

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

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

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

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

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

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

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

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

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

  991. 

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

  993.             v_edge_pos--;

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

  995.             uvsrc_y--;

  996.         if (use_ic

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

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

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

  1000.             s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcU, s->uvlinesize,

  1001.                                      5, (5 << fieldmv), uvsrc_x, uvsrc_y,

  1002.                                      s->h_edge_pos >> 1, v_edge_pos);

  1003.             s->vdsp.emulated_edge_mc(s->edge_emu_buffer + 16, srcV, s->uvlinesize,

  1004.                                      5, (5 << fieldmv), uvsrc_x, uvsrc_y,

  1005.                                      s->h_edge_pos >> 1, v_edge_pos);

  1006.             srcU = s->edge_emu_buffer;

  1007.             srcV = s->edge_emu_buffer + 16;

  1008. 

  1009.             /* if we deal with intensity compensation we need to scale source blocks */

  1010.             if (use_ic) {

  1011.                 int i, j;

  1012.                 uint8_t *src, *src2;

  1013.                 const uint8_t *lutuv = v->last_lutuv[v->ref_field_type[0]];

  1014. 

  1015.                 src  = srcU;

  1016.                 src2 = srcV;

  1017.                 for (j = 0; j < 5; j++) {

  1018.                     for (i = 0; i < 5; i++) {

  1019.                         src[i]  = lutuv[src[i]];

  1020.                         src2[i] = lutuv[src2[i]];

  1021.                     }

  1022.                     src  += s->uvlinesize << 1;

  1023.                     src2 += s->uvlinesize << 1;

  1024.                 }

  1025.             }

  1026.         }

  1027.         if (!v->rnd) {

  1028.             h264chroma->put_h264_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);

  1029.             h264chroma->put_h264_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);

  1030.         } else {

  1031.             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]);

  1032.             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]);

  1033.         }

  1034.     }

  1035. }

  1036. 

  1037. /***********************************************************************/

  1038. /**

  1039.  * @name VC-1 Block-level functions

  1040.  * @see 7.1.4, p91 and 8.1.1.7, p(1)04

  1041.  * @{

  1042.  */

  1043. 

  1044. /**

  1045.  * @def GET_MQUANT

  1046.  * @brief Get macroblock-level quantizer scale

  1047.  */

  1048. #define GET_MQUANT()                                           \

  1049.     if (v->dquantfrm) {                                        \

  1050.         int edges = 0;                                         \

  1051.         if (v->dqprofile == DQPROFILE_ALL_MBS) {               \

  1052.             if (v->dqbilevel) {                                \

  1053.                 mquant = (get_bits1(gb)) ? v->altpq : v->pq;   \

  1054.             } else {                                           \

  1055.                 mqdiff = get_bits(gb, 3);                      \

  1056.                 if (mqdiff != 7)                               \

  1057.                     mquant = v->pq + mqdiff;                   \

  1058.                 else                                           \

  1059.                     mquant = get_bits(gb, 5);                  \

  1060.             }                                                  \

  1061.         }                                                      \

  1062.         if (v->dqprofile == DQPROFILE_SINGLE_EDGE)             \

  1063.             edges = 1 << v->dqsbedge;                          \

  1064.         else if (v->dqprofile == DQPROFILE_DOUBLE_EDGES)       \

  1065.             edges = (3 << v->dqsbedge) % 15;                   \

  1066.         else if (v->dqprofile == DQPROFILE_FOUR_EDGES)         \

  1067.             edges = 15;                                        \

  1068.         if ((edges&1) && !s->mb_x)                             \

  1069.             mquant = v->altpq;                                 \

  1070.         if ((edges&2) && s->first_slice_line)                  \

  1071.             mquant = v->altpq;                                 \

  1072.         if ((edges&4) && s->mb_x == (s->mb_width - 1))         \

  1073.             mquant = v->altpq;                                 \

  1074.         if ((edges&8) && s->mb_y == (s->mb_height - 1))        \

  1075.             mquant = v->altpq;                                 \

  1076.         if (!mquant || mquant > 31) {                          \

  1077.             av_log(v->s.avctx, AV_LOG_ERROR,                   \

  1078.                    "Overriding invalid mquant %d\n", mquant);  \

  1079.             mquant = 1;                                        \

  1080.         }                                                      \

  1081.     }

  1082. 

  1083. /**

  1084.  * @def GET_MVDATA(_dmv_x, _dmv_y)

  1085.  * @brief Get MV differentials

  1086.  * @see MVDATA decoding from 8.3.5.2, p(1)20

  1087.  * @param _dmv_x Horizontal differential for decoded MV

  1088.  * @param _dmv_y Vertical differential for decoded MV

  1089.  */

  1090. #define GET_MVDATA(_dmv_x, _dmv_y)                                      \

  1091.     index = 1 + get_vlc2(gb, ff_vc1_mv_diff_vlc[s->mv_table_index].table, \

  1092.                          VC1_MV_DIFF_VLC_BITS, 2);                      \

  1093.     if (index > 36) {                                                   \

  1094.         mb_has_coeffs = 1;                                              \

  1095.         index -= 37;                                                    \

  1096.     } else                                                              \

  1097.         mb_has_coeffs = 0;                                              \

  1098.     s->mb_intra = 0;                                                    \

  1099.     if (!index) {                                                       \

  1100.         _dmv_x = _dmv_y = 0;                                            \

  1101.     } else if (index == 35) {                                           \

  1102.         _dmv_x = get_bits(gb, v->k_x - 1 + s->quarter_sample);          \

  1103.         _dmv_y = get_bits(gb, v->k_y - 1 + s->quarter_sample);          \

  1104.     } else if (index == 36) {                                           \

  1105.         _dmv_x = 0;                                                     \

  1106.         _dmv_y = 0;                                                     \

  1107.         s->mb_intra = 1;                                                \

  1108.     } else {                                                            \

  1109.         index1 = index % 6;                                             \

  1110.         if (!s->quarter_sample && index1 == 5) val = 1;                 \

  1111.         else                                   val = 0;                 \

  1112.         if (size_table[index1] - val > 0)                               \

  1113.             val = get_bits(gb, size_table[index1] - val);               \

  1114.         else                                   val = 0;                 \

  1115.         sign = 0 - (val&1);                                             \

  1116.         _dmv_x = (sign ^ ((val>>1) + offset_table[index1])) - sign;     \

  1117.                                                                         \

  1118.         index1 = index / 6;                                             \

  1119.         if (!s->quarter_sample && index1 == 5) val = 1;                 \

  1120.         else                                   val = 0;                 \

  1121.         if (size_table[index1] - val > 0)                               \

  1122.             val = get_bits(gb, size_table[index1] - val);               \

  1123.         else                                   val = 0;                 \

  1124.         sign = 0 - (val & 1);                                           \

  1125.         _dmv_y = (sign ^ ((val >> 1) + offset_table[index1])) - sign;   \

  1126.     }

  1127. 

  1128. static av_always_inline void get_mvdata_interlaced(VC1Context *v, int *dmv_x,

  1129.                                                    int *dmv_y, int *pred_flag)

  1130. {

  1131.     int index, index1;

  1132.     int extend_x = 0, extend_y = 0;

  1133.     GetBitContext *gb = &v->s.gb;

  1134.     int bits, esc;

  1135.     int val, sign;

  1136.     const int* offs_tab;

  1137. 

  1138.     if (v->numref) {

  1139.         bits = VC1_2REF_MVDATA_VLC_BITS;

  1140.         esc  = 125;

  1141.     } else {

  1142.         bits = VC1_1REF_MVDATA_VLC_BITS;

  1143.         esc  = 71;

  1144.     }

  1145.     switch (v->dmvrange) {

  1146.     case 1:

  1147.         extend_x = 1;

  1148.         break;

  1149.     case 2:

  1150.         extend_y = 1;

  1151.         break;

  1152.     case 3:

  1153.         extend_x = extend_y = 1;

  1154.         break;

  1155.     }

  1156.     index = get_vlc2(gb, v->imv_vlc->table, bits, 3);

  1157.     if (index == esc) {

  1158.         *dmv_x = get_bits(gb, v->k_x);

  1159.         *dmv_y = get_bits(gb, v->k_y);

  1160.         if (v->numref) {

  1161.             if (pred_flag) {

  1162.                 *pred_flag = *dmv_y & 1;

  1163.                 *dmv_y     = (*dmv_y + *pred_flag) >> 1;

  1164.             } else {

  1165.                 *dmv_y     = (*dmv_y + (*dmv_y & 1)) >> 1;

  1166.             }

  1167.         }

  1168.     }

  1169.     else {

  1170.         if (extend_x)

  1171.             offs_tab = offset_table2;

  1172.         else

  1173.             offs_tab = offset_table1;

  1174.         index1 = (index + 1) % 9;

  1175.         if (index1 != 0) {

  1176.             val    = get_bits(gb, index1 + extend_x);

  1177.             sign   = 0 -(val & 1);

  1178.             *dmv_x = (sign ^ ((val >> 1) + offs_tab[index1])) - sign;

  1179.         } else

  1180.             *dmv_x = 0;

  1181.         if (extend_y)

  1182.             offs_tab = offset_table2;

  1183.         else

  1184.             offs_tab = offset_table1;

  1185.         index1 = (index + 1) / 9;

  1186.         if (index1 > v->numref) {

  1187.             val    = get_bits(gb, (index1 + (extend_y << v->numref)) >> v->numref);

  1188.             sign   = 0 - (val & 1);

  1189.             *dmv_y = (sign ^ ((val >> 1) + offs_tab[index1 >> v->numref])) - sign;

  1190.         } else

  1191.             *dmv_y = 0;

  1192.         if (v->numref && pred_flag)

  1193.             *pred_flag = index1 & 1;

  1194.     }

  1195. }

  1196. 

  1197. static av_always_inline int scaleforsame_x(VC1Context *v, int n /* MV */, int dir)

  1198. {

  1199.     int scaledvalue, refdist;

  1200.     int scalesame1, scalesame2;

  1201.     int scalezone1_x, zone1offset_x;

  1202.     int table_index = dir ^ v->second_field;

  1203. 

  1204.     if (v->s.pict_type != AV_PICTURE_TYPE_B)

  1205.         refdist = v->refdist;

  1206.     else

  1207.         refdist = dir ? v->brfd : v->frfd;

  1208.     if (refdist > 3)

  1209.         refdist = 3;

  1210.     scalesame1    = ff_vc1_field_mvpred_scales[table_index][1][refdist];

  1211.     scalesame2    = ff_vc1_field_mvpred_scales[table_index][2][refdist];

  1212.     scalezone1_x  = ff_vc1_field_mvpred_scales[table_index][3][refdist];

  1213.     zone1offset_x = ff_vc1_field_mvpred_scales[table_index][5][refdist];

  1214. 

  1215.     if (FFABS(n) > 255)

  1216.         scaledvalue = n;

  1217.     else {

  1218.         if (FFABS(n) < scalezone1_x)

  1219.             scaledvalue = (n * scalesame1) >> 8;

  1220.         else {

  1221.             if (n < 0)

  1222.                 scaledvalue = ((n * scalesame2) >> 8) - zone1offset_x;

  1223.             else

  1224.                 scaledvalue = ((n * scalesame2) >> 8) + zone1offset_x;

  1225.         }

  1226.     }

  1227.     return av_clip(scaledvalue, -v->range_x, v->range_x - 1);

  1228. }

  1229. 

  1230. static av_always_inline int scaleforsame_y(VC1Context *v, int i, int n /* MV */, int dir)

  1231. {

  1232.     int scaledvalue, refdist;

  1233.     int scalesame1, scalesame2;

  1234.     int scalezone1_y, zone1offset_y;

  1235.     int table_index = dir ^ v->second_field;

  1236. 

  1237.     if (v->s.pict_type != AV_PICTURE_TYPE_B)

  1238.         refdist = v->refdist;

  1239.     else

  1240.         refdist = dir ? v->brfd : v->frfd;

  1241.     if (refdist > 3)

  1242.         refdist = 3;

  1243.     scalesame1    = ff_vc1_field_mvpred_scales[table_index][1][refdist];

  1244.     scalesame2    = ff_vc1_field_mvpred_scales[table_index][2][refdist];

  1245.     scalezone1_y  = ff_vc1_field_mvpred_scales[table_index][4][refdist];

  1246.     zone1offset_y = ff_vc1_field_mvpred_scales[table_index][6][refdist];

  1247. 

  1248.     if (FFABS(n) > 63)

  1249.         scaledvalue = n;

  1250.     else {

  1251.         if (FFABS(n) < scalezone1_y)

  1252.             scaledvalue = (n * scalesame1) >> 8;

  1253.         else {

  1254.             if (n < 0)

  1255.                 scaledvalue = ((n * scalesame2) >> 8) - zone1offset_y;

  1256.             else

  1257.                 scaledvalue = ((n * scalesame2) >> 8) + zone1offset_y;

  1258.         }

  1259.     }

  1260. 

  1261.     if (v->cur_field_type && !v->ref_field_type[dir])

  1262.         return av_clip(scaledvalue, -v->range_y / 2 + 1, v->range_y / 2);

  1263.     else

  1264.         return av_clip(scaledvalue, -v->range_y / 2, v->range_y / 2 - 1);

  1265. }

  1266. 

  1267. static av_always_inline int scaleforopp_x(VC1Context *v, int n /* MV */)

  1268. {

  1269.     int scalezone1_x, zone1offset_x;

  1270.     int scaleopp1, scaleopp2, brfd;

  1271.     int scaledvalue;

  1272. 

  1273.     brfd = FFMIN(v->brfd, 3);

  1274.     scalezone1_x  = ff_vc1_b_field_mvpred_scales[3][brfd];

  1275.     zone1offset_x = ff_vc1_b_field_mvpred_scales[5][brfd];

  1276.     scaleopp1     = ff_vc1_b_field_mvpred_scales[1][brfd];

  1277.     scaleopp2     = ff_vc1_b_field_mvpred_scales[2][brfd];

  1278. 

  1279.     if (FFABS(n) > 255)

  1280.         scaledvalue = n;

  1281.     else {

  1282.         if (FFABS(n) < scalezone1_x)

  1283.             scaledvalue = (n * scaleopp1) >> 8;

  1284.         else {

  1285.             if (n < 0)

  1286.                 scaledvalue = ((n * scaleopp2) >> 8) - zone1offset_x;

  1287.             else

  1288.                 scaledvalue = ((n * scaleopp2) >> 8) + zone1offset_x;

  1289.         }

  1290.     }

  1291.     return av_clip(scaledvalue, -v->range_x, v->range_x - 1);

  1292. }

  1293. 

  1294. static av_always_inline int scaleforopp_y(VC1Context *v, int n /* MV */, int dir)

  1295. {

  1296.     int scalezone1_y, zone1offset_y;

  1297.     int scaleopp1, scaleopp2, brfd;

  1298.     int scaledvalue;

  1299. 

  1300.     brfd = FFMIN(v->brfd, 3);

  1301.     scalezone1_y  = ff_vc1_b_field_mvpred_scales[4][brfd];

  1302.     zone1offset_y = ff_vc1_b_field_mvpred_scales[6][brfd];

  1303.     scaleopp1     = ff_vc1_b_field_mvpred_scales[1][brfd];

  1304.     scaleopp2     = ff_vc1_b_field_mvpred_scales[2][brfd];

  1305. 

  1306.     if (FFABS(n) > 63)

  1307.         scaledvalue = n;

  1308.     else {

  1309.         if (FFABS(n) < scalezone1_y)

  1310.             scaledvalue = (n * scaleopp1) >> 8;

  1311.         else {

  1312.             if (n < 0)

  1313.                 scaledvalue = ((n * scaleopp2) >> 8) - zone1offset_y;

  1314.             else

  1315.                 scaledvalue = ((n * scaleopp2) >> 8) + zone1offset_y;

  1316.         }

  1317.     }

  1318.     if (v->cur_field_type && !v->ref_field_type[dir]) {

  1319.         return av_clip(scaledvalue, -v->range_y / 2 + 1, v->range_y / 2);

  1320.     } else {

  1321.         return av_clip(scaledvalue, -v->range_y / 2, v->range_y / 2 - 1);

  1322.     }

  1323. }

  1324. 

  1325. static av_always_inline int scaleforsame(VC1Context *v, int i, int n /* MV */,

  1326.                                          int dim, int dir)

  1327. {

  1328.     int brfd, scalesame;

  1329.     int hpel = 1 - v->s.quarter_sample;

  1330. 

  1331.     n >>= hpel;

  1332.     if (v->s.pict_type != AV_PICTURE_TYPE_B || v->second_field || !dir) {

  1333.         if (dim)

  1334.             n = scaleforsame_y(v, i, n, dir) << hpel;

  1335.         else

  1336.             n = scaleforsame_x(v, n, dir) << hpel;

  1337.         return n;

  1338.     }

  1339.     brfd      = FFMIN(v->brfd, 3);

  1340.     scalesame = ff_vc1_b_field_mvpred_scales[0][brfd];

  1341. 

  1342.     n = (n * scalesame >> 8) << hpel;

  1343.     return n;

  1344. }

  1345. 

  1346. static av_always_inline int scaleforopp(VC1Context *v, int n /* MV */,

  1347.                                         int dim, int dir)

  1348. {

  1349.     int refdist, scaleopp;

  1350.     int hpel = 1 - v->s.quarter_sample;

  1351. 

  1352.     n >>= hpel;

  1353.     if (v->s.pict_type == AV_PICTURE_TYPE_B && !v->second_field && dir == 1) {

  1354.         if (dim)

  1355.             n = scaleforopp_y(v, n, dir) << hpel;

  1356.         else

  1357.             n = scaleforopp_x(v, n) << hpel;

  1358.         return n;

  1359.     }

  1360.     if (v->s.pict_type != AV_PICTURE_TYPE_B)

  1361.         refdist = FFMIN(v->refdist, 3);

  1362.     else

  1363.         refdist = dir ? v->brfd : v->frfd;

  1364.     scaleopp = ff_vc1_field_mvpred_scales[dir ^ v->second_field][0][refdist];

  1365. 

  1366.     n = (n * scaleopp >> 8) << hpel;

  1367.     return n;

  1368. }

  1369. 

  1370. /** Predict and set motion vector

  1371.  */

  1372. static inline void vc1_pred_mv(VC1Context *v, int n, int dmv_x, int dmv_y,

  1373.                                int mv1, int r_x, int r_y, uint8_t* is_intra,

  1374.                                int pred_flag, int dir)

  1375. {

  1376.     MpegEncContext *s = &v->s;

  1377.     int xy, wrap, off = 0;

  1378.     int16_t *A, *B, *C;

  1379.     int px, py;

  1380.     int sum;

  1381.     int mixedmv_pic, num_samefield = 0, num_oppfield = 0;

  1382.     int opposite, a_f, b_f, c_f;

  1383.     int16_t field_predA[2];

  1384.     int16_t field_predB[2];

  1385.     int16_t field_predC[2];

  1386.     int a_valid, b_valid, c_valid;

  1387.     int hybridmv_thresh, y_bias = 0;

  1388. 

  1389.     if (v->mv_mode == MV_PMODE_MIXED_MV ||

  1390.         ((v->mv_mode == MV_PMODE_INTENSITY_COMP) && (v->mv_mode2 == MV_PMODE_MIXED_MV)))

  1391.         mixedmv_pic = 1;

  1392.     else

  1393.         mixedmv_pic = 0;

  1394.     /* scale MV difference to be quad-pel */

  1395.     dmv_x <<= 1 - s->quarter_sample;

  1396.     dmv_y <<= 1 - s->quarter_sample;

  1397. 

  1398.     wrap = s->b8_stride;

  1399.     xy   = s->block_index[n];

  1400. 

  1401.     if (s->mb_intra) {

  1402.         s->mv[0][n][0] = s->current_picture.motion_val[0][xy + v->blocks_off][0] = 0;

  1403.         s->mv[0][n][1] = s->current_picture.motion_val[0][xy + v->blocks_off][1] = 0;

  1404.         s->current_picture.motion_val[1][xy + v->blocks_off][0] = 0;

  1405.         s->current_picture.motion_val[1][xy + v->blocks_off][1] = 0;

  1406.         if (mv1) { /* duplicate motion data for 1-MV block */

  1407.             s->current_picture.motion_val[0][xy + 1 + v->blocks_off][0]        = 0;

  1408.             s->current_picture.motion_val[0][xy + 1 + v->blocks_off][1]        = 0;

  1409.             s->current_picture.motion_val[0][xy + wrap + v->blocks_off][0]     = 0;

  1410.             s->current_picture.motion_val[0][xy + wrap + v->blocks_off][1]     = 0;

  1411.             s->current_picture.motion_val[0][xy + wrap + 1 + v->blocks_off][0] = 0;

  1412.             s->current_picture.motion_val[0][xy + wrap + 1 + v->blocks_off][1] = 0;

  1413.             v->luma_mv[s->mb_x][0] = v->luma_mv[s->mb_x][1] = 0;

  1414.             s->current_picture.motion_val[1][xy + 1 + v->blocks_off][0]        = 0;

  1415.             s->current_picture.motion_val[1][xy + 1 + v->blocks_off][1]        = 0;

  1416.             s->current_picture.motion_val[1][xy + wrap][0]                     = 0;

  1417.             s->current_picture.motion_val[1][xy + wrap + v->blocks_off][1]     = 0;

  1418.             s->current_picture.motion_val[1][xy + wrap + 1 + v->blocks_off][0] = 0;

  1419.             s->current_picture.motion_val[1][xy + wrap + 1 + v->blocks_off][1] = 0;

  1420.         }

  1421.         return;

  1422.     }

  1423. 

  1424.     C = s->current_picture.motion_val[dir][xy -    1 + v->blocks_off];

  1425.     A = s->current_picture.motion_val[dir][xy - wrap + v->blocks_off];

  1426.     if (mv1) {

  1427.         if (v->field_mode && mixedmv_pic)

  1428.             off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;

  1429.         else

  1430.             off = (s->mb_x == (s->mb_width - 1)) ? -1 : 2;

  1431.     } else {

  1432.         //in 4-MV mode different blocks have different B predictor position

  1433.         switch (n) {

  1434.         case 0:

  1435.             off = (s->mb_x > 0) ? -1 : 1;

  1436.             break;

  1437.         case 1:

  1438.             off = (s->mb_x == (s->mb_width - 1)) ? -1 : 1;

  1439.             break;

  1440.         case 2:

  1441.             off = 1;

  1442.             break;

  1443.         case 3:

  1444.             off = -1;

  1445.         }

  1446.     }

  1447.     B = s->current_picture.motion_val[dir][xy - wrap + off + v->blocks_off];

  1448. 

  1449.     a_valid = !s->first_slice_line || (n == 2 || n == 3);

  1450.     b_valid = a_valid && (s->mb_width > 1);

  1451.     c_valid = s->mb_x || (n == 1 || n == 3);

  1452.     if (v->field_mode) {

  1453.         a_valid = a_valid && !is_intra[xy - wrap];

  1454.         b_valid = b_valid && !is_intra[xy - wrap + off];

  1455.         c_valid = c_valid && !is_intra[xy - 1];

  1456.     }

  1457. 

  1458.     if (a_valid) {

  1459.         a_f = v->mv_f[dir][xy - wrap + v->blocks_off];

  1460.         num_oppfield  += a_f;

  1461.         num_samefield += 1 - a_f;

  1462.         field_predA[0] = A[0];

  1463.         field_predA[1] = A[1];

  1464.     } else {

  1465.         field_predA[0] = field_predA[1] = 0;

  1466.         a_f = 0;

  1467.     }

  1468.     if (b_valid) {

  1469.         b_f = v->mv_f[dir][xy - wrap + off + v->blocks_off];

  1470.         num_oppfield  += b_f;

  1471.         num_samefield += 1 - b_f;

  1472.         field_predB[0] = B[0];

  1473.         field_predB[1] = B[1];

  1474.     } else {

  1475.         field_predB[0] = field_predB[1] = 0;

  1476.         b_f = 0;

  1477.     }

  1478.     if (c_valid) {

  1479.         c_f = v->mv_f[dir][xy - 1 + v->blocks_off];

  1480.         num_oppfield  += c_f;

  1481.         num_samefield += 1 - c_f;

  1482.         field_predC[0] = C[0];

  1483.         field_predC[1] = C[1];

  1484.     } else {

  1485.         field_predC[0] = field_predC[1] = 0;

  1486.         c_f = 0;

  1487.     }

  1488. 

  1489.     if (v->field_mode) {

  1490.         if (!v->numref)

  1491.             // REFFIELD determines if the last field or the second-last field is

  1492.             // to be used as reference

  1493.             opposite = 1 - v->reffield;

  1494.         else {

  1495.             if (num_samefield <= num_oppfield)

  1496.                 opposite = 1 - pred_flag;

  1497.             else

  1498.                 opposite = pred_flag;

  1499.         }

  1500.     } else

  1501.         opposite = 0;

  1502.     if (opposite) {

  1503.         if (a_valid && !a_f) {

  1504.             field_predA[0] = scaleforopp(v, field_predA[0], 0, dir);

  1505.             field_predA[1] = scaleforopp(v, field_predA[1], 1, dir);

  1506.         }

  1507.         if (b_valid && !b_f) {

  1508.             field_predB[0] = scaleforopp(v, field_predB[0], 0, dir);

  1509.             field_predB[1] = scaleforopp(v, field_predB[1], 1, dir);

  1510.         }

  1511.         if (c_valid && !c_f) {

  1512.             field_predC[0] = scaleforopp(v, field_predC[0], 0, dir);

  1513.             field_predC[1] = scaleforopp(v, field_predC[1], 1, dir);

  1514.         }

  1515.         v->mv_f[dir][xy + v->blocks_off] = 1;

  1516.         v->ref_field_type[dir] = !v->cur_field_type;

  1517.     } else {

  1518.         if (a_valid && a_f) {

  1519.             field_predA[0] = scaleforsame(v, n, field_predA[0], 0, dir);

  1520.             field_predA[1] = scaleforsame(v, n, field_predA[1], 1, dir);

  1521.         }

  1522.         if (b_valid && b_f) {

  1523.             field_predB[0] = scaleforsame(v, n, field_predB[0], 0, dir);

  1524.             field_predB[1] = scaleforsame(v, n, field_predB[1], 1, dir);

  1525.         }

  1526.         if (c_valid && c_f) {

  1527.             field_predC[0] = scaleforsame(v, n, field_predC[0], 0, dir);

  1528.             field_predC[1] = scaleforsame(v, n, field_predC[1], 1, dir);

  1529.         }

  1530.         v->mv_f[dir][xy + v->blocks_off] = 0;

  1531.         v->ref_field_type[dir] = v->cur_field_type;

  1532.     }

  1533. 

  1534.     if (a_valid) {

  1535.         px = field_predA[0];

  1536.         py = field_predA[1];

  1537.     } else if (c_valid) {

  1538.         px = field_predC[0];

  1539.         py = field_predC[1];

  1540.     } else if (b_valid) {

  1541.         px = field_predB[0];

  1542.         py = field_predB[1];

  1543.     } else {

  1544.         px = 0;

  1545.         py = 0;

  1546.     }

  1547. 

  1548.     if (num_samefield + num_oppfield > 1) {

  1549.         px = mid_pred(field_predA[0], field_predB[0], field_predC[0]);

  1550.         py = mid_pred(field_predA[1], field_predB[1], field_predC[1]);

  1551.     }

  1552. 

  1553.     /* Pullback MV as specified in 8.3.5.3.4 */

  1554.     if (!v->field_mode) {

  1555.         int qx, qy, X, Y;

  1556.         qx = (s->mb_x << 6) + ((n == 1 || n == 3) ? 32 : 0);

  1557.         qy = (s->mb_y << 6) + ((n == 2 || n == 3) ? 32 : 0);

  1558.         X  = (s->mb_width  << 6) - 4;

  1559.         Y  = (s->mb_height << 6) - 4;

  1560.         if (mv1) {

  1561.             if (qx + px < -60) px = -60 - qx;

  1562.             if (qy + py < -60) py = -60 - qy;

  1563.         } else {

  1564.             if (qx + px < -28) px = -28 - qx;

  1565.             if (qy + py < -28) py = -28 - qy;

  1566.         }

  1567.         if (qx + px > X) px = X - qx;

  1568.         if (qy + py > Y) py = Y - qy;

  1569.     }

  1570. 

  1571.     if (!v->field_mode || s->pict_type != AV_PICTURE_TYPE_B) {

  1572.         /* Calculate hybrid prediction as specified in 8.3.5.3.5 (also 10.3.5.4.3.5) */

  1573.         hybridmv_thresh = 32;

  1574.         if (a_valid && c_valid) {

  1575.             if (is_intra[xy - wrap])

  1576.                 sum = FFABS(px) + FFABS(py);

  1577.             else

  1578.                 sum = FFABS(px - field_predA[0]) + FFABS(py - field_predA[1]);

  1579.             if (sum > hybridmv_thresh) {

  1580.                 if (get_bits1(&s->gb)) {     // read HYBRIDPRED bit

  1581.                     px = field_predA[0];

  1582.                     py = field_predA[1];

  1583.                 } else {

  1584.                     px = field_predC[0];

  1585.                     py = field_predC[1];

  1586.                 }

  1587.             } else {

  1588.                 if (is_intra[xy - 1])

  1589.                     sum = FFABS(px) + FFABS(py);

  1590.                 else

  1591.                     sum = FFABS(px - field_predC[0]) + FFABS(py - field_predC[1]);

  1592.                 if (sum > hybridmv_thresh) {

  1593.                     if (get_bits1(&s->gb)) {

  1594.                         px = field_predA[0];

  1595.                         py = field_predA[1];

  1596.                     } else {

  1597.                         px = field_predC[0];

  1598.                         py = field_predC[1];

  1599.                     }

  1600.                 }

  1601.             }

  1602.         }

  1603.     }

  1604. 

  1605.     if (v->field_mode && v->numref)

  1606.         r_y >>= 1;

  1607.     if (v->field_mode && v->cur_field_type && v->ref_field_type[dir] == 0)

  1608.         y_bias = 1;

  1609.     /* store MV using signed modulus of MV range defined in 4.11 */

  1610.     s->mv[dir][n][0] = s->current_picture.motion_val[dir][xy + v->blocks_off][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x;

  1611.     s->mv[dir][n][1] = s->current_picture.motion_val[dir][xy + v->blocks_off][1] = ((py + dmv_y + r_y - y_bias) & ((r_y << 1) - 1)) - r_y + y_bias;

  1612.     if (mv1) { /* duplicate motion data for 1-MV block */

  1613.         s->current_picture.motion_val[dir][xy +    1 +     v->blocks_off][0] = s->current_picture.motion_val[dir][xy + v->blocks_off][0];

  1614.         s->current_picture.motion_val[dir][xy +    1 +     v->blocks_off][1] = s->current_picture.motion_val[dir][xy + v->blocks_off][1];

  1615.         s->current_picture.motion_val[dir][xy + wrap +     v->blocks_off][0] = s->current_picture.motion_val[dir][xy + v->blocks_off][0];

  1616.         s->current_picture.motion_val[dir][xy + wrap +     v->blocks_off][1] = s->current_picture.motion_val[dir][xy + v->blocks_off][1];

  1617.         s->current_picture.motion_val[dir][xy + wrap + 1 + v->blocks_off][0] = s->current_picture.motion_val[dir][xy + v->blocks_off][0];

  1618.         s->current_picture.motion_val[dir][xy + wrap + 1 + v->blocks_off][1] = s->current_picture.motion_val[dir][xy + v->blocks_off][1];

  1619.         v->mv_f[dir][xy +    1 + v->blocks_off] = v->mv_f[dir][xy +            v->blocks_off];

  1620.         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];

  1621.     }

  1622. }

  1623. 

  1624. /** Predict and set motion vector for interlaced frame picture MBs

  1625.  */

  1626. static inline void vc1_pred_mv_intfr(VC1Context *v, int n, int dmv_x, int dmv_y,

  1627.                                      int mvn, int r_x, int r_y, uint8_t* is_intra, int dir)

  1628. {

  1629.     MpegEncContext *s = &v->s;

  1630.     int xy, wrap, off = 0;

  1631.     int A[2], B[2], C[2];

  1632.     int px, py;

  1633.     int a_valid = 0, b_valid = 0, c_valid = 0;

  1634.     int field_a, field_b, field_c; // 0: same, 1: opposit

  1635.     int total_valid, num_samefield, num_oppfield;

  1636.     int pos_c, pos_b, n_adj;

  1637. 

  1638.     wrap = s->b8_stride;

  1639.     xy = s->block_index[n];

  1640. 

  1641.     if (s->mb_intra) {

  1642.         s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = 0;

  1643.         s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = 0;

  1644.         s->current_picture.motion_val[1][xy][0] = 0;

  1645.         s->current_picture.motion_val[1][xy][1] = 0;

  1646.         if (mvn == 1) { /* duplicate motion data for 1-MV block */

  1647.             s->current_picture.motion_val[0][xy + 1][0]        = 0;

  1648.             s->current_picture.motion_val[0][xy + 1][1]        = 0;

  1649.             s->current_picture.motion_val[0][xy + wrap][0]     = 0;

  1650.             s->current_picture.motion_val[0][xy + wrap][1]     = 0;

  1651.             s->current_picture.motion_val[0][xy + wrap + 1][0] = 0;

  1652.             s->current_picture.motion_val[0][xy + wrap + 1][1] = 0;

  1653.             v->luma_mv[s->mb_x][0] = v->luma_mv[s->mb_x][1] = 0;

  1654.             s->current_picture.motion_val[1][xy + 1][0]        = 0;

  1655.             s->current_picture.motion_val[1][xy + 1][1]        = 0;

  1656.             s->current_picture.motion_val[1][xy + wrap][0]     = 0;

  1657.             s->current_picture.motion_val[1][xy + wrap][1]     = 0;

  1658.             s->current_picture.motion_val[1][xy + wrap + 1][0] = 0;

  1659.             s->current_picture.motion_val[1][xy + wrap + 1][1] = 0;

  1660.         }

  1661.         return;

  1662.     }

  1663. 

  1664.     off = ((n == 0) || (n == 1)) ? 1 : -1;

  1665.     /* predict A */

  1666.     if (s->mb_x || (n == 1) || (n == 3)) {

  1667.         if ((v->blk_mv_type[xy]) // current block (MB) has a field MV

  1668.             || (!v->blk_mv_type[xy] && !v->blk_mv_type[xy - 1])) { // or both have frame MV

  1669.             A[0] = s->current_picture.motion_val[dir][xy - 1][0];

  1670.             A[1] = s->current_picture.motion_val[dir][xy - 1][1];

  1671.             a_valid = 1;

  1672.         } else { // current block has frame mv and cand. has field MV (so average)

  1673.             A[0] = (s->current_picture.motion_val[dir][xy - 1][0]

  1674.                     + s->current_picture.motion_val[dir][xy - 1 + off * wrap][0] + 1) >> 1;

  1675.             A[1] = (s->current_picture.motion_val[dir][xy - 1][1]

  1676.                     + s->current_picture.motion_val[dir][xy - 1 + off * wrap][1] + 1) >> 1;

  1677.             a_valid = 1;

  1678.         }

  1679.         if (!(n & 1) && v->is_intra[s->mb_x - 1]) {

  1680.             a_valid = 0;

  1681.             A[0] = A[1] = 0;

  1682.         }

  1683.     } else

  1684.         A[0] = A[1] = 0;

  1685.     /* Predict B and C */

  1686.     B[0] = B[1] = C[0] = C[1] = 0;

  1687.     if (n == 0 || n == 1 || v->blk_mv_type[xy]) {

  1688.         if (!s->first_slice_line) {

  1689.             if (!v->is_intra[s->mb_x - s->mb_stride]) {

  1690.                 b_valid = 1;

  1691.                 n_adj   = n | 2;

  1692.                 pos_b   = s->block_index[n_adj] - 2 * wrap;

  1693.                 if (v->blk_mv_type[pos_b] && v->blk_mv_type[xy]) {

  1694.                     n_adj = (n & 2) | (n & 1);

  1695.                 }

  1696.                 B[0] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap][0];

  1697.                 B[1] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap][1];

  1698.                 if (v->blk_mv_type[pos_b] && !v->blk_mv_type[xy]) {

  1699.                     B[0] = (B[0] + s->current_picture.motion_val[dir][s->block_index[n_adj ^ 2] - 2 * wrap][0] + 1) >> 1;

  1700.                     B[1] = (B[1] + s->current_picture.motion_val[dir][s->block_index[n_adj ^ 2] - 2 * wrap][1] + 1) >> 1;

  1701.                 }

  1702.             }

  1703.             if (s->mb_width > 1) {

  1704.                 if (!v->is_intra[s->mb_x - s->mb_stride + 1]) {

  1705.                     c_valid = 1;

  1706.                     n_adj   = 2;

  1707.                     pos_c   = s->block_index[2] - 2 * wrap + 2;

  1708.                     if (v->blk_mv_type[pos_c] && v->blk_mv_type[xy]) {

  1709.                         n_adj = n & 2;

  1710.                     }

  1711.                     C[0] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap + 2][0];

  1712.                     C[1] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap + 2][1];

  1713.                     if (v->blk_mv_type[pos_c] && !v->blk_mv_type[xy]) {

  1714.                         C[0] = (1 + C[0] + (s->current_picture.motion_val[dir][s->block_index[n_adj ^ 2] - 2 * wrap + 2][0])) >> 1;

  1715.                         C[1] = (1 + C[1] + (s->current_picture.motion_val[dir][s->block_index[n_adj ^ 2] - 2 * wrap + 2][1])) >> 1;

  1716.                     }

  1717.                     if (s->mb_x == s->mb_width - 1) {

  1718.                         if (!v->is_intra[s->mb_x - s->mb_stride - 1]) {

  1719.                             c_valid = 1;

  1720.                             n_adj   = 3;

  1721.                             pos_c   = s->block_index[3] - 2 * wrap - 2;

  1722.                             if (v->blk_mv_type[pos_c] && v->blk_mv_type[xy]) {

  1723.                                 n_adj = n | 1;

  1724.                             }

  1725.                             C[0] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap - 2][0];

  1726.                             C[1] = s->current_picture.motion_val[dir][s->block_index[n_adj] - 2 * wrap - 2][1];

  1727.                             if (v->blk_mv_type[pos_c] && !v->blk_mv_type[xy]) {

  1728.                                 C[0] = (1 + C[0] + s->current_picture.motion_val[dir][s->block_index[1] - 2 * wrap - 2][0]) >> 1;

  1729.                                 C[1] = (1 + C[1] + s->current_picture.motion_val[dir][s->block_index[1] - 2 * wrap - 2][1]) >> 1;

  1730.                             }

  1731.                         } else

  1732.                             c_valid = 0;

  1733.                     }

  1734.                 }

  1735.             }

  1736.         }

  1737.     } else {

  1738.         pos_b   = s->block_index[1];

  1739.         b_valid = 1;

  1740.         B[0]    = s->current_picture.motion_val[dir][pos_b][0];

  1741.         B[1]    = s->current_picture.motion_val[dir][pos_b][1];

  1742.         pos_c   = s->block_index[0];

  1743.         c_valid = 1;

  1744.         C[0]    = s->current_picture.motion_val[dir][pos_c][0];

  1745.         C[1]    = s->current_picture.motion_val[dir][pos_c][1];

  1746.     }

  1747. 

  1748.     total_valid = a_valid + b_valid + c_valid;

  1749.     // check if predictor A is out of bounds

  1750.     if (!s->mb_x && !(n == 1 || n == 3)) {

  1751.         A[0] = A[1] = 0;

  1752.     }

  1753.     // check if predictor B is out of bounds

  1754.     if ((s->first_slice_line && v->blk_mv_type[xy]) || (s->first_slice_line && !(n & 2))) {

  1755.         B[0] = B[1] = C[0] = C[1] = 0;

  1756.     }

  1757.     if (!v->blk_mv_type[xy]) {

  1758.         if (s->mb_width == 1) {

  1759.             px = B[0];

  1760.             py = B[1];

  1761.         } else {

  1762.             if (total_valid >= 2) {

  1763.                 px = mid_pred(A[0], B[0], C[0]);

  1764.                 py = mid_pred(A[1], B[1], C[1]);

  1765.             } else if (total_valid) {

  1766.                 if (a_valid) { px = A[0]; py = A[1]; }

  1767.                 if (b_valid) { px = B[0]; py = B[1]; }

  1768.                 if (c_valid) { px = C[0]; py = C[1]; }

  1769.             } else

  1770.                 px = py = 0;

  1771.         }

  1772.     } else {

  1773.         if (a_valid)

  1774.             field_a = (A[1] & 4) ? 1 : 0;

  1775.         else

  1776.             field_a = 0;

  1777.         if (b_valid)

  1778.             field_b = (B[1] & 4) ? 1 : 0;

  1779.         else

  1780.             field_b = 0;

  1781.         if (c_valid)

  1782.             field_c = (C[1] & 4) ? 1 : 0;

  1783.         else

  1784.             field_c = 0;

  1785. 

  1786.         num_oppfield  = field_a + field_b + field_c;

  1787.         num_samefield = total_valid - num_oppfield;

  1788.         if (total_valid == 3) {

  1789.             if ((num_samefield == 3) || (num_oppfield == 3)) {

  1790.                 px = mid_pred(A[0], B[0], C[0]);

  1791.                 py = mid_pred(A[1], B[1], C[1]);

  1792.             } else if (num_samefield >= num_oppfield) {

  1793.                 /* take one MV from same field set depending on priority

  1794.                 the check for B may not be necessary */

  1795.                 px = !field_a ? A[0] : B[0];

  1796.                 py = !field_a ? A[1] : B[1];

  1797.             } else {

  1798.                 px =  field_a ? A[0] : B[0];

  1799.                 py =  field_a ? A[1] : B[1];

  1800.             }

  1801.         } else if (total_valid == 2) {

  1802.             if (num_samefield >= num_oppfield) {

  1803.                 if (!field_a && a_valid) {

  1804.                     px = A[0];

  1805.                     py = A[1];

  1806.                 } else if (!field_b && b_valid) {

  1807.                     px = B[0];

  1808.                     py = B[1];

  1809.                 } else if (c_valid) {

  1810.                     px = C[0];

  1811.                     py = C[1];

  1812.                 } else px = py = 0;

  1813.             } else {

  1814.                 if (field_a && a_valid) {

  1815.                     px = A[0];

  1816.                     py = A[1];

  1817.                 } else if (field_b && b_valid) {

  1818.                     px = B[0];

  1819.                     py = B[1];

  1820.                 } else if (c_valid) {

  1821.                     px = C[0];

  1822.                     py = C[1];

  1823.                 }

  1824.             }

  1825.         } else if (total_valid == 1) {

  1826.             px = (a_valid) ? A[0] : ((b_valid) ? B[0] : C[0]);

  1827.             py = (a_valid) ? A[1] : ((b_valid) ? B[1] : C[1]);

  1828.         } else

  1829.             px = py = 0;

  1830.     }

  1831. 

  1832.     /* store MV using signed modulus of MV range defined in 4.11 */

  1833.     s->mv[dir][n][0] = s->current_picture.motion_val[dir][xy][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x;

  1834.     s->mv[dir][n][1] = s->current_picture.motion_val[dir][xy][1] = ((py + dmv_y + r_y) & ((r_y << 1) - 1)) - r_y;

  1835.     if (mvn == 1) { /* duplicate motion data for 1-MV block */

  1836.         s->current_picture.motion_val[dir][xy +    1    ][0] = s->current_picture.motion_val[dir][xy][0];

  1837.         s->current_picture.motion_val[dir][xy +    1    ][1] = s->current_picture.motion_val[dir][xy][1];

  1838.         s->current_picture.motion_val[dir][xy + wrap    ][0] = s->current_picture.motion_val[dir][xy][0];

  1839.         s->current_picture.motion_val[dir][xy + wrap    ][1] = s->current_picture.motion_val[dir][xy][1];

  1840.         s->current_picture.motion_val[dir][xy + wrap + 1][0] = s->current_picture.motion_val[dir][xy][0];

  1841.         s->current_picture.motion_val[dir][xy + wrap + 1][1] = s->current_picture.motion_val[dir][xy][1];

  1842.     } else if (mvn == 2) { /* duplicate motion data for 2-Field MV block */

  1843.         s->current_picture.motion_val[dir][xy + 1][0] = s->current_picture.motion_val[dir][xy][0];

  1844.         s->current_picture.motion_val[dir][xy + 1][1] = s->current_picture.motion_val[dir][xy][1];

  1845.         s->mv[dir][n + 1][0] = s->mv[dir][n][0];

  1846.         s->mv[dir][n + 1][1] = s->mv[dir][n][1];

  1847.     }

  1848. }

  1849. 

  1850. /** Motion compensation for direct or interpolated blocks in B-frames

  1851.  */

  1852. static void vc1_interp_mc(VC1Context *v)

  1853. {

  1854.     MpegEncContext *s = &v->s;

  1855.     H264ChromaContext *h264chroma = &v->h264chroma;

  1856.     uint8_t *srcY, *srcU, *srcV;

  1857.     int dxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;

  1858.     int off, off_uv;

  1859.     int v_edge_pos = s->v_edge_pos >> v->field_mode;

  1860.     int use_ic = v->next_use_ic;

  1861. 

  1862.     if (!v->field_mode && !v->s.next_picture.f.data[0])

  1863.         return;

  1864. 

  1865.     mx   = s->mv[1][0][0];

  1866.     my   = s->mv[1][0][1];

  1867.     uvmx = (mx + ((mx & 3) == 3)) >> 1;

  1868.     uvmy = (my + ((my & 3) == 3)) >> 1;

  1869.     if (v->field_mode) {

  1870.         if (v->cur_field_type != v->ref_field_type[1])

  1871.             my   = my   - 2 + 4 * v->cur_field_type;

  1872.             uvmy = uvmy - 2 + 4 * v->cur_field_type;

  1873.     }

  1874.     if (v->fastuvmc) {

  1875.         uvmx = uvmx + ((uvmx < 0) ? -(uvmx & 1) : (uvmx & 1));

  1876.         uvmy = uvmy + ((uvmy < 0) ? -(uvmy & 1) : (uvmy & 1));

  1877.     }

  1878.     srcY = s->next_picture.f.data[0];

  1879.     srcU = s->next_picture.f.data[1];

  1880.     srcV = s->next_picture.f.data[2];

  1881. 

  1882.     src_x   = s->mb_x * 16 + (mx   >> 2);

  1883.     src_y   = s->mb_y * 16 + (my   >> 2);

  1884.     uvsrc_x = s->mb_x *  8 + (uvmx >> 2);

  1885.     uvsrc_y = s->mb_y *  8 + (uvmy >> 2);

  1886. 

  1887.     if (v->profile != PROFILE_ADVANCED) {

  1888.         src_x   = av_clip(  src_x, -16, s->mb_width  * 16);

  1889.         src_y   = av_clip(  src_y, -16, s->mb_height * 16);

  1890.         uvsrc_x = av_clip(uvsrc_x,  -8, s->mb_width  *  8);

  1891.         uvsrc_y = av_clip(uvsrc_y,  -8, s->mb_height *  8);

  1892.     } else {

  1893.         src_x   = av_clip(  src_x, -17, s->avctx->coded_width);

  1894.         src_y   = av_clip(  src_y, -18, s->avctx->coded_height + 1);

  1895.         uvsrc_x = av_clip(uvsrc_x,  -8, s->avctx->coded_width  >> 1);

  1896.         uvsrc_y = av_clip(uvsrc_y,  -8, s->avctx->coded_height >> 1);

  1897.     }

  1898. 

  1899.     srcY += src_y   * s->linesize   + src_x;

  1900.     srcU += uvsrc_y * s->uvlinesize + uvsrc_x;

  1901.     srcV += uvsrc_y * s->uvlinesize + uvsrc_x;

  1902. 

  1903.     if (v->field_mode && v->ref_field_type[1]) {

  1904.         srcY += s->current_picture_ptr->f.linesize[0];

  1905.         srcU += s->current_picture_ptr->f.linesize[1];

  1906.         srcV += s->current_picture_ptr->f.linesize[2];

  1907.     }

  1908. 

  1909.     /* for grayscale we should not try to read from unknown area */

  1910.     if (s->flags & CODEC_FLAG_GRAY) {

  1911.         srcU = s->edge_emu_buffer + 18 * s->linesize;

  1912.         srcV = s->edge_emu_buffer + 18 * s->linesize;

  1913.     }

  1914. 

  1915.     if (v->rangeredfrm || s->h_edge_pos < 22 || v_edge_pos < 22 || use_ic

  1916.         || (unsigned)(src_x - 1) > s->h_edge_pos - (mx & 3) - 16 - 3

  1917.         || (unsigned)(src_y - 1) > v_edge_pos    - (my & 3) - 16 - 3) {

  1918.         uint8_t *uvbuf = s->edge_emu_buffer + 19 * s->linesize;

  1919. 

  1920.         srcY -= s->mspel * (1 + s->linesize);

  1921.         s->vdsp.emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize,

  1922.                                  17 + s->mspel * 2, 17 + s->mspel * 2,

  1923.                                  src_x - s->mspel, src_y - s->mspel,

  1924.                                  s->h_edge_pos, v_edge_pos);

  1925.         srcY = s->edge_emu_buffer;

  1926.         s->vdsp.emulated_edge_mc(uvbuf     , srcU, s->uvlinesize, 8 + 1, 8 + 1,

  1927.                                  uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1);

  1928.         s->vdsp.emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8 + 1, 8 + 1,

  1929.                                  uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1);

  1930.         srcU = uvbuf;

  1931.         srcV = uvbuf + 16;

  1932.         /* if we deal with range reduction we need to scale source blocks */

  1933.         if (v->rangeredfrm) {

  1934.             int i, j;

  1935.             uint8_t *src, *src2;

  1936. 

  1937.             src = srcY;

  1938.             for (j = 0; j < 17 + s->mspel * 2; j++) {

  1939.                 for (i = 0; i < 17 + s->mspel * 2; i++)

  1940.                     src[i] = ((src[i] - 128) >> 1) + 128;

  1941.                 src += s->linesize;

  1942.             }

  1943.             src = srcU;

  1944.             src2 = srcV;

  1945.             for (j = 0; j < 9; j++) {

  1946.                 for (i = 0; i < 9; i++) {

  1947.                     src[i]  = ((src[i]  - 128) >> 1) + 128;

  1948.                     src2[i] = ((src2[i] - 128) >> 1) + 128;

  1949.                 }

  1950.                 src  += s->uvlinesize;

  1951.                 src2 += s->uvlinesize;

  1952.             }

  1953.         }

  1954. 

  1955.         if (use_ic) {

  1956.             const uint8_t *luty = v->next_luty [v->ref_field_type[1]];

  1957.             const uint8_t *lutuv= v->next_lutuv[v->ref_field_type[1]];

  1958.             int i, j;

  1959.             uint8_t *src, *src2;

  1960. 

  1961.             src = srcY;

  1962.             for (j = 0; j < 17 + s->mspel * 2; j++) {

  1963.                 for (i = 0; i < 17 + s->mspel * 2; i++)

  1964.                     src[i] = luty[src[i]];

  1965.                 src += s->linesize;

  1966.             }

  1967.             src  = srcU;

  1968.             src2 = srcV;

  1969.             for (j = 0; j < 9; j++) {

  1970.                 for (i = 0; i < 9; i++) {

  1971.                     src[i]  = lutuv[src[i]];

  1972.                     src2[i] = lutuv[src2[i]];

  1973.                 }

  1974.                 src  += s->uvlinesize;

  1975.                 src2 += s->uvlinesize;

  1976.             }

  1977.         }

  1978.         srcY += s->mspel * (1 + s->linesize);

  1979.     }

  1980. 

  1981.     off    = 0;

  1982.     off_uv = 0;

  1983. 

  1984.     if (s->mspel) {

  1985.         dxy = ((my & 3) << 2) | (mx & 3);

  1986.         v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off    , srcY    , s->linesize, v->rnd);

  1987.         v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8, srcY + 8, s->linesize, v->rnd);

  1988.         srcY += s->linesize * 8;

  1989.         v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8 * s->linesize    , srcY    , s->linesize, v->rnd);

  1990.         v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd);

  1991.     } else { // hpel mc

  1992.         dxy = (my & 2) | ((mx & 2) >> 1);

  1993. 

  1994.         if (!v->rnd)

  1995.             s->hdsp.avg_pixels_tab[0][dxy](s->dest[0] + off, srcY, s->linesize, 16);

  1996.         else

  1997.             s->hdsp.avg_no_rnd_pixels_tab[dxy](s->dest[0] + off, srcY, s->linesize, 16);

  1998.     }

  1999. 

  2000.     if (s->flags & CODEC_FLAG_GRAY) return;

  2001.     /* Chroma MC always uses qpel blilinear */

  2002.     uvmx = (uvmx & 3) << 1;

  2003.     uvmy = (uvmy & 3) << 1;

  2004.     if (!v->rnd) {

  2005.         h264chroma->avg_h264_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy);

  2006.         h264chroma->avg_h264_chroma_pixels_tab[0](s->dest[2] + off_uv, srcV, s->uvlinesize, 8, uvmx, uvmy);

  2007.     } else {

  2008.         v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy);

  2009.         v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2] + off_uv, srcV, s->uvlinesize, 8, uvmx, uvmy);

  2010.     }

  2011. }

  2012. 

  2013. static av_always_inline int scale_mv(int value, int bfrac, int inv, int qs)

  2014. {

  2015.     int n = bfrac;

  2016. 

  2017. #if B_FRACTION_DEN==256

  2018.     if (inv)

  2019.         n -= 256;

  2020.     if (!qs)

  2021.         return 2 * ((value * n + 255) >> 9);

  2022.     return (value * n + 128) >> 8;

  2023. #else

  2024.     if (inv)

  2025.         n -= B_FRACTION_DEN;

  2026.     if (!qs)

  2027.         return 2 * ((value * n + B_FRACTION_DEN - 1) / (2 * B_FRACTION_DEN));

  2028.     return (value * n + B_FRACTION_DEN/2) / B_FRACTION_DEN;

  2029. #endif

  2030. }

  2031. 

  2032. /** Reconstruct motion vector for B-frame and do motion compensation

  2033.  */

  2034. static inline void vc1_b_mc(VC1Context *v, int dmv_x[2], int dmv_y[2],

  2035.                             int direct, int mode)

  2036. {

  2037.     int use_ic = v->next_use_ic || v->curr_use_ic || v->last_use_ic;

  2038.     if (use_ic) {

  2039.         v->mv_mode2 = v->mv_mode;

  2040.         v->mv_mode  = MV_PMODE_INTENSITY_COMP;

  2041.     }

  2042.     if (direct) {

  2043.         vc1_mc_1mv(v, 0);

  2044.         vc1_interp_mc(v);

  2045.         if (use_ic)

  2046.             v->mv_mode = v->mv_mode2;

  2047.         return;

  2048.     }

  2049.     if (mode == BMV_TYPE_INTERPOLATED) {

  2050.         vc1_mc_1mv(v, 0);

  2051.         vc1_interp_mc(v);

  2052.         if (use_ic)

  2053.             v->mv_mode = v->mv_mode2;

  2054.         return;

  2055.     }

  2056. 

  2057.     if (use_ic && (mode == BMV_TYPE_BACKWARD))

  2058.         v->mv_mode = v->mv_mode2;

  2059.     vc1_mc_1mv(v, (mode == BMV_TYPE_BACKWARD));

  2060.     if (use_ic)

  2061.         v->mv_mode = v->mv_mode2;

  2062. }

  2063. 

  2064. static inline void vc1_pred_b_mv(VC1Context *v, int dmv_x[2], int dmv_y[2],

  2065.                                  int direct, int mvtype)

  2066. {

  2067.     MpegEncContext *s = &v->s;

  2068.     int xy, wrap, off = 0;

  2069.     int16_t *A, *B, *C;

  2070.     int px, py;

  2071.     int sum;

  2072.     int r_x, r_y;

  2073.     const uint8_t *is_intra = v->mb_type[0];

  2074. 

  2075.     r_x = v->range_x;

  2076.     r_y = v->range_y;

  2077.     /* scale MV difference to be quad-pel */

  2078.     dmv_x[0] <<= 1 - s->quarter_sample;

  2079.     dmv_y[0] <<= 1 - s->quarter_sample;

  2080.     dmv_x[1] <<= 1 - s->quarter_sample;

  2081.     dmv_y[1] <<= 1 - s->quarter_sample;

  2082. 

  2083.     wrap = s->b8_stride;

  2084.     xy = s->block_index[0];

  2085. 

  2086.     if (s->mb_intra) {

  2087.         s->current_picture.motion_val[0][xy + v->blocks_off][0] =

  2088.         s->current_picture.motion_val[0][xy + v->blocks_off][1] =

  2089.         s->current_picture.motion_val[1][xy + v->blocks_off][0] =

  2090.         s->current_picture.motion_val[1][xy + v->blocks_off][1] = 0;

  2091.         return;

  2092.     }

  2093.     if (!v->field_mode) {

  2094.         s->mv[0][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 0, s->quarter_sample);

  2095.         s->mv[0][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 0, s->quarter_sample);

  2096.         s->mv[1][0][0] = scale_mv(s->next_picture.motion_val[1][xy][0], v->bfraction, 1, s->quarter_sample);

  2097.         s->mv[1][0][1] = scale_mv(s->next_picture.motion_val[1][xy][1], v->bfraction, 1, s->quarter_sample);

  2098. 

  2099.         /* Pullback predicted motion vectors as specified in 8.4.5.4 */

  2100.         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));

  2101.         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));

  2102.         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));

  2103.         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));

  2104.     }

  2105.     if (direct) {

  2106.         s->current_picture.motion_val[0][xy + v->blocks_off][0] = s->mv[0][0][0];

  2107.         s->current_picture.motion_val[0][xy + v->blocks_off][1] = s->mv[0][0][1];

  2108.         s->current_picture.motion_val[1][xy + v->blocks_off][0] = s->mv[1][0][0];

  2109.         s->current_picture.motion_val[1][xy + v->blocks_off][1] = s->mv[1][0][1];

  2110.         return;

  2111.     }

  2112. 

  2113.     if ((mvtype == BMV_TYPE_FORWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {

  2114.         C   = s->current_picture.motion_val[0][xy - 2];

  2115.         A   = s->current_picture.motion_val[0][xy - wrap * 2];

  2116.         off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;

  2117.         B   = s->current_picture.motion_val[0][xy - wrap * 2 + off];

  2118. 

  2119.         if (!s->mb_x) C[0] = C[1] = 0;

  2120.         if (!s->first_slice_line) { // predictor A is not out of bounds

  2121.             if (s->mb_width == 1) {

  2122.                 px = A[0];

  2123.                 py = A[1];

  2124.             } else {

  2125.                 px = mid_pred(A[0], B[0], C[0]);

  2126.                 py = mid_pred(A[1], B[1], C[1]);

  2127.             }

  2128.         } else if (s->mb_x) { // predictor C is not out of bounds

  2129.             px = C[0];

  2130.             py = C[1];

  2131.         } else {

  2132.             px = py = 0;

  2133.         }

  2134.         /* Pullback MV as specified in 8.3.5.3.4 */

  2135.         {

  2136.             int qx, qy, X, Y;

  2137.             if (v->profile < PROFILE_ADVANCED) {

  2138.                 qx = (s->mb_x << 5);

  2139.                 qy = (s->mb_y << 5);

  2140.                 X  = (s->mb_width  << 5) - 4;

  2141.                 Y  = (s->mb_height << 5) - 4;

  2142.                 if (qx + px < -28) px = -28 - qx;

  2143.                 if (qy + py < -28) py = -28 - qy;

  2144.                 if (qx + px > X) px = X - qx;

  2145.                 if (qy + py > Y) py = Y - qy;

  2146.             } else {

  2147.                 qx = (s->mb_x << 6);

  2148.                 qy = (s->mb_y << 6);

  2149.                 X  = (s->mb_width  << 6) - 4;

  2150.                 Y  = (s->mb_height << 6) - 4;

  2151.                 if (qx + px < -60) px = -60 - qx;

  2152.                 if (qy + py < -60) py = -60 - qy;

  2153.                 if (qx + px > X) px = X - qx;

  2154.                 if (qy + py > Y) py = Y - qy;

  2155.             }

  2156.         }

  2157.         /* Calculate hybrid prediction as specified in 8.3.5.3.5 */

  2158.         if (0 && !s->first_slice_line && s->mb_x) {

  2159.             if (is_intra[xy - wrap])

  2160.                 sum = FFABS(px) + FFABS(py);

  2161.             else

  2162.                 sum = FFABS(px - A[0]) + FFABS(py - A[1]);

  2163.             if (sum > 32) {

  2164.                 if (get_bits1(&s->gb)) {

  2165.                     px = A[0];

  2166.                     py = A[1];

  2167.                 } else {

  2168.                     px = C[0];

  2169.                     py = C[1];

  2170.                 }

  2171.             } else {

  2172.                 if (is_intra[xy - 2])

  2173.                     sum = FFABS(px) + FFABS(py);

  2174.                 else

  2175.                     sum = FFABS(px - C[0]) + FFABS(py - C[1]);

  2176.                 if (sum > 32) {

  2177.                     if (get_bits1(&s->gb)) {

  2178.                         px = A[0];

  2179.                         py = A[1];

  2180.                     } else {

  2181.                         px = C[0];

  2182.                         py = C[1];

  2183.                     }

  2184.                 }

  2185.             }

  2186.         }

  2187.         /* store MV using signed modulus of MV range defined in 4.11 */

  2188.         s->mv[0][0][0] = ((px + dmv_x[0] + r_x) & ((r_x << 1) - 1)) - r_x;

  2189.         s->mv[0][0][1] = ((py + dmv_y[0] + r_y) & ((r_y << 1) - 1)) - r_y;

  2190.     }

  2191.     if ((mvtype == BMV_TYPE_BACKWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {

  2192.         C   = s->current_picture.motion_val[1][xy - 2];

  2193.         A   = s->current_picture.motion_val[1][xy - wrap * 2];

  2194.         off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;

  2195.         B   = s->current_picture.motion_val[1][xy - wrap * 2 + off];

  2196. 

  2197.         if (!s->mb_x)

  2198.             C[0] = C[1] = 0;

  2199.         if (!s->first_slice_line) { // predictor A is not out of bounds

  2200.             if (s->mb_width == 1) {

  2201.                 px = A[0];

  2202.                 py = A[1];

  2203.             } else {

  2204.                 px = mid_pred(A[0], B[0], C[0]);

  2205.                 py = mid_pred(A[1], B[1], C[1]);

  2206.             }

  2207.         } else if (s->mb_x) { // predictor C is not out of bounds

  2208.             px = C[0];

  2209.             py = C[1];

  2210.         } else {

  2211.             px = py = 0;

  2212.         }

  2213.         /* Pullback MV as specified in 8.3.5.3.4 */

  2214.         {

  2215.             int qx, qy, X, Y;

  2216.             if (v->profile < PROFILE_ADVANCED) {

  2217.                 qx = (s->mb_x << 5);

  2218.                 qy = (s->mb_y << 5);

  2219.                 X  = (s->mb_width  << 5) - 4;

  2220.                 Y  = (s->mb_height << 5) - 4;

  2221.                 if (qx + px < -28) px = -28 - qx;

  2222.                 if (qy + py < -28) py = -28 - qy;

  2223.                 if (qx + px > X) px = X - qx;

  2224.                 if (qy + py > Y) py = Y - qy;

  2225.             } else {

  2226.                 qx = (s->mb_x << 6);

  2227.                 qy = (s->mb_y << 6);

  2228.                 X  = (s->mb_width  << 6) - 4;

  2229.                 Y  = (s->mb_height << 6) - 4;

  2230.                 if (qx + px < -60) px = -60 - qx;

  2231.                 if (qy + py < -60) py = -60 - qy;

  2232.                 if (qx + px > X) px = X - qx;

  2233.                 if (qy + py > Y) py = Y - qy;

  2234.             }

  2235.         }

  2236.         /* Calculate hybrid prediction as specified in 8.3.5.3.5 */

  2237.         if (0 && !s->first_slice_line && s->mb_x) {

  2238.             if (is_intra[xy - wrap])

  2239.                 sum = FFABS(px) + FFABS(py);

  2240.             else

  2241.                 sum = FFABS(px - A[0]) + FFABS(py - A[1]);

  2242.             if (sum > 32) {

  2243.                 if (get_bits1(&s->gb)) {

  2244.                     px = A[0];

  2245.                     py = A[1];

  2246.                 } else {

  2247.                     px = C[0];

  2248.                     py = C[1];

  2249.                 }

  2250.             } else {

  2251.                 if (is_intra[xy - 2])

  2252.                     sum = FFABS(px) + FFABS(py);

  2253.                 else

  2254.                     sum = FFABS(px - C[0]) + FFABS(py - C[1]);

  2255.                 if (sum > 32) {

  2256.                     if (get_bits1(&s->gb)) {

  2257.                         px = A[0];

  2258.                         py = A[1];

  2259.                     } else {

  2260.                         px = C[0];

  2261.                         py = C[1];

  2262.                     }

  2263.                 }

  2264.             }

  2265.         }

  2266.         /* store MV using signed modulus of MV range defined in 4.11 */

  2267. 

  2268.         s->mv[1][0][0] = ((px + dmv_x[1] + r_x) & ((r_x << 1) - 1)) - r_x;

  2269.         s->mv[1][0][1] = ((py + dmv_y[1] + r_y) & ((r_y << 1) - 1)) - r_y;

  2270.     }

  2271.     s->current_picture.motion_val[0][xy][0] = s->mv[0][0][0];

  2272.     s->current_picture.motion_val[0][xy][1] = s->mv[0][0][1];

  2273.     s->current_picture.motion_val[1][xy][0] = s->mv[1][0][0];

  2274.     s->current_picture.motion_val[1][xy][1] = s->mv[1][0][1];

  2275. }

  2276. 

  2277. static inline void vc1_pred_b_mv_intfi(VC1Context *v, int n, int *dmv_x, int *dmv_y, int mv1, int *pred_flag)

  2278. {

  2279.     int dir = (v->bmvtype == BMV_TYPE_BACKWARD) ? 1 : 0;

  2280.     MpegEncContext *s = &v->s;

  2281.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  2282. 

  2283.     if (v->bmvtype == BMV_TYPE_DIRECT) {

  2284.         int total_opp, k, f;

  2285.         if (s->next_picture.mb_type[mb_pos + v->mb_off] != MB_TYPE_INTRA) {

  2286.             s->mv[0][0][0] = scale_mv(s->next_picture.motion_val[1][s->block_index[0] + v->blocks_off][0],

  2287.                                       v->bfraction, 0, s->quarter_sample);

  2288.             s->mv[0][0][1] = scale_mv(s->next_picture.motion_val[1][s->block_index[0] + v->blocks_off][1],

  2289.                                       v->bfraction, 0, s->quarter_sample);

  2290.             s->mv[1][0][0] = scale_mv(s->next_picture.motion_val[1][s->block_index[0] + v->blocks_off][0],

  2291.                                       v->bfraction, 1, s->quarter_sample);

  2292.             s->mv[1][0][1] = scale_mv(s->next_picture.motion_val[1][s->block_index[0] + v->blocks_off][1],

  2293.                                       v->bfraction, 1, s->quarter_sample);

  2294. 

  2295.             total_opp = v->mv_f_next[0][s->block_index[0] + v->blocks_off]

  2296.                       + v->mv_f_next[0][s->block_index[1] + v->blocks_off]

  2297.                       + v->mv_f_next[0][s->block_index[2] + v->blocks_off]

  2298.                       + v->mv_f_next[0][s->block_index[3] + v->blocks_off];

  2299.             f = (total_opp > 2) ? 1 : 0;

  2300.         } else {

  2301.             s->mv[0][0][0] = s->mv[0][0][1] = 0;

  2302.             s->mv[1][0][0] = s->mv[1][0][1] = 0;

  2303.             f = 0;

  2304.         }

  2305.         v->ref_field_type[0] = v->ref_field_type[1] = v->cur_field_type ^ f;

  2306.         for (k = 0; k < 4; k++) {

  2307.             s->current_picture.motion_val[0][s->block_index[k] + v->blocks_off][0] = s->mv[0][0][0];

  2308.             s->current_picture.motion_val[0][s->block_index[k] + v->blocks_off][1] = s->mv[0][0][1];

  2309.             s->current_picture.motion_val[1][s->block_index[k] + v->blocks_off][0] = s->mv[1][0][0];

  2310.             s->current_picture.motion_val[1][s->block_index[k] + v->blocks_off][1] = s->mv[1][0][1];

  2311.             v->mv_f[0][s->block_index[k] + v->blocks_off] = f;

  2312.             v->mv_f[1][s->block_index[k] + v->blocks_off] = f;

  2313.         }

  2314.         return;

  2315.     }

  2316.     if (v->bmvtype == BMV_TYPE_INTERPOLATED) {

  2317.         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);

  2318.         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);

  2319.         return;

  2320.     }

  2321.     if (dir) { // backward

  2322.         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);

  2323.         if (n == 3 || mv1) {

  2324.             vc1_pred_mv(v, 0, dmv_x[0], dmv_y[0],   1, v->range_x, v->range_y, v->mb_type[0], 0, 0);

  2325.         }

  2326.     } else { // forward

  2327.         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);

  2328.         if (n == 3 || mv1) {

  2329.             vc1_pred_mv(v, 0, dmv_x[1], dmv_y[1],   1, v->range_x, v->range_y, v->mb_type[0], 0, 1);

  2330.         }

  2331.     }

  2332. }

  2333. 

  2334. /** Get predicted DC value for I-frames only

  2335.  * prediction dir: left=0, top=1

  2336.  * @param s MpegEncContext

  2337.  * @param overlap flag indicating that overlap filtering is used

  2338.  * @param pq integer part of picture quantizer

  2339.  * @param[in] n block index in the current MB

  2340.  * @param dc_val_ptr Pointer to DC predictor

  2341.  * @param dir_ptr Prediction direction for use in AC prediction

  2342.  */

  2343. static inline int vc1_i_pred_dc(MpegEncContext *s, int overlap, int pq, int n,

  2344.                                 int16_t **dc_val_ptr, int *dir_ptr)

  2345. {

  2346.     int a, b, c, wrap, pred, scale;

  2347.     int16_t *dc_val;

  2348.     static const uint16_t dcpred[32] = {

  2349.         -1, 1024,  512,  341,  256,  205,  171,  146,  128,

  2350.              114,  102,   93,   85,   79,   73,   68,   64,

  2351.               60,   57,   54,   51,   49,   47,   45,   43,

  2352.               41,   39,   38,   37,   35,   34,   33

  2353.     };

  2354. 

  2355.     /* find prediction - wmv3_dc_scale always used here in fact */

  2356.     if (n < 4) scale = s->y_dc_scale;

  2357.     else       scale = s->c_dc_scale;

  2358. 

  2359.     wrap   = s->block_wrap[n];

  2360.     dc_val = s->dc_val[0] + s->block_index[n];

  2361. 

  2362.     /* B A

  2363.      * C X

  2364.      */

  2365.     c = dc_val[ - 1];

  2366.     b = dc_val[ - 1 - wrap];

  2367.     a = dc_val[ - wrap];

  2368. 

  2369.     if (pq < 9 || !overlap) {

  2370.         /* Set outer values */

  2371.         if (s->first_slice_line && (n != 2 && n != 3))

  2372.             b = a = dcpred[scale];

  2373.         if (s->mb_x == 0 && (n != 1 && n != 3))

  2374.             b = c = dcpred[scale];

  2375.     } else {

  2376.         /* Set outer values */

  2377.         if (s->first_slice_line && (n != 2 && n != 3))

  2378.             b = a = 0;

  2379.         if (s->mb_x == 0 && (n != 1 && n != 3))

  2380.             b = c = 0;

  2381.     }

  2382. 

  2383.     if (abs(a - b) <= abs(b - c)) {

  2384.         pred     = c;

  2385.         *dir_ptr = 1; // left

  2386.     } else {

  2387.         pred     = a;

  2388.         *dir_ptr = 0; // top

  2389.     }

  2390. 

  2391.     /* update predictor */

  2392.     *dc_val_ptr = &dc_val[0];

  2393.     return pred;

  2394. }

  2395. 

  2396. 

  2397. /** Get predicted DC value

  2398.  * prediction dir: left=0, top=1

  2399.  * @param s MpegEncContext

  2400.  * @param overlap flag indicating that overlap filtering is used

  2401.  * @param pq integer part of picture quantizer

  2402.  * @param[in] n block index in the current MB

  2403.  * @param a_avail flag indicating top block availability

  2404.  * @param c_avail flag indicating left block availability

  2405.  * @param dc_val_ptr Pointer to DC predictor

  2406.  * @param dir_ptr Prediction direction for use in AC prediction

  2407.  */

  2408. static inline int vc1_pred_dc(MpegEncContext *s, int overlap, int pq, int n,

  2409.                               int a_avail, int c_avail,

  2410.                               int16_t **dc_val_ptr, int *dir_ptr)

  2411. {

  2412.     int a, b, c, wrap, pred;

  2413.     int16_t *dc_val;

  2414.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  2415.     int q1, q2 = 0;

  2416.     int dqscale_index;

  2417. 

  2418.     wrap = s->block_wrap[n];

  2419.     dc_val = s->dc_val[0] + s->block_index[n];

  2420. 

  2421.     /* B A

  2422.      * C X

  2423.      */

  2424.     c = dc_val[ - 1];

  2425.     b = dc_val[ - 1 - wrap];

  2426.     a = dc_val[ - wrap];

  2427.     /* scale predictors if needed */

  2428.     q1 = s->current_picture.qscale_table[mb_pos];

  2429.     dqscale_index = s->y_dc_scale_table[q1] - 1;

  2430.     if (dqscale_index < 0)

  2431.         return 0;

  2432.     if (c_avail && (n != 1 && n != 3)) {

  2433.         q2 = s->current_picture.qscale_table[mb_pos - 1];

  2434.         if (q2 && q2 != q1)

  2435.             c = (c * s->y_dc_scale_table[q2] * ff_vc1_dqscale[dqscale_index] + 0x20000) >> 18;

  2436.     }

  2437.     if (a_avail && (n != 2 && n != 3)) {

  2438.         q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];

  2439.         if (q2 && q2 != q1)

  2440.             a = (a * s->y_dc_scale_table[q2] * ff_vc1_dqscale[dqscale_index] + 0x20000) >> 18;

  2441.     }

  2442.     if (a_avail && c_avail && (n != 3)) {

  2443.         int off = mb_pos;

  2444.         if (n != 1)

  2445.             off--;

  2446.         if (n != 2)

  2447.             off -= s->mb_stride;

  2448.         q2 = s->current_picture.qscale_table[off];

  2449.         if (q2 && q2 != q1)

  2450.             b = (b * s->y_dc_scale_table[q2] * ff_vc1_dqscale[dqscale_index] + 0x20000) >> 18;

  2451.     }

  2452. 

  2453.     if (a_avail && c_avail) {

  2454.         if (abs(a - b) <= abs(b - c)) {

  2455.             pred     = c;

  2456.             *dir_ptr = 1; // left

  2457.         } else {

  2458.             pred     = a;

  2459.             *dir_ptr = 0; // top

  2460.         }

  2461.     } else if (a_avail) {

  2462.         pred     = a;

  2463.         *dir_ptr = 0; // top

  2464.     } else if (c_avail) {

  2465.         pred     = c;

  2466.         *dir_ptr = 1; // left

  2467.     } else {

  2468.         pred     = 0;

  2469.         *dir_ptr = 1; // left

  2470.     }

  2471. 

  2472.     /* update predictor */

  2473.     *dc_val_ptr = &dc_val[0];

  2474.     return pred;

  2475. }

  2476. 

  2477. /** @} */ // Block group

  2478. 

  2479. /**

  2480.  * @name VC1 Macroblock-level functions in Simple/Main Profiles

  2481.  * @see 7.1.4, p91 and 8.1.1.7, p(1)04

  2482.  * @{

  2483.  */

  2484. 

  2485. static inline int vc1_coded_block_pred(MpegEncContext * s, int n,

  2486.                                        uint8_t **coded_block_ptr)

  2487. {

  2488.     int xy, wrap, pred, a, b, c;

  2489. 

  2490.     xy   = s->block_index[n];

  2491.     wrap = s->b8_stride;

  2492. 

  2493.     /* B C

  2494.      * A X

  2495.      */

  2496.     a = s->coded_block[xy - 1       ];

  2497.     b = s->coded_block[xy - 1 - wrap];

  2498.     c = s->coded_block[xy     - wrap];

  2499. 

  2500.     if (b == c) {

  2501.         pred = a;

  2502.     } else {

  2503.         pred = c;

  2504.     }

  2505. 

  2506.     /* store value */

  2507.     *coded_block_ptr = &s->coded_block[xy];

  2508. 

  2509.     return pred;

  2510. }

  2511. 

  2512. /**

  2513.  * Decode one AC coefficient

  2514.  * @param v The VC1 context

  2515.  * @param last Last coefficient

  2516.  * @param skip How much zero coefficients to skip

  2517.  * @param value Decoded AC coefficient value

  2518.  * @param codingset set of VLC to decode data

  2519.  * @see 8.1.3.4

  2520.  */

  2521. static void vc1_decode_ac_coeff(VC1Context *v, int *last, int *skip,

  2522.                                 int *value, int codingset)

  2523. {

  2524.     GetBitContext *gb = &v->s.gb;

  2525.     int index, escape, run = 0, level = 0, lst = 0;

  2526. 

  2527.     index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);

  2528.     if (index != ff_vc1_ac_sizes[codingset] - 1) {

  2529.         run   = vc1_index_decode_table[codingset][index][0];

  2530.         level = vc1_index_decode_table[codingset][index][1];

  2531.         lst   = index >= vc1_last_decode_table[codingset] || get_bits_left(gb) < 0;

  2532.         if (get_bits1(gb))

  2533.             level = -level;

  2534.     } else {

  2535.         escape = decode210(gb);

  2536.         if (escape != 2) {

  2537.             index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);

  2538.             run   = vc1_index_decode_table[codingset][index][0];

  2539.             level = vc1_index_decode_table[codingset][index][1];

  2540.             lst   = index >= vc1_last_decode_table[codingset];

  2541.             if (escape == 0) {

  2542.                 if (lst)

  2543.                     level += vc1_last_delta_level_table[codingset][run];

  2544.                 else

  2545.                     level += vc1_delta_level_table[codingset][run];

  2546.             } else {

  2547.                 if (lst)

  2548.                     run += vc1_last_delta_run_table[codingset][level] + 1;

  2549.                 else

  2550.                     run += vc1_delta_run_table[codingset][level] + 1;

  2551.             }

  2552.             if (get_bits1(gb))

  2553.                 level = -level;

  2554.         } else {

  2555.             int sign;

  2556.             lst = get_bits1(gb);

  2557.             if (v->s.esc3_level_length == 0) {

  2558.                 if (v->pq < 8 || v->dquantfrm) { // table 59

  2559.                     v->s.esc3_level_length = get_bits(gb, 3);

  2560.                     if (!v->s.esc3_level_length)

  2561.                         v->s.esc3_level_length = get_bits(gb, 2) + 8;

  2562.                 } else { // table 60

  2563.                     v->s.esc3_level_length = get_unary(gb, 1, 6) + 2;

  2564.                 }

  2565.                 v->s.esc3_run_length = 3 + get_bits(gb, 2);

  2566.             }

  2567.             run   = get_bits(gb, v->s.esc3_run_length);

  2568.             sign  = get_bits1(gb);

  2569.             level = get_bits(gb, v->s.esc3_level_length);

  2570.             if (sign)

  2571.                 level = -level;

  2572.         }

  2573.     }

  2574. 

  2575.     *last  = lst;

  2576.     *skip  = run;

  2577.     *value = level;

  2578. }

  2579. 

  2580. /** Decode intra block in intra frames - should be faster than decode_intra_block

  2581.  * @param v VC1Context

  2582.  * @param block block to decode

  2583.  * @param[in] n subblock index

  2584.  * @param coded are AC coeffs present or not

  2585.  * @param codingset set of VLC to decode data

  2586.  */

  2587. static int vc1_decode_i_block(VC1Context *v, int16_t block[64], int n,

  2588.                               int coded, int codingset)

  2589. {

  2590.     GetBitContext *gb = &v->s.gb;

  2591.     MpegEncContext *s = &v->s;

  2592.     int dc_pred_dir = 0; /* Direction of the DC prediction used */

  2593.     int i;

  2594.     int16_t *dc_val;

  2595.     int16_t *ac_val, *ac_val2;

  2596.     int dcdiff;

  2597. 

  2598.     /* Get DC differential */

  2599.     if (n < 4) {

  2600.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2601.     } else {

  2602.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2603.     }

  2604.     if (dcdiff < 0) {

  2605.         av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");

  2606.         return -1;

  2607.     }

  2608.     if (dcdiff) {

  2609.         if (dcdiff == 119 /* ESC index value */) {

  2610.             /* TODO: Optimize */

  2611.             if (v->pq == 1)      dcdiff = get_bits(gb, 10);

  2612.             else if (v->pq == 2) dcdiff = get_bits(gb, 9);

  2613.             else                 dcdiff = get_bits(gb, 8);

  2614.         } else {

  2615.             if (v->pq == 1)

  2616.                 dcdiff = (dcdiff << 2) + get_bits(gb, 2) - 3;

  2617.             else if (v->pq == 2)

  2618.                 dcdiff = (dcdiff << 1) + get_bits1(gb)   - 1;

  2619.         }

  2620.         if (get_bits1(gb))

  2621.             dcdiff = -dcdiff;

  2622.     }

  2623. 

  2624.     /* Prediction */

  2625.     dcdiff += vc1_i_pred_dc(&v->s, v->overlap, v->pq, n, &dc_val, &dc_pred_dir);

  2626.     *dc_val = dcdiff;

  2627. 

  2628.     /* Store the quantized DC coeff, used for prediction */

  2629.     if (n < 4) {

  2630.         block[0] = dcdiff * s->y_dc_scale;

  2631.     } else {

  2632.         block[0] = dcdiff * s->c_dc_scale;

  2633.     }

  2634.     /* Skip ? */

  2635.     if (!coded) {

  2636.         goto not_coded;

  2637.     }

  2638. 

  2639.     // AC Decoding

  2640.     i = 1;

  2641. 

  2642.     {

  2643.         int last = 0, skip, value;

  2644.         const uint8_t *zz_table;

  2645.         int scale;

  2646.         int k;

  2647. 

  2648.         scale = v->pq * 2 + v->halfpq;

  2649. 

  2650.         if (v->s.ac_pred) {

  2651.             if (!dc_pred_dir)

  2652.                 zz_table = v->zz_8x8[2];

  2653.             else

  2654.                 zz_table = v->zz_8x8[3];

  2655.         } else

  2656.             zz_table = v->zz_8x8[1];

  2657. 

  2658.         ac_val  = s->ac_val[0][0] + s->block_index[n] * 16;

  2659.         ac_val2 = ac_val;

  2660.         if (dc_pred_dir) // left

  2661.             ac_val -= 16;

  2662.         else // top

  2663.             ac_val -= 16 * s->block_wrap[n];

  2664. 

  2665.         while (!last) {

  2666.             vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);

  2667.             i += skip;

  2668.             if (i > 63)

  2669.                 break;

  2670.             block[zz_table[i++]] = value;

  2671.         }

  2672. 

  2673.         /* apply AC prediction if needed */

  2674.         if (s->ac_pred) {

  2675.             if (dc_pred_dir) { // left

  2676.                 for (k = 1; k < 8; k++)

  2677.                     block[k << v->left_blk_sh] += ac_val[k];

  2678.             } else { // top

  2679.                 for (k = 1; k < 8; k++)

  2680.                     block[k << v->top_blk_sh] += ac_val[k + 8];

  2681.             }

  2682.         }

  2683.         /* save AC coeffs for further prediction */

  2684.         for (k = 1; k < 8; k++) {

  2685.             ac_val2[k]     = block[k << v->left_blk_sh];

  2686.             ac_val2[k + 8] = block[k << v->top_blk_sh];

  2687.         }

  2688. 

  2689.         /* scale AC coeffs */

  2690.         for (k = 1; k < 64; k++)

  2691.             if (block[k]) {

  2692.                 block[k] *= scale;

  2693.                 if (!v->pquantizer)

  2694.                     block[k] += (block[k] < 0) ? -v->pq : v->pq;

  2695.             }

  2696. 

  2697.         if (s->ac_pred) i = 63;

  2698.     }

  2699. 

  2700. not_coded:

  2701.     if (!coded) {

  2702.         int k, scale;

  2703.         ac_val  = s->ac_val[0][0] + s->block_index[n] * 16;

  2704.         ac_val2 = ac_val;

  2705. 

  2706.         i = 0;

  2707.         scale = v->pq * 2 + v->halfpq;

  2708.         memset(ac_val2, 0, 16 * 2);

  2709.         if (dc_pred_dir) { // left

  2710.             ac_val -= 16;

  2711.             if (s->ac_pred)

  2712.                 memcpy(ac_val2, ac_val, 8 * 2);

  2713.         } else { // top

  2714.             ac_val -= 16 * s->block_wrap[n];

  2715.             if (s->ac_pred)

  2716.                 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);

  2717.         }

  2718. 

  2719.         /* apply AC prediction if needed */

  2720.         if (s->ac_pred) {

  2721.             if (dc_pred_dir) { //left

  2722.                 for (k = 1; k < 8; k++) {

  2723.                     block[k << v->left_blk_sh] = ac_val[k] * scale;

  2724.                     if (!v->pquantizer && block[k << v->left_blk_sh])

  2725.                         block[k << v->left_blk_sh] += (block[k << v->left_blk_sh] < 0) ? -v->pq : v->pq;

  2726.                 }

  2727.             } else { // top

  2728.                 for (k = 1; k < 8; k++) {

  2729.                     block[k << v->top_blk_sh] = ac_val[k + 8] * scale;

  2730.                     if (!v->pquantizer && block[k << v->top_blk_sh])

  2731.                         block[k << v->top_blk_sh] += (block[k << v->top_blk_sh] < 0) ? -v->pq : v->pq;

  2732.                 }

  2733.             }

  2734.             i = 63;

  2735.         }

  2736.     }

  2737.     s->block_last_index[n] = i;

  2738. 

  2739.     return 0;

  2740. }

  2741. 

  2742. /** Decode intra block in intra frames - should be faster than decode_intra_block

  2743.  * @param v VC1Context

  2744.  * @param block block to decode

  2745.  * @param[in] n subblock number

  2746.  * @param coded are AC coeffs present or not

  2747.  * @param codingset set of VLC to decode data

  2748.  * @param mquant quantizer value for this macroblock

  2749.  */

  2750. static int vc1_decode_i_block_adv(VC1Context *v, int16_t block[64], int n,

  2751.                                   int coded, int codingset, int mquant)

  2752. {

  2753.     GetBitContext *gb = &v->s.gb;

  2754.     MpegEncContext *s = &v->s;

  2755.     int dc_pred_dir = 0; /* Direction of the DC prediction used */

  2756.     int i;

  2757.     int16_t *dc_val;

  2758.     int16_t *ac_val, *ac_val2;

  2759.     int dcdiff;

  2760.     int a_avail = v->a_avail, c_avail = v->c_avail;

  2761.     int use_pred = s->ac_pred;

  2762.     int scale;

  2763.     int q1, q2 = 0;

  2764.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  2765. 

  2766.     /* Get DC differential */

  2767.     if (n < 4) {

  2768.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2769.     } else {

  2770.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2771.     }

  2772.     if (dcdiff < 0) {

  2773.         av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");

  2774.         return -1;

  2775.     }

  2776.     if (dcdiff) {

  2777.         if (dcdiff == 119 /* ESC index value */) {

  2778.             /* TODO: Optimize */

  2779.             if (mquant == 1)      dcdiff = get_bits(gb, 10);

  2780.             else if (mquant == 2) dcdiff = get_bits(gb, 9);

  2781.             else                  dcdiff = get_bits(gb, 8);

  2782.         } else {

  2783.             if (mquant == 1)

  2784.                 dcdiff = (dcdiff << 2) + get_bits(gb, 2) - 3;

  2785.             else if (mquant == 2)

  2786.                 dcdiff = (dcdiff << 1) + get_bits1(gb)   - 1;

  2787.         }

  2788.         if (get_bits1(gb))

  2789.             dcdiff = -dcdiff;

  2790.     }

  2791. 

  2792.     /* Prediction */

  2793.     dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, v->a_avail, v->c_avail, &dc_val, &dc_pred_dir);

  2794.     *dc_val = dcdiff;

  2795. 

  2796.     /* Store the quantized DC coeff, used for prediction */

  2797.     if (n < 4) {

  2798.         block[0] = dcdiff * s->y_dc_scale;

  2799.     } else {

  2800.         block[0] = dcdiff * s->c_dc_scale;

  2801.     }

  2802. 

  2803.     //AC Decoding

  2804.     i = 1;

  2805. 

  2806.     /* check if AC is needed at all */

  2807.     if (!a_avail && !c_avail)

  2808.         use_pred = 0;

  2809.     ac_val  = s->ac_val[0][0] + s->block_index[n] * 16;

  2810.     ac_val2 = ac_val;

  2811. 

  2812.     scale = mquant * 2 + ((mquant == v->pq) ? v->halfpq : 0);

  2813. 

  2814.     if (dc_pred_dir) // left

  2815.         ac_val -= 16;

  2816.     else // top

  2817.         ac_val -= 16 * s->block_wrap[n];

  2818. 

  2819.     q1 = s->current_picture.qscale_table[mb_pos];

  2820.     if ( dc_pred_dir && c_avail && mb_pos)

  2821.         q2 = s->current_picture.qscale_table[mb_pos - 1];

  2822.     if (!dc_pred_dir && a_avail && mb_pos >= s->mb_stride)

  2823.         q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];

  2824.     if ( dc_pred_dir && n == 1)

  2825.         q2 = q1;

  2826.     if (!dc_pred_dir && n == 2)

  2827.         q2 = q1;

  2828.     if (n == 3)

  2829.         q2 = q1;

  2830. 

  2831.     if (coded) {

  2832.         int last = 0, skip, value;

  2833.         const uint8_t *zz_table;

  2834.         int k;

  2835. 

  2836.         if (v->s.ac_pred) {

  2837.             if (!use_pred && v->fcm == ILACE_FRAME) {

  2838.                 zz_table = v->zzi_8x8;

  2839.             } else {

  2840.                 if (!dc_pred_dir) // top

  2841.                     zz_table = v->zz_8x8[2];

  2842.                 else // left

  2843.                     zz_table = v->zz_8x8[3];

  2844.             }

  2845.         } else {

  2846.             if (v->fcm != ILACE_FRAME)

  2847.                 zz_table = v->zz_8x8[1];

  2848.             else

  2849.                 zz_table = v->zzi_8x8;

  2850.         }

  2851. 

  2852.         while (!last) {

  2853.             vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);

  2854.             i += skip;

  2855.             if (i > 63)

  2856.                 break;

  2857.             block[zz_table[i++]] = value;

  2858.         }

  2859. 

  2860.         /* apply AC prediction if needed */

  2861.         if (use_pred) {

  2862.             /* scale predictors if needed*/

  2863.             if (q2 && q1 != q2) {

  2864.                 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  2865.                 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  2866. 

  2867.                 if (q1 < 1)

  2868.                     return AVERROR_INVALIDDATA;

  2869.                 if (dc_pred_dir) { // left

  2870.                     for (k = 1; k < 8; k++)

  2871.                         block[k << v->left_blk_sh] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2872.                 } else { // top

  2873.                     for (k = 1; k < 8; k++)

  2874.                         block[k << v->top_blk_sh] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2875.                 }

  2876.             } else {

  2877.                 if (dc_pred_dir) { //left

  2878.                     for (k = 1; k < 8; k++)

  2879.                         block[k << v->left_blk_sh] += ac_val[k];

  2880.                 } else { //top

  2881.                     for (k = 1; k < 8; k++)

  2882.                         block[k << v->top_blk_sh] += ac_val[k + 8];

  2883.                 }

  2884.             }

  2885.         }

  2886.         /* save AC coeffs for further prediction */

  2887.         for (k = 1; k < 8; k++) {

  2888.             ac_val2[k    ] = block[k << v->left_blk_sh];

  2889.             ac_val2[k + 8] = block[k << v->top_blk_sh];

  2890.         }

  2891. 

  2892.         /* scale AC coeffs */

  2893.         for (k = 1; k < 64; k++)

  2894.             if (block[k]) {

  2895.                 block[k] *= scale;

  2896.                 if (!v->pquantizer)

  2897.                     block[k] += (block[k] < 0) ? -mquant : mquant;

  2898.             }

  2899. 

  2900.         if (use_pred) i = 63;

  2901.     } else { // no AC coeffs

  2902.         int k;

  2903. 

  2904.         memset(ac_val2, 0, 16 * 2);

  2905.         if (dc_pred_dir) { // left

  2906.             if (use_pred) {

  2907.                 memcpy(ac_val2, ac_val, 8 * 2);

  2908.                 if (q2 && q1 != q2) {

  2909.                     q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  2910.                     q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  2911.                     if (q1 < 1)

  2912.                         return AVERROR_INVALIDDATA;

  2913.                     for (k = 1; k < 8; k++)

  2914.                         ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2915.                 }

  2916.             }

  2917.         } else { // top

  2918.             if (use_pred) {

  2919.                 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);

  2920.                 if (q2 && q1 != q2) {

  2921.                     q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  2922.                     q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  2923.                     if (q1 < 1)

  2924.                         return AVERROR_INVALIDDATA;

  2925.                     for (k = 1; k < 8; k++)

  2926.                         ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2927.                 }

  2928.             }

  2929.         }

  2930. 

  2931.         /* apply AC prediction if needed */

  2932.         if (use_pred) {

  2933.             if (dc_pred_dir) { // left

  2934.                 for (k = 1; k < 8; k++) {

  2935.                     block[k << v->left_blk_sh] = ac_val2[k] * scale;

  2936.                     if (!v->pquantizer && block[k << v->left_blk_sh])

  2937.                         block[k << v->left_blk_sh] += (block[k << v->left_blk_sh] < 0) ? -mquant : mquant;

  2938.                 }

  2939.             } else { // top

  2940.                 for (k = 1; k < 8; k++) {

  2941.                     block[k << v->top_blk_sh] = ac_val2[k + 8] * scale;

  2942.                     if (!v->pquantizer && block[k << v->top_blk_sh])

  2943.                         block[k << v->top_blk_sh] += (block[k << v->top_blk_sh] < 0) ? -mquant : mquant;

  2944.                 }

  2945.             }

  2946.             i = 63;

  2947.         }

  2948.     }

  2949.     s->block_last_index[n] = i;

  2950. 

  2951.     return 0;

  2952. }

  2953. 

  2954. /** Decode intra block in inter frames - more generic version than vc1_decode_i_block

  2955.  * @param v VC1Context

  2956.  * @param block block to decode

  2957.  * @param[in] n subblock index

  2958.  * @param coded are AC coeffs present or not

  2959.  * @param mquant block quantizer

  2960.  * @param codingset set of VLC to decode data

  2961.  */

  2962. static int vc1_decode_intra_block(VC1Context *v, int16_t block[64], int n,

  2963.                                   int coded, int mquant, int codingset)

  2964. {

  2965.     GetBitContext *gb = &v->s.gb;

  2966.     MpegEncContext *s = &v->s;

  2967.     int dc_pred_dir = 0; /* Direction of the DC prediction used */

  2968.     int i;

  2969.     int16_t *dc_val;

  2970.     int16_t *ac_val, *ac_val2;

  2971.     int dcdiff;

  2972.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  2973.     int a_avail = v->a_avail, c_avail = v->c_avail;

  2974.     int use_pred = s->ac_pred;

  2975.     int scale;

  2976.     int q1, q2 = 0;

  2977. 

  2978.     s->dsp.clear_block(block);

  2979. 

  2980.     /* XXX: Guard against dumb values of mquant */

  2981.     mquant = (mquant < 1) ? 0 : ((mquant > 31) ? 31 : mquant);

  2982. 

  2983.     /* Set DC scale - y and c use the same */

  2984.     s->y_dc_scale = s->y_dc_scale_table[mquant];

  2985.     s->c_dc_scale = s->c_dc_scale_table[mquant];

  2986. 

  2987.     /* Get DC differential */

  2988.     if (n < 4) {

  2989.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2990.     } else {

  2991.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2992.     }

  2993.     if (dcdiff < 0) {

  2994.         av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");

  2995.         return -1;

  2996.     }

  2997.     if (dcdiff) {

  2998.         if (dcdiff == 119 /* ESC index value */) {

  2999.             /* TODO: Optimize */

  3000.             if (mquant == 1)      dcdiff = get_bits(gb, 10);

  3001.             else if (mquant == 2) dcdiff = get_bits(gb, 9);

  3002.             else                  dcdiff = get_bits(gb, 8);

  3003.         } else {

  3004.             if (mquant == 1)

  3005.                 dcdiff = (dcdiff << 2) + get_bits(gb, 2) - 3;

  3006.             else if (mquant == 2)

  3007.                 dcdiff = (dcdiff << 1) + get_bits1(gb)   - 1;

  3008.         }

  3009.         if (get_bits1(gb))

  3010.             dcdiff = -dcdiff;

  3011.     }

  3012. 

  3013.     /* Prediction */

  3014.     dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, a_avail, c_avail, &dc_val, &dc_pred_dir);

  3015.     *dc_val = dcdiff;

  3016. 

  3017.     /* Store the quantized DC coeff, used for prediction */

  3018. 

  3019.     if (n < 4) {

  3020.         block[0] = dcdiff * s->y_dc_scale;

  3021.     } else {

  3022.         block[0] = dcdiff * s->c_dc_scale;

  3023.     }

  3024. 

  3025.     //AC Decoding

  3026.     i = 1;

  3027. 

  3028.     /* check if AC is needed at all and adjust direction if needed */

  3029.     if (!a_avail) dc_pred_dir = 1;

  3030.     if (!c_avail) dc_pred_dir = 0;

  3031.     if (!a_avail && !c_avail) use_pred = 0;

  3032.     ac_val = s->ac_val[0][0] + s->block_index[n] * 16;

  3033.     ac_val2 = ac_val;

  3034. 

  3035.     scale = mquant * 2 + v->halfpq;

  3036. 

  3037.     if (dc_pred_dir) //left

  3038.         ac_val -= 16;

  3039.     else //top

  3040.         ac_val -= 16 * s->block_wrap[n];

  3041. 

  3042.     q1 = s->current_picture.qscale_table[mb_pos];

  3043.     if (dc_pred_dir && c_avail && mb_pos)

  3044.         q2 = s->current_picture.qscale_table[mb_pos - 1];

  3045.     if (!dc_pred_dir && a_avail && mb_pos >= s->mb_stride)

  3046.         q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];

  3047.     if ( dc_pred_dir && n == 1)

  3048.         q2 = q1;

  3049.     if (!dc_pred_dir && n == 2)

  3050.         q2 = q1;

  3051.     if (n == 3) q2 = q1;

  3052. 

  3053.     if (coded) {

  3054.         int last = 0, skip, value;

  3055.         int k;

  3056. 

  3057.         while (!last) {

  3058.             vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);

  3059.             i += skip;

  3060.             if (i > 63)

  3061.                 break;

  3062.             if (v->fcm == PROGRESSIVE)

  3063.                 block[v->zz_8x8[0][i++]] = value;

  3064.             else {

  3065.                 if (use_pred && (v->fcm == ILACE_FRAME)) {

  3066.                     if (!dc_pred_dir) // top

  3067.                         block[v->zz_8x8[2][i++]] = value;

  3068.                     else // left

  3069.                         block[v->zz_8x8[3][i++]] = value;

  3070.                 } else {

  3071.                     block[v->zzi_8x8[i++]] = value;

  3072.                 }

  3073.             }

  3074.         }

  3075. 

  3076.         /* apply AC prediction if needed */

  3077.         if (use_pred) {

  3078.             /* scale predictors if needed*/

  3079.             if (q2 && q1 != q2) {

  3080.                 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  3081.                 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  3082. 

  3083.                 if (q1 < 1)

  3084.                     return AVERROR_INVALIDDATA;

  3085.                 if (dc_pred_dir) { // left

  3086.                     for (k = 1; k < 8; k++)

  3087.                         block[k << v->left_blk_sh] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  3088.                 } else { //top

  3089.                     for (k = 1; k < 8; k++)

  3090.                         block[k << v->top_blk_sh] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  3091.                 }

  3092.             } else {

  3093.                 if (dc_pred_dir) { // left

  3094.                     for (k = 1; k < 8; k++)

  3095.                         block[k << v->left_blk_sh] += ac_val[k];

  3096.                 } else { // top

  3097.                     for (k = 1; k < 8; k++)

  3098.                         block[k << v->top_blk_sh] += ac_val[k + 8];

  3099.                 }

  3100.             }

  3101.         }

  3102.         /* save AC coeffs for further prediction */

  3103.         for (k = 1; k < 8; k++) {

  3104.             ac_val2[k    ] = block[k << v->left_blk_sh];

  3105.             ac_val2[k + 8] = block[k << v->top_blk_sh];

  3106.         }

  3107. 

  3108.         /* scale AC coeffs */

  3109.         for (k = 1; k < 64; k++)

  3110.             if (block[k]) {

  3111.                 block[k] *= scale;

  3112.                 if (!v->pquantizer)

  3113.                     block[k] += (block[k] < 0) ? -mquant : mquant;

  3114.             }

  3115. 

  3116.         if (use_pred) i = 63;

  3117.     } else { // no AC coeffs

  3118.         int k;

  3119. 

  3120.         memset(ac_val2, 0, 16 * 2);

  3121.         if (dc_pred_dir) { // left

  3122.             if (use_pred) {

  3123.                 memcpy(ac_val2, ac_val, 8 * 2);

  3124.                 if (q2 && q1 != q2) {

  3125.                     q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  3126.                     q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  3127.                     if (q1 < 1)

  3128.                         return AVERROR_INVALIDDATA;

  3129.                     for (k = 1; k < 8; k++)

  3130.                         ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  3131.                 }

  3132.             }

  3133.         } else { // top

  3134.             if (use_pred) {

  3135.                 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);

  3136.                 if (q2 && q1 != q2) {

  3137.                     q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  3138.                     q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  3139.                     if (q1 < 1)

  3140.                         return AVERROR_INVALIDDATA;

  3141.                     for (k = 1; k < 8; k++)

  3142.                         ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  3143.                 }

  3144.             }

  3145.         }

  3146. 

  3147.         /* apply AC prediction if needed */

  3148.         if (use_pred) {

  3149.             if (dc_pred_dir) { // left

  3150.                 for (k = 1; k < 8; k++) {

  3151.                     block[k << v->left_blk_sh] = ac_val2[k] * scale;

  3152.                     if (!v->pquantizer && block[k << v->left_blk_sh])

  3153.                         block[k << v->left_blk_sh] += (block[k << v->left_blk_sh] < 0) ? -mquant : mquant;

  3154.                 }

  3155.             } else { // top

  3156.                 for (k = 1; k < 8; k++) {

  3157.                     block[k << v->top_blk_sh] = ac_val2[k + 8] * scale;

  3158.                     if (!v->pquantizer && block[k << v->top_blk_sh])

  3159.                         block[k << v->top_blk_sh] += (block[k << v->top_blk_sh] < 0) ? -mquant : mquant;

  3160.                 }

  3161.             }

  3162.             i = 63;

  3163.         }

  3164.     }

  3165.     s->block_last_index[n] = i;

  3166. 

  3167.     return 0;

  3168. }

  3169. 

  3170. /** Decode P block

  3171.  */

  3172. static int vc1_decode_p_block(VC1Context *v, int16_t block[64], int n,

  3173.                               int mquant, int ttmb, int first_block,

  3174.                               uint8_t *dst, int linesize, int skip_block,

  3175.                               int *ttmb_out)

  3176. {

  3177.     MpegEncContext *s = &v->s;

  3178.     GetBitContext *gb = &s->gb;

  3179.     int i, j;

  3180.     int subblkpat = 0;

  3181.     int scale, off, idx, last, skip, value;

  3182.     int ttblk = ttmb & 7;

  3183.     int pat = 0;

  3184. 

  3185.     s->dsp.clear_block(block);

  3186. 

  3187.     if (ttmb == -1) {

  3188.         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)];

  3189.     }

  3190.     if (ttblk == TT_4X4) {

  3191.         subblkpat = ~(get_vlc2(gb, ff_vc1_subblkpat_vlc[v->tt_index].table, VC1_SUBBLKPAT_VLC_BITS, 1) + 1);

  3192.     }

  3193.     if ((ttblk != TT_8X8 && ttblk != TT_4X4)

  3194.         && ((v->ttmbf || (ttmb != -1 && (ttmb & 8) && !first_block))

  3195.             || (!v->res_rtm_flag && !first_block))) {

  3196.         subblkpat = decode012(gb);

  3197.         if (subblkpat)

  3198.             subblkpat ^= 3; // swap decoded pattern bits

  3199.         if (ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM)

  3200.             ttblk = TT_8X4;

  3201.         if (ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT)

  3202.             ttblk = TT_4X8;

  3203.     }

  3204.     scale = 2 * mquant + ((v->pq == mquant) ? v->halfpq : 0);

  3205. 

  3206.     // convert transforms like 8X4_TOP to generic TT and SUBBLKPAT

  3207.     if (ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) {

  3208.         subblkpat = 2 - (ttblk == TT_8X4_TOP);

  3209.         ttblk     = TT_8X4;

  3210.     }

  3211.     if (ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) {

  3212.         subblkpat = 2 - (ttblk == TT_4X8_LEFT);

  3213.         ttblk     = TT_4X8;

  3214.     }

  3215.     switch (ttblk) {

  3216.     case TT_8X8:

  3217.         pat  = 0xF;

  3218.         i    = 0;

  3219.         last = 0;

  3220.         while (!last) {

  3221.             vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  3222.             i += skip;

  3223.             if (i > 63)

  3224.                 break;

  3225.             if (!v->fcm)

  3226.                 idx = v->zz_8x8[0][i++];

  3227.             else

  3228.                 idx = v->zzi_8x8[i++];

  3229.             block[idx] = value * scale;

  3230.             if (!v->pquantizer)

  3231.                 block[idx] += (block[idx] < 0) ? -mquant : mquant;

  3232.         }

  3233.         if (!skip_block) {

  3234.             if (i == 1)

  3235.                 v->vc1dsp.vc1_inv_trans_8x8_dc(dst, linesize, block);

  3236.             else {

  3237.                 v->vc1dsp.vc1_inv_trans_8x8(block);

  3238.                 s->dsp.add_pixels_clamped(block, dst, linesize);

  3239.             }

  3240.         }

  3241.         break;

  3242.     case TT_4X4:

  3243.         pat = ~subblkpat & 0xF;

  3244.         for (j = 0; j < 4; j++) {

  3245.             last = subblkpat & (1 << (3 - j));

  3246.             i    = 0;

  3247.             off  = (j & 1) * 4 + (j & 2) * 16;

  3248.             while (!last) {

  3249.                 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  3250.                 i += skip;

  3251.                 if (i > 15)

  3252.                     break;

  3253.                 if (!v->fcm)

  3254.                     idx = ff_vc1_simple_progressive_4x4_zz[i++];

  3255.                 else

  3256.                     idx = ff_vc1_adv_interlaced_4x4_zz[i++];

  3257.                 block[idx + off] = value * scale;

  3258.                 if (!v->pquantizer)

  3259.                     block[idx + off] += (block[idx + off] < 0) ? -mquant : mquant;

  3260.             }

  3261.             if (!(subblkpat & (1 << (3 - j))) && !skip_block) {

  3262.                 if (i == 1)

  3263.                     v->vc1dsp.vc1_inv_trans_4x4_dc(dst + (j & 1) * 4 + (j & 2) * 2 * linesize, linesize, block + off);

  3264.                 else

  3265.                     v->vc1dsp.vc1_inv_trans_4x4(dst + (j & 1) * 4 + (j & 2) *  2 * linesize, linesize, block + off);

  3266.             }

  3267.         }

  3268.         break;

  3269.     case TT_8X4:

  3270.         pat = ~((subblkpat & 2) * 6 + (subblkpat & 1) * 3) & 0xF;

  3271.         for (j = 0; j < 2; j++) {

  3272.             last = subblkpat & (1 << (1 - j));

  3273.             i    = 0;

  3274.             off  = j * 32;

  3275.             while (!last) {

  3276.                 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  3277.                 i += skip;

  3278.                 if (i > 31)

  3279.                     break;

  3280.                 if (!v->fcm)

  3281.                     idx = v->zz_8x4[i++] + off;

  3282.                 else

  3283.                     idx = ff_vc1_adv_interlaced_8x4_zz[i++] + off;

  3284.                 block[idx] = value * scale;

  3285.                 if (!v->pquantizer)

  3286.                     block[idx] += (block[idx] < 0) ? -mquant : mquant;

  3287.             }

  3288.             if (!(subblkpat & (1 << (1 - j))) && !skip_block) {

  3289.                 if (i == 1)

  3290.                     v->vc1dsp.vc1_inv_trans_8x4_dc(dst + j * 4 * linesize, linesize, block + off);

  3291.                 else

  3292.                     v->vc1dsp.vc1_inv_trans_8x4(dst + j * 4 * linesize, linesize, block + off);

  3293.             }

  3294.         }

  3295.         break;

  3296.     case TT_4X8:

  3297.         pat = ~(subblkpat * 5) & 0xF;

  3298.         for (j = 0; j < 2; j++) {

  3299.             last = subblkpat & (1 << (1 - j));

  3300.             i    = 0;

  3301.             off  = j * 4;

  3302.             while (!last) {

  3303.                 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  3304.                 i += skip;

  3305.                 if (i > 31)

  3306.                     break;

  3307.                 if (!v->fcm)

  3308.                     idx = v->zz_4x8[i++] + off;

  3309.                 else

  3310.                     idx = ff_vc1_adv_interlaced_4x8_zz[i++] + off;

  3311.                 block[idx] = value * scale;

  3312.                 if (!v->pquantizer)

  3313.                     block[idx] += (block[idx] < 0) ? -mquant : mquant;

  3314.             }

  3315.             if (!(subblkpat & (1 << (1 - j))) && !skip_block) {

  3316.                 if (i == 1)

  3317.                     v->vc1dsp.vc1_inv_trans_4x8_dc(dst + j * 4, linesize, block + off);

  3318.                 else

  3319.                     v->vc1dsp.vc1_inv_trans_4x8(dst + j*4, linesize, block + off);

  3320.             }

  3321.         }

  3322.         break;

  3323.     }

  3324.     if (ttmb_out)

  3325.         *ttmb_out |= ttblk << (n * 4);

  3326.     return pat;

  3327. }

  3328. 

  3329. /** @} */ // Macroblock group

  3330. 

  3331. static const int size_table  [6] = { 0, 2, 3, 4,  5,  8 };

  3332. static const int offset_table[6] = { 0, 1, 3, 7, 15, 31 };

  3333. 

  3334. static av_always_inline void vc1_apply_p_v_loop_filter(VC1Context *v, int block_num)

  3335. {

  3336.     MpegEncContext *s  = &v->s;

  3337.     int mb_cbp         = v->cbp[s->mb_x - s->mb_stride],

  3338.         block_cbp      = mb_cbp      >> (block_num * 4), bottom_cbp,

  3339.         mb_is_intra    = v->is_intra[s->mb_x - s->mb_stride],

  3340.         block_is_intra = mb_is_intra >> (block_num * 4), bottom_is_intra;

  3341.     int idx, linesize  = block_num > 3 ? s->uvlinesize : s->linesize, ttblk;

  3342.     uint8_t *dst;

  3343. 

  3344.     if (block_num > 3) {

  3345.         dst      = s->dest[block_num - 3];

  3346.     } else {

  3347.         dst      = s->dest[0] + (block_num & 1) * 8 + ((block_num & 2) * 4 - 8) * linesize;

  3348.     }

  3349.     if (s->mb_y != s->end_mb_y || block_num < 2) {

  3350.         int16_t (*mv)[2];

  3351.         int mv_stride;

  3352. 

  3353.         if (block_num > 3) {

  3354.             bottom_cbp      = v->cbp[s->mb_x]      >> (block_num * 4);

  3355.             bottom_is_intra = v->is_intra[s->mb_x] >> (block_num * 4);

  3356.             mv              = &v->luma_mv[s->mb_x - s->mb_stride];

  3357.             mv_stride       = s->mb_stride;

  3358.         } else {

  3359.             bottom_cbp      = (block_num < 2) ? (mb_cbp               >> ((block_num + 2) * 4))

  3360.                                               : (v->cbp[s->mb_x]      >> ((block_num - 2) * 4));

  3361.             bottom_is_intra = (block_num < 2) ? (mb_is_intra          >> ((block_num + 2) * 4))

  3362.                                               : (v->is_intra[s->mb_x] >> ((block_num - 2) * 4));

  3363.             mv_stride       = s->b8_stride;

  3364.             mv              = &s->current_picture.motion_val[0][s->block_index[block_num] - 2 * mv_stride];

  3365.         }

  3366. 

  3367.         if (bottom_is_intra & 1 || block_is_intra & 1 ||

  3368.             mv[0][0] != mv[mv_stride][0] || mv[0][1] != mv[mv_stride][1]) {

  3369.             v->vc1dsp.vc1_v_loop_filter8(dst, linesize, v->pq);

  3370.         } else {

  3371.             idx = ((bottom_cbp >> 2) | block_cbp) & 3;

  3372.             if (idx == 3) {

  3373.                 v->vc1dsp.vc1_v_loop_filter8(dst, linesize, v->pq);

  3374.             } else if (idx) {

  3375.                 if (idx == 1)

  3376.                     v->vc1dsp.vc1_v_loop_filter4(dst + 4, linesize, v->pq);

  3377.                 else

  3378.                     v->vc1dsp.vc1_v_loop_filter4(dst,     linesize, v->pq);

  3379.             }

  3380.         }

  3381.     }

  3382. 

  3383.     dst -= 4 * linesize;

  3384.     ttblk = (v->ttblk[s->mb_x - s->mb_stride] >> (block_num * 4)) & 0xF;

  3385.     if (ttblk == TT_4X4 || ttblk == TT_8X4) {

  3386.         idx = (block_cbp | (block_cbp >> 2)) & 3;

  3387.         if (idx == 3) {

  3388.             v->vc1dsp.vc1_v_loop_filter8(dst, linesize, v->pq);

  3389.         } else if (idx) {

  3390.             if (idx == 1)

  3391.                 v->vc1dsp.vc1_v_loop_filter4(dst + 4, linesize, v->pq);

  3392.             else

  3393.                 v->vc1dsp.vc1_v_loop_filter4(dst,     linesize, v->pq);

  3394.         }

  3395.     }

  3396. }

  3397. 

  3398. static av_always_inline void vc1_apply_p_h_loop_filter(VC1Context *v, int block_num)

  3399. {

  3400.     MpegEncContext *s  = &v->s;

  3401.     int mb_cbp         = v->cbp[s->mb_x - 1 - s->mb_stride],

  3402.         block_cbp      = mb_cbp      >> (block_num * 4), right_cbp,

  3403.         mb_is_intra    = v->is_intra[s->mb_x - 1 - s->mb_stride],

  3404.         block_is_intra = mb_is_intra >> (block_num * 4), right_is_intra;

  3405.     int idx, linesize  = block_num > 3 ? s->uvlinesize : s->linesize, ttblk;

  3406.     uint8_t *dst;

  3407. 

  3408.     if (block_num > 3) {

  3409.         dst = s->dest[block_num - 3] - 8 * linesize;

  3410.     } else {

  3411.         dst = s->dest[0] + (block_num & 1) * 8 + ((block_num & 2) * 4 - 16) * linesize - 8;

  3412.     }

  3413. 

  3414.     if (s->mb_x != s->mb_width || !(block_num & 5)) {

  3415.         int16_t (*mv)[2];

  3416. 

  3417.         if (block_num > 3) {

  3418.             right_cbp      = v->cbp[s->mb_x - s->mb_stride] >> (block_num * 4);

  3419.             right_is_intra = v->is_intra[s->mb_x - s->mb_stride] >> (block_num * 4);

  3420.             mv             = &v->luma_mv[s->mb_x - s->mb_stride - 1];

  3421.         } else {

  3422.             right_cbp      = (block_num & 1) ? (v->cbp[s->mb_x - s->mb_stride]      >> ((block_num - 1) * 4))

  3423.                                              : (mb_cbp                              >> ((block_num + 1) * 4));

  3424.             right_is_intra = (block_num & 1) ? (v->is_intra[s->mb_x - s->mb_stride] >> ((block_num - 1) * 4))

  3425.                                              : (mb_is_intra                         >> ((block_num + 1) * 4));

  3426.             mv             = &s->current_picture.motion_val[0][s->block_index[block_num] - s->b8_stride * 2 - 2];

  3427.         }

  3428.         if (block_is_intra & 1 || right_is_intra & 1 || mv[0][0] != mv[1][0] || mv[0][1] != mv[1][1]) {

  3429.             v->vc1dsp.vc1_h_loop_filter8(dst, linesize, v->pq);

  3430.         } else {

  3431.             idx = ((right_cbp >> 1) | block_cbp) & 5; // FIXME check

  3432.             if (idx == 5) {

  3433.                 v->vc1dsp.vc1_h_loop_filter8(dst, linesize, v->pq);

  3434.             } else if (idx) {

  3435.                 if (idx == 1)

  3436.                     v->vc1dsp.vc1_h_loop_filter4(dst + 4 * linesize, linesize, v->pq);

  3437.                 else

  3438.                     v->vc1dsp.vc1_h_loop_filter4(dst,                linesize, v->pq);

  3439.             }

  3440.         }

  3441.     }

  3442. 

  3443.     dst -= 4;

  3444.     ttblk = (v->ttblk[s->mb_x - s->mb_stride - 1] >> (block_num * 4)) & 0xf;

  3445.     if (ttblk == TT_4X4 || ttblk == TT_4X8) {

  3446.         idx = (block_cbp | (block_cbp >> 1)) & 5;

  3447.         if (idx == 5) {

  3448.             v->vc1dsp.vc1_h_loop_filter8(dst, linesize, v->pq);

  3449.         } else if (idx) {

  3450.             if (idx == 1)

  3451.                 v->vc1dsp.vc1_h_loop_filter4(dst + linesize * 4, linesize, v->pq);

  3452.             else

  3453.                 v->vc1dsp.vc1_h_loop_filter4(dst,                linesize, v->pq);

  3454.         }

  3455.     }

  3456. }

  3457. 

  3458. static void vc1_apply_p_loop_filter(VC1Context *v)

  3459. {

  3460.     MpegEncContext *s = &v->s;

  3461.     int i;

  3462. 

  3463.     for (i = 0; i < 6; i++) {

  3464.         vc1_apply_p_v_loop_filter(v, i);

  3465.     }

  3466. 

  3467.     /* V always precedes H, therefore we run H one MB before V;

  3468.      * at the end of a row, we catch up to complete the row */

  3469.     if (s->mb_x) {

  3470.         for (i = 0; i < 6; i++) {

  3471.             vc1_apply_p_h_loop_filter(v, i);

  3472.         }

  3473.         if (s->mb_x == s->mb_width - 1) {

  3474.             s->mb_x++;

  3475.             ff_update_block_index(s);

  3476.             for (i = 0; i < 6; i++) {

  3477.                 vc1_apply_p_h_loop_filter(v, i);

  3478.             }

  3479.         }

  3480.     }

  3481. }

  3482. 

  3483. /** Decode one P-frame MB

  3484.  */

  3485. static int vc1_decode_p_mb(VC1Context *v)

  3486. {

  3487.     MpegEncContext *s = &v->s;

  3488.     GetBitContext *gb = &s->gb;

  3489.     int i, j;

  3490.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  3491.     int cbp; /* cbp decoding stuff */

  3492.     int mqdiff, mquant; /* MB quantization */

  3493.     int ttmb = v->ttfrm; /* MB Transform type */

  3494. 

  3495.     int mb_has_coeffs = 1; /* last_flag */

  3496.     int dmv_x, dmv_y; /* Differential MV components */

  3497.     int index, index1; /* LUT indexes */

  3498.     int val, sign; /* temp values */

  3499.     int first_block = 1;

  3500.     int dst_idx, off;

  3501.     int skipped, fourmv;

  3502.     int block_cbp = 0, pat, block_tt = 0, block_intra = 0;

  3503. 

  3504.     mquant = v->pq; /* lossy initialization */

  3505. 

  3506.     if (v->mv_type_is_raw)

  3507.         fourmv = get_bits1(gb);

  3508.     else

  3509.         fourmv = v->mv_type_mb_plane[mb_pos];

  3510.     if (v->skip_is_raw)

  3511.         skipped = get_bits1(gb);

  3512.     else

  3513.         skipped = v->s.mbskip_table[mb_pos];

  3514. 

  3515.     if (!fourmv) { /* 1MV mode */

  3516.         if (!skipped) {

  3517.             GET_MVDATA(dmv_x, dmv_y);

  3518. 

  3519.             if (s->mb_intra) {

  3520.                 s->current_picture.motion_val[1][s->block_index[0]][0] = 0;

  3521.                 s->current_picture.motion_val[1][s->block_index[0]][1] = 0;

  3522.             }

  3523.             s->current_picture.mb_type[mb_pos] = s->mb_intra ? MB_TYPE_INTRA : MB_TYPE_16x16;

  3524.             vc1_pred_mv(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], 0, 0);

  3525. 

  3526.             /* FIXME Set DC val for inter block ? */

  3527.             if (s->mb_intra && !mb_has_coeffs) {

  3528.                 GET_MQUANT();

  3529.                 s->ac_pred = get_bits1(gb);

  3530.                 cbp        = 0;

  3531.             } else if (mb_has_coeffs) {

  3532.                 if (s->mb_intra)

  3533.                     s->ac_pred = get_bits1(gb);

  3534.                 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  3535.                 GET_MQUANT();

  3536.             } else {

  3537.                 mquant = v->pq;

  3538.                 cbp    = 0;

  3539.             }

  3540.             s->current_picture.qscale_table[mb_pos] = mquant;

  3541. 

  3542.             if (!v->ttmbf && !s->mb_intra && mb_has_coeffs)

  3543.                 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table,

  3544.                                 VC1_TTMB_VLC_BITS, 2);

  3545.             if (!s->mb_intra) vc1_mc_1mv(v, 0);

  3546.             dst_idx = 0;

  3547.             for (i = 0; i < 6; i++) {

  3548.                 s->dc_val[0][s->block_index[i]] = 0;

  3549.                 dst_idx += i >> 2;

  3550.                 val = ((cbp >> (5 - i)) & 1);

  3551.                 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  3552.                 v->mb_type[0][s->block_index[i]] = s->mb_intra;

  3553.                 if (s->mb_intra) {

  3554.                     /* check if prediction blocks A and C are available */

  3555.                     v->a_avail = v->c_avail = 0;

  3556.                     if (i == 2 || i == 3 || !s->first_slice_line)

  3557.                         v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  3558.                     if (i == 1 || i == 3 || s->mb_x)

  3559.                         v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  3560. 

  3561.                     vc1_decode_intra_block(v, s->block[i], i, val, mquant,

  3562.                                            (i & 4) ? v->codingset2 : v->codingset);

  3563.                     if ((i>3) && (s->flags & CODEC_FLAG_GRAY))

  3564.                         continue;

  3565.                     v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);

  3566.                     if (v->rangeredfrm)

  3567.                         for (j = 0; j < 64; j++)

  3568.                             s->block[i][j] <<= 1;

  3569.                     s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);

  3570.                     if (v->pq >= 9 && v->overlap) {

  3571.                         if (v->c_avail)

  3572.                             v->vc1dsp.vc1_h_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);

  3573.                         if (v->a_avail)

  3574.                             v->vc1dsp.vc1_v_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);

  3575.                     }

  3576.                     block_cbp   |= 0xF << (i << 2);

  3577.                     block_intra |= 1 << i;

  3578.                 } else if (val) {

  3579.                     pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block,

  3580.                                              s->dest[dst_idx] + off, (i & 4) ? s->uvlinesize : s->linesize,

  3581.                                              (i & 4) && (s->flags & CODEC_FLAG_GRAY), &block_tt);

  3582.                     block_cbp |= pat << (i << 2);

  3583.                     if (!v->ttmbf && ttmb < 8)

  3584.                         ttmb = -1;

  3585.                     first_block = 0;

  3586.                 }

  3587.             }

  3588.         } else { // skipped

  3589.             s->mb_intra = 0;

  3590.             for (i = 0; i < 6; i++) {

  3591.                 v->mb_type[0][s->block_index[i]] = 0;

  3592.                 s->dc_val[0][s->block_index[i]]  = 0;

  3593.             }

  3594.             s->current_picture.mb_type[mb_pos]      = MB_TYPE_SKIP;

  3595.             s->current_picture.qscale_table[mb_pos] = 0;

  3596.             vc1_pred_mv(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0], 0, 0);

  3597.             vc1_mc_1mv(v, 0);

  3598.         }

  3599.     } else { // 4MV mode

  3600.         if (!skipped /* unskipped MB */) {

  3601.             int intra_count = 0, coded_inter = 0;

  3602.             int is_intra[6], is_coded[6];

  3603.             /* Get CBPCY */

  3604.             cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  3605.             for (i = 0; i < 6; i++) {

  3606.                 val = ((cbp >> (5 - i)) & 1);

  3607.                 s->dc_val[0][s->block_index[i]] = 0;

  3608.                 s->mb_intra                     = 0;

  3609.                 if (i < 4) {

  3610.                     dmv_x = dmv_y = 0;

  3611.                     s->mb_intra   = 0;

  3612.                     mb_has_coeffs = 0;

  3613.                     if (val) {

  3614.                         GET_MVDATA(dmv_x, dmv_y);

  3615.                     }

  3616.                     vc1_pred_mv(v, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0], 0, 0);

  3617.                     if (!s->mb_intra)

  3618.                         vc1_mc_4mv_luma(v, i, 0, 0);

  3619.                     intra_count += s->mb_intra;

  3620.                     is_intra[i]  = s->mb_intra;

  3621.                     is_coded[i]  = mb_has_coeffs;

  3622.                 }

  3623.                 if (i & 4) {

  3624.                     is_intra[i] = (intra_count >= 3);

  3625.                     is_coded[i] = val;

  3626.                 }

  3627.                 if (i == 4)

  3628.                     vc1_mc_4mv_chroma(v, 0);

  3629.                 v->mb_type[0][s->block_index[i]] = is_intra[i];

  3630.                 if (!coded_inter)

  3631.                     coded_inter = !is_intra[i] & is_coded[i];

  3632.             }

  3633.             // if there are no coded blocks then don't do anything more

  3634.             dst_idx = 0;

  3635.             if (!intra_count && !coded_inter)

  3636.                 goto end;

  3637.             GET_MQUANT();

  3638.             s->current_picture.qscale_table[mb_pos] = mquant;

  3639.             /* test if block is intra and has pred */

  3640.             {

  3641.                 int intrapred = 0;

  3642.                 for (i = 0; i < 6; i++)

  3643.                     if (is_intra[i]) {

  3644.                         if (((!s->first_slice_line || (i == 2 || i == 3)) && v->mb_type[0][s->block_index[i] - s->block_wrap[i]])

  3645.                             || ((s->mb_x || (i == 1 || i == 3)) && v->mb_type[0][s->block_index[i] - 1])) {

  3646.                             intrapred = 1;

  3647.                             break;

  3648.                         }

  3649.                     }

  3650.                 if (intrapred)

  3651.                     s->ac_pred = get_bits1(gb);

  3652.                 else

  3653.                     s->ac_pred = 0;

  3654.             }

  3655.             if (!v->ttmbf && coded_inter)

  3656.                 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  3657.             for (i = 0; i < 6; i++) {

  3658.                 dst_idx    += i >> 2;

  3659.                 off         = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  3660.                 s->mb_intra = is_intra[i];

  3661.                 if (is_intra[i]) {

  3662.                     /* check if prediction blocks A and C are available */

  3663.                     v->a_avail = v->c_avail = 0;

  3664.                     if (i == 2 || i == 3 || !s->first_slice_line)

  3665.                         v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  3666.                     if (i == 1 || i == 3 || s->mb_x)

  3667.                         v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  3668. 

  3669.                     vc1_decode_intra_block(v, s->block[i], i, is_coded[i], mquant,

  3670.                                            (i & 4) ? v->codingset2 : v->codingset);

  3671.                     if ((i>3) && (s->flags & CODEC_FLAG_GRAY))

  3672.                         continue;

  3673.                     v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);

  3674.                     if (v->rangeredfrm)

  3675.                         for (j = 0; j < 64; j++)

  3676.                             s->block[i][j] <<= 1;

  3677.                     s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off,

  3678.                                                      (i & 4) ? s->uvlinesize : s->linesize);

  3679.                     if (v->pq >= 9 && v->overlap) {

  3680.                         if (v->c_avail)

  3681.                             v->vc1dsp.vc1_h_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);

  3682.                         if (v->a_avail)

  3683.                             v->vc1dsp.vc1_v_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);

  3684.                     }

  3685.                     block_cbp   |= 0xF << (i << 2);

  3686.                     block_intra |= 1 << i;

  3687.                 } else if (is_coded[i]) {

  3688.                     pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,

  3689.                                              first_block, s->dest[dst_idx] + off,

  3690.                                              (i & 4) ? s->uvlinesize : s->linesize,

  3691.                                              (i & 4) && (s->flags & CODEC_FLAG_GRAY),

  3692.                                              &block_tt);

  3693.                     block_cbp |= pat << (i << 2);

  3694.                     if (!v->ttmbf && ttmb < 8)

  3695.                         ttmb = -1;

  3696.                     first_block = 0;

  3697.                 }

  3698.             }

  3699.         } else { // skipped MB

  3700.             s->mb_intra                               = 0;

  3701.             s->current_picture.qscale_table[mb_pos] = 0;

  3702.             for (i = 0; i < 6; i++) {

  3703.                 v->mb_type[0][s->block_index[i]] = 0;

  3704.                 s->dc_val[0][s->block_index[i]]  = 0;

  3705.             }

  3706.             for (i = 0; i < 4; i++) {

  3707.                 vc1_pred_mv(v, i, 0, 0, 0, v->range_x, v->range_y, v->mb_type[0], 0, 0);

  3708.                 vc1_mc_4mv_luma(v, i, 0, 0);

  3709.             }

  3710.             vc1_mc_4mv_chroma(v, 0);

  3711.             s->current_picture.qscale_table[mb_pos] = 0;

  3712.         }

  3713.     }

  3714. end:

  3715.     v->cbp[s->mb_x]      = block_cbp;

  3716.     v->ttblk[s->mb_x]    = block_tt;

  3717.     v->is_intra[s->mb_x] = block_intra;

  3718. 

  3719.     return 0;

  3720. }

  3721. 

  3722. /* Decode one macroblock in an interlaced frame p picture */

  3723. 

  3724. static int vc1_decode_p_mb_intfr(VC1Context *v)

  3725. {

  3726.     MpegEncContext *s = &v->s;

  3727.     GetBitContext *gb = &s->gb;

  3728.     int i;

  3729.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  3730.     int cbp = 0; /* cbp decoding stuff */

  3731.     int mqdiff, mquant; /* MB quantization */

  3732.     int ttmb = v->ttfrm; /* MB Transform type */

  3733. 

  3734.     int mb_has_coeffs = 1; /* last_flag */

  3735.     int dmv_x, dmv_y; /* Differential MV components */

  3736.     int val; /* temp value */

  3737.     int first_block = 1;

  3738.     int dst_idx, off;

  3739.     int skipped, fourmv = 0, twomv = 0;

  3740.     int block_cbp = 0, pat, block_tt = 0;

  3741.     int idx_mbmode = 0, mvbp;

  3742.     int stride_y, fieldtx;

  3743. 

  3744.     mquant = v->pq; /* Loosy initialization */

  3745. 

  3746.     if (v->skip_is_raw)

  3747.         skipped = get_bits1(gb);

  3748.     else

  3749.         skipped = v->s.mbskip_table[mb_pos];

  3750.     if (!skipped) {

  3751.         if (v->fourmvswitch)

  3752.             idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_INTFR_4MV_MBMODE_VLC_BITS, 2); // try getting this done

  3753.         else

  3754.             idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_INTFR_NON4MV_MBMODE_VLC_BITS, 2); // in a single line

  3755.         switch (ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0]) {

  3756.         /* store the motion vector type in a flag (useful later) */

  3757.         case MV_PMODE_INTFR_4MV:

  3758.             fourmv = 1;

  3759.             v->blk_mv_type[s->block_index[0]] = 0;

  3760.             v->blk_mv_type[s->block_index[1]] = 0;

  3761.             v->blk_mv_type[s->block_index[2]] = 0;

  3762.             v->blk_mv_type[s->block_index[3]] = 0;

  3763.             break;

  3764.         case MV_PMODE_INTFR_4MV_FIELD:

  3765.             fourmv = 1;

  3766.             v->blk_mv_type[s->block_index[0]] = 1;

  3767.             v->blk_mv_type[s->block_index[1]] = 1;

  3768.             v->blk_mv_type[s->block_index[2]] = 1;

  3769.             v->blk_mv_type[s->block_index[3]] = 1;

  3770.             break;

  3771.         case MV_PMODE_INTFR_2MV_FIELD:

  3772.             twomv = 1;

  3773.             v->blk_mv_type[s->block_index[0]] = 1;

  3774.             v->blk_mv_type[s->block_index[1]] = 1;

  3775.             v->blk_mv_type[s->block_index[2]] = 1;

  3776.             v->blk_mv_type[s->block_index[3]] = 1;

  3777.             break;

  3778.         case MV_PMODE_INTFR_1MV:

  3779.             v->blk_mv_type[s->block_index[0]] = 0;

  3780.             v->blk_mv_type[s->block_index[1]] = 0;

  3781.             v->blk_mv_type[s->block_index[2]] = 0;

  3782.             v->blk_mv_type[s->block_index[3]] = 0;

  3783.             break;

  3784.         }

  3785.         if (ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0] == MV_PMODE_INTFR_INTRA) { // intra MB

  3786.             for (i = 0; i < 4; i++) {

  3787.                 s->current_picture.motion_val[1][s->block_index[i]][0] = 0;

  3788.                 s->current_picture.motion_val[1][s->block_index[i]][1] = 0;

  3789.             }

  3790.             s->current_picture.mb_type[mb_pos]                     = MB_TYPE_INTRA;

  3791.             s->mb_intra = v->is_intra[s->mb_x] = 1;

  3792.             for (i = 0; i < 6; i++)

  3793.                 v->mb_type[0][s->block_index[i]] = 1;

  3794.             fieldtx = v->fieldtx_plane[mb_pos] = get_bits1(gb);

  3795.             mb_has_coeffs = get_bits1(gb);

  3796.             if (mb_has_coeffs)

  3797.                 cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  3798.             v->s.ac_pred = v->acpred_plane[mb_pos] = get_bits1(gb);

  3799.             GET_MQUANT();

  3800.             s->current_picture.qscale_table[mb_pos] = mquant;

  3801.             /* Set DC scale - y and c use the same (not sure if necessary here) */

  3802.             s->y_dc_scale = s->y_dc_scale_table[mquant];

  3803.             s->c_dc_scale = s->c_dc_scale_table[mquant];

  3804.             dst_idx = 0;

  3805.             for (i = 0; i < 6; i++) {

  3806.                 s->dc_val[0][s->block_index[i]] = 0;

  3807.                 dst_idx += i >> 2;

  3808.                 val = ((cbp >> (5 - i)) & 1);

  3809.                 v->mb_type[0][s->block_index[i]] = s->mb_intra;

  3810.                 v->a_avail = v->c_avail = 0;

  3811.                 if (i == 2 || i == 3 || !s->first_slice_line)

  3812.                     v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  3813.                 if (i == 1 || i == 3 || s->mb_x)

  3814.                     v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  3815. 

  3816.                 vc1_decode_intra_block(v, s->block[i], i, val, mquant,

  3817.                                        (i & 4) ? v->codingset2 : v->codingset);

  3818.                 if ((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;

  3819.                 v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);

  3820.                 if (i < 4) {

  3821.                     stride_y = s->linesize << fieldtx;

  3822.                     off = (fieldtx) ? ((i & 1) * 8) + ((i & 2) >> 1) * s->linesize : (i & 1) * 8 + 4 * (i & 2) * s->linesize;

  3823.                 } else {

  3824.                     stride_y = s->uvlinesize;

  3825.                     off = 0;

  3826.                 }

  3827.                 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, stride_y);

  3828.                 //TODO: loop filter

  3829.             }

  3830. 

  3831.         } else { // inter MB

  3832.             mb_has_coeffs = ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][3];

  3833.             if (mb_has_coeffs)

  3834.                 cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  3835.             if (ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0] == MV_PMODE_INTFR_2MV_FIELD) {

  3836.                 v->twomvbp = get_vlc2(gb, v->twomvbp_vlc->table, VC1_2MV_BLOCK_PATTERN_VLC_BITS, 1);

  3837.             } else {

  3838.                 if ((ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0] == MV_PMODE_INTFR_4MV)

  3839.                     || (ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0] == MV_PMODE_INTFR_4MV_FIELD)) {

  3840.                     v->fourmvbp = get_vlc2(gb, v->fourmvbp_vlc->table, VC1_4MV_BLOCK_PATTERN_VLC_BITS, 1);

  3841.                 }

  3842.             }

  3843.             s->mb_intra = v->is_intra[s->mb_x] = 0;

  3844.             for (i = 0; i < 6; i++)

  3845.                 v->mb_type[0][s->block_index[i]] = 0;

  3846.             fieldtx = v->fieldtx_plane[mb_pos] = ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][1];

  3847.             /* for all motion vector read MVDATA and motion compensate each block */

  3848.             dst_idx = 0;

  3849.             if (fourmv) {

  3850.                 mvbp = v->fourmvbp;

  3851.                 for (i = 0; i < 6; i++) {

  3852.                     if (i < 4) {

  3853.                         dmv_x = dmv_y = 0;

  3854.                         val   = ((mvbp >> (3 - i)) & 1);

  3855.                         if (val) {

  3856.                             get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);

  3857.                         }

  3858.                         vc1_pred_mv_intfr(v, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0], 0);

  3859.                         vc1_mc_4mv_luma(v, i, 0, 0);

  3860.                     } else if (i == 4) {

  3861.                         vc1_mc_4mv_chroma4(v);

  3862.                     }

  3863.                 }

  3864.             } else if (twomv) {

  3865.                 mvbp  = v->twomvbp;

  3866.                 dmv_x = dmv_y = 0;

  3867.                 if (mvbp & 2) {

  3868.                     get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);

  3869.                 }

  3870.                 vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0], 0);

  3871.                 vc1_mc_4mv_luma(v, 0, 0, 0);

  3872.                 vc1_mc_4mv_luma(v, 1, 0, 0);

  3873.                 dmv_x = dmv_y = 0;

  3874.                 if (mvbp & 1) {

  3875.                     get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);

  3876.                 }

  3877.                 vc1_pred_mv_intfr(v, 2, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0], 0);

  3878.                 vc1_mc_4mv_luma(v, 2, 0, 0);

  3879.                 vc1_mc_4mv_luma(v, 3, 0, 0);

  3880.                 vc1_mc_4mv_chroma4(v);

  3881.             } else {

  3882.                 mvbp = ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][2];

  3883.                 dmv_x = dmv_y = 0;

  3884.                 if (mvbp) {

  3885.                     get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);

  3886.                 }

  3887.                 vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], 0);

  3888.                 vc1_mc_1mv(v, 0);

  3889.             }

  3890.             if (cbp)

  3891.                 GET_MQUANT();  // p. 227

  3892.             s->current_picture.qscale_table[mb_pos] = mquant;

  3893.             if (!v->ttmbf && cbp)

  3894.                 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  3895.             for (i = 0; i < 6; i++) {

  3896.                 s->dc_val[0][s->block_index[i]] = 0;

  3897.                 dst_idx += i >> 2;

  3898.                 val = ((cbp >> (5 - i)) & 1);

  3899.                 if (!fieldtx)

  3900.                     off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  3901.                 else

  3902.                     off = (i & 4) ? 0 : ((i & 1) * 8 + ((i > 1) * s->linesize));

  3903.                 if (val) {

  3904.                     pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,

  3905.                                              first_block, s->dest[dst_idx] + off,

  3906.                                              (i & 4) ? s->uvlinesize : (s->linesize << fieldtx),

  3907.                                              (i & 4) && (s->flags & CODEC_FLAG_GRAY), &block_tt);

  3908.                     block_cbp |= pat << (i << 2);

  3909.                     if (!v->ttmbf && ttmb < 8)

  3910.                         ttmb = -1;

  3911.                     first_block = 0;

  3912.                 }

  3913.             }

  3914.         }

  3915.     } else { // skipped

  3916.         s->mb_intra = v->is_intra[s->mb_x] = 0;

  3917.         for (i = 0; i < 6; i++) {

  3918.             v->mb_type[0][s->block_index[i]] = 0;

  3919.             s->dc_val[0][s->block_index[i]] = 0;

  3920.         }

  3921.         s->current_picture.mb_type[mb_pos]      = MB_TYPE_SKIP;

  3922.         s->current_picture.qscale_table[mb_pos] = 0;

  3923.         v->blk_mv_type[s->block_index[0]] = 0;

  3924.         v->blk_mv_type[s->block_index[1]] = 0;

  3925.         v->blk_mv_type[s->block_index[2]] = 0;

  3926.         v->blk_mv_type[s->block_index[3]] = 0;

  3927.         vc1_pred_mv_intfr(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0], 0);

  3928.         vc1_mc_1mv(v, 0);

  3929.     }

  3930.     if (s->mb_x == s->mb_width - 1)

  3931.         memmove(v->is_intra_base, v->is_intra, sizeof(v->is_intra_base[0])*s->mb_stride);

  3932.     return 0;

  3933. }

  3934. 

  3935. static int vc1_decode_p_mb_intfi(VC1Context *v)

  3936. {

  3937.     MpegEncContext *s = &v->s;

  3938.     GetBitContext *gb = &s->gb;

  3939.     int i;

  3940.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  3941.     int cbp = 0; /* cbp decoding stuff */

  3942.     int mqdiff, mquant; /* MB quantization */

  3943.     int ttmb = v->ttfrm; /* MB Transform type */

  3944. 

  3945.     int mb_has_coeffs = 1; /* last_flag */

  3946.     int dmv_x, dmv_y; /* Differential MV components */

  3947.     int val; /* temp values */

  3948.     int first_block = 1;

  3949.     int dst_idx, off;

  3950.     int pred_flag;

  3951.     int block_cbp = 0, pat, block_tt = 0;

  3952.     int idx_mbmode = 0;

  3953. 

  3954.     mquant = v->pq; /* Loosy initialization */

  3955. 

  3956.     idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_IF_MBMODE_VLC_BITS, 2);

  3957.     if (idx_mbmode <= 1) { // intra MB

  3958.         s->mb_intra = v->is_intra[s->mb_x] = 1;

  3959.         s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][0] = 0;

  3960.         s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][1] = 0;

  3961.         s->current_picture.mb_type[mb_pos + v->mb_off] = MB_TYPE_INTRA;

  3962.         GET_MQUANT();

  3963.         s->current_picture.qscale_table[mb_pos] = mquant;

  3964.         /* Set DC scale - y and c use the same (not sure if necessary here) */

  3965.         s->y_dc_scale = s->y_dc_scale_table[mquant];

  3966.         s->c_dc_scale = s->c_dc_scale_table[mquant];

  3967.         v->s.ac_pred  = v->acpred_plane[mb_pos] = get_bits1(gb);

  3968.         mb_has_coeffs = idx_mbmode & 1;

  3969.         if (mb_has_coeffs)

  3970.             cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_ICBPCY_VLC_BITS, 2);

  3971.         dst_idx = 0;

  3972.         for (i = 0; i < 6; i++) {

  3973.             s->dc_val[0][s->block_index[i]]  = 0;

  3974.             v->mb_type[0][s->block_index[i]] = 1;

  3975.             dst_idx += i >> 2;

  3976.             val = ((cbp >> (5 - i)) & 1);

  3977.             v->a_avail = v->c_avail = 0;

  3978.             if (i == 2 || i == 3 || !s->first_slice_line)

  3979.                 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  3980.             if (i == 1 || i == 3 || s->mb_x)

  3981.                 v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  3982. 

  3983.             vc1_decode_intra_block(v, s->block[i], i, val, mquant,

  3984.                                    (i & 4) ? v->codingset2 : v->codingset);

  3985.             if ((i>3) && (s->flags & CODEC_FLAG_GRAY))

  3986.                 continue;

  3987.             v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);

  3988.             off  = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  3989.             s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i & 4) ? s->uvlinesize : s->linesize);

  3990.             // TODO: loop filter

  3991.         }

  3992.     } else {

  3993.         s->mb_intra = v->is_intra[s->mb_x] = 0;

  3994.         s->current_picture.mb_type[mb_pos + v->mb_off] = MB_TYPE_16x16;

  3995.         for (i = 0; i < 6; i++) v->mb_type[0][s->block_index[i]] = 0;

  3996.         if (idx_mbmode <= 5) { // 1-MV

  3997.             dmv_x = dmv_y = pred_flag = 0;

  3998.             if (idx_mbmode & 1) {

  3999.                 get_mvdata_interlaced(v, &dmv_x, &dmv_y, &pred_flag);

  4000.             }

  4001.             vc1_pred_mv(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], pred_flag, 0);

  4002.             vc1_mc_1mv(v, 0);

  4003.             mb_has_coeffs = !(idx_mbmode & 2);

  4004.         } else { // 4-MV

  4005.             v->fourmvbp = get_vlc2(gb, v->fourmvbp_vlc->table, VC1_4MV_BLOCK_PATTERN_VLC_BITS, 1);

  4006.             for (i = 0; i < 6; i++) {

  4007.                 if (i < 4) {

  4008.                     dmv_x = dmv_y = pred_flag = 0;

  4009.                     val   = ((v->fourmvbp >> (3 - i)) & 1);

  4010.                     if (val) {

  4011.                         get_mvdata_interlaced(v, &dmv_x, &dmv_y, &pred_flag);

  4012.                     }

  4013.                     vc1_pred_mv(v, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0], pred_flag, 0);

  4014.                     vc1_mc_4mv_luma(v, i, 0, 0);

  4015.                 } else if (i == 4)

  4016.                     vc1_mc_4mv_chroma(v, 0);

  4017.             }

  4018.             mb_has_coeffs = idx_mbmode & 1;

  4019.         }

  4020.         if (mb_has_coeffs)

  4021.             cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  4022.         if (cbp) {

  4023.             GET_MQUANT();

  4024.         }

  4025.         s->current_picture.qscale_table[mb_pos] = mquant;

  4026.         if (!v->ttmbf && cbp) {

  4027.             ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  4028.         }

  4029.         dst_idx = 0;

  4030.         for (i = 0; i < 6; i++) {

  4031.             s->dc_val[0][s->block_index[i]] = 0;

  4032.             dst_idx += i >> 2;

  4033.             val = ((cbp >> (5 - i)) & 1);

  4034.             off = (i & 4) ? 0 : (i & 1) * 8 + (i & 2) * 4 * s->linesize;

  4035.             if (val) {

  4036.                 pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,

  4037.                                          first_block, s->dest[dst_idx] + off,

  4038.                                          (i & 4) ? s->uvlinesize : s->linesize,

  4039.                                          (i & 4) && (s->flags & CODEC_FLAG_GRAY),

  4040.                                          &block_tt);

  4041.                 block_cbp |= pat << (i << 2);

  4042.                 if (!v->ttmbf && ttmb < 8) ttmb = -1;

  4043.                 first_block = 0;

  4044.             }

  4045.         }

  4046.     }

  4047.     if (s->mb_x == s->mb_width - 1)

  4048.         memmove(v->is_intra_base, v->is_intra, sizeof(v->is_intra_base[0]) * s->mb_stride);

  4049.     return 0;

  4050. }

  4051. 

  4052. /** Decode one B-frame MB (in Main profile)

  4053.  */

  4054. static void vc1_decode_b_mb(VC1Context *v)

  4055. {

  4056.     MpegEncContext *s = &v->s;

  4057.     GetBitContext *gb = &s->gb;

  4058.     int i, j;

  4059.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  4060.     int cbp = 0; /* cbp decoding stuff */

  4061.     int mqdiff, mquant; /* MB quantization */

  4062.     int ttmb = v->ttfrm; /* MB Transform type */

  4063.     int mb_has_coeffs = 0; /* last_flag */

  4064.     int index, index1; /* LUT indexes */

  4065.     int val, sign; /* temp values */

  4066.     int first_block = 1;

  4067.     int dst_idx, off;

  4068.     int skipped, direct;

  4069.     int dmv_x[2], dmv_y[2];

  4070.     int bmvtype = BMV_TYPE_BACKWARD;

  4071. 

  4072.     mquant      = v->pq; /* lossy initialization */

  4073.     s->mb_intra = 0;

  4074. 

  4075.     if (v->dmb_is_raw)

  4076.         direct = get_bits1(gb);

  4077.     else

  4078.         direct = v->direct_mb_plane[mb_pos];

  4079.     if (v->skip_is_raw)

  4080.         skipped = get_bits1(gb);

  4081.     else

  4082.         skipped = v->s.mbskip_table[mb_pos];

  4083. 

  4084.     dmv_x[0] = dmv_x[1] = dmv_y[0] = dmv_y[1] = 0;

  4085.     for (i = 0; i < 6; i++) {

  4086.         v->mb_type[0][s->block_index[i]] = 0;

  4087.         s->dc_val[0][s->block_index[i]]  = 0;

  4088.     }

  4089.     s->current_picture.qscale_table[mb_pos] = 0;

  4090. 

  4091.     if (!direct) {

  4092.         if (!skipped) {

  4093.             GET_MVDATA(dmv_x[0], dmv_y[0]);

  4094.             dmv_x[1] = dmv_x[0];

  4095.             dmv_y[1] = dmv_y[0];

  4096.         }

  4097.         if (skipped || !s->mb_intra) {

  4098.             bmvtype = decode012(gb);

  4099.             switch (bmvtype) {

  4100.             case 0:

  4101.                 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_BACKWARD : BMV_TYPE_FORWARD;

  4102.                 break;

  4103.             case 1:

  4104.                 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_FORWARD : BMV_TYPE_BACKWARD;

  4105.                 break;

  4106.             case 2:

  4107.                 bmvtype  = BMV_TYPE_INTERPOLATED;

  4108.                 dmv_x[0] = dmv_y[0] = 0;

  4109.             }

  4110.         }

  4111.     }

  4112.     for (i = 0; i < 6; i++)

  4113.         v->mb_type[0][s->block_index[i]] = s->mb_intra;

  4114. 

  4115.     if (skipped) {

  4116.         if (direct)

  4117.             bmvtype = BMV_TYPE_INTERPOLATED;

  4118.         vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  4119.         vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  4120.         return;

  4121.     }

  4122.     if (direct) {

  4123.         cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  4124.         GET_MQUANT();

  4125.         s->mb_intra = 0;

  4126.         s->current_picture.qscale_table[mb_pos] = mquant;

  4127.         if (!v->ttmbf)

  4128.             ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  4129.         dmv_x[0] = dmv_y[0] = dmv_x[1] = dmv_y[1] = 0;

  4130.         vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  4131.         vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  4132.     } else {

  4133.         if (!mb_has_coeffs && !s->mb_intra) {

  4134.             /* no coded blocks - effectively skipped */

  4135.             vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  4136.             vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  4137.             return;

  4138.         }

  4139.         if (s->mb_intra && !mb_has_coeffs) {

  4140.             GET_MQUANT();

  4141.             s->current_picture.qscale_table[mb_pos] = mquant;

  4142.             s->ac_pred = get_bits1(gb);

  4143.             cbp = 0;

  4144.             vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  4145.         } else {

  4146.             if (bmvtype == BMV_TYPE_INTERPOLATED) {

  4147.                 GET_MVDATA(dmv_x[0], dmv_y[0]);

  4148.                 if (!mb_has_coeffs) {

  4149.                     /* interpolated skipped block */

  4150.                     vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  4151.                     vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  4152.                     return;

  4153.                 }

  4154.             }

  4155.             vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  4156.             if (!s->mb_intra) {

  4157.                 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  4158.             }

  4159.             if (s->mb_intra)

  4160.                 s->ac_pred = get_bits1(gb);

  4161.             cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  4162.             GET_MQUANT();

  4163.             s->current_picture.qscale_table[mb_pos] = mquant;

  4164.             if (!v->ttmbf && !s->mb_intra && mb_has_coeffs)

  4165.                 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  4166.         }

  4167.     }

  4168.     dst_idx = 0;

  4169.     for (i = 0; i < 6; i++) {

  4170.         s->dc_val[0][s->block_index[i]] = 0;

  4171.         dst_idx += i >> 2;

  4172.         val = ((cbp >> (5 - i)) & 1);

  4173.         off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  4174.         v->mb_type[0][s->block_index[i]] = s->mb_intra;

  4175.         if (s->mb_intra) {

  4176.             /* check if prediction blocks A and C are available */

  4177.             v->a_avail = v->c_avail = 0;

  4178.             if (i == 2 || i == 3 || !s->first_slice_line)

  4179.                 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  4180.             if (i == 1 || i == 3 || s->mb_x)

  4181.                 v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  4182. 

  4183.             vc1_decode_intra_block(v, s->block[i], i, val, mquant,

  4184.                                    (i & 4) ? v->codingset2 : v->codingset);

  4185.             if ((i>3) && (s->flags & CODEC_FLAG_GRAY))

  4186.                 continue;

  4187.             v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);

  4188.             if (v->rangeredfrm)

  4189.                 for (j = 0; j < 64; j++)

  4190.                     s->block[i][j] <<= 1;

  4191.             s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);

  4192.         } else if (val) {

  4193.             vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,

  4194.                                first_block, s->dest[dst_idx] + off,

  4195.                                (i & 4) ? s->uvlinesize : s->linesize,

  4196.                                (i & 4) && (s->flags & CODEC_FLAG_GRAY), NULL);

  4197.             if (!v->ttmbf && ttmb < 8)

  4198.                 ttmb = -1;

  4199.             first_block = 0;

  4200.         }

  4201.     }

  4202. }

  4203. 

  4204. /** Decode one B-frame MB (in interlaced field B picture)

  4205.  */

  4206. static void vc1_decode_b_mb_intfi(VC1Context *v)

  4207. {

  4208.     MpegEncContext *s = &v->s;

  4209.     GetBitContext *gb = &s->gb;

  4210.     int i, j;

  4211.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  4212.     int cbp = 0; /* cbp decoding stuff */

  4213.     int mqdiff, mquant; /* MB quantization */

  4214.     int ttmb = v->ttfrm; /* MB Transform type */

  4215.     int mb_has_coeffs = 0; /* last_flag */

  4216.     int val; /* temp value */

  4217.     int first_block = 1;

  4218.     int dst_idx, off;

  4219.     int fwd;

  4220.     int dmv_x[2], dmv_y[2], pred_flag[2];

  4221.     int bmvtype = BMV_TYPE_BACKWARD;

  4222.     int idx_mbmode, interpmvp;

  4223. 

  4224.     mquant      = v->pq; /* Loosy initialization */

  4225.     s->mb_intra = 0;

  4226. 

  4227.     idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_IF_MBMODE_VLC_BITS, 2);

  4228.     if (idx_mbmode <= 1) { // intra MB

  4229.         s->mb_intra = v->is_intra[s->mb_x] = 1;

  4230.         s->current_picture.motion_val[1][s->block_index[0]][0] = 0;

  4231.         s->current_picture.motion_val[1][s->block_index[0]][1] = 0;

  4232.         s->current_picture.mb_type[mb_pos + v->mb_off]         = MB_TYPE_INTRA;

  4233.         GET_MQUANT();

  4234.         s->current_picture.qscale_table[mb_pos] = mquant;

  4235.         /* Set DC scale - y and c use the same (not sure if necessary here) */

  4236.         s->y_dc_scale = s->y_dc_scale_table[mquant];

  4237.         s->c_dc_scale = s->c_dc_scale_table[mquant];

  4238.         v->s.ac_pred  = v->acpred_plane[mb_pos] = get_bits1(gb);

  4239.         mb_has_coeffs = idx_mbmode & 1;

  4240.         if (mb_has_coeffs)

  4241.             cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_ICBPCY_VLC_BITS, 2);

  4242.         dst_idx = 0;

  4243.         for (i = 0; i < 6; i++) {

  4244.             s->dc_val[0][s->block_index[i]] = 0;

  4245.             dst_idx += i >> 2;

  4246.             val = ((cbp >> (5 - i)) & 1);

  4247.             v->mb_type[0][s->block_index[i]] = s->mb_intra;

  4248.             v->a_avail                       = v->c_avail = 0;

  4249.             if (i == 2 || i == 3 || !s->first_slice_line)

  4250.                 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  4251.             if (i == 1 || i == 3 || s->mb_x)

  4252.                 v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  4253. 

  4254.             vc1_decode_intra_block(v, s->block[i], i, val, mquant,

  4255.                                    (i & 4) ? v->codingset2 : v->codingset);

  4256.             if ((i>3) && (s->flags & CODEC_FLAG_GRAY))

  4257.                 continue;

  4258.             v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);

  4259.             if (v->rangeredfrm)

  4260.                 for (j = 0; j < 64; j++)

  4261.                     s->block[i][j] <<= 1;

  4262.             off  = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  4263.             s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i & 4) ? s->uvlinesize : s->linesize);

  4264.             // TODO: yet to perform loop filter

  4265.         }

  4266.     } else {

  4267.         s->mb_intra = v->is_intra[s->mb_x] = 0;

  4268.         s->current_picture.mb_type[mb_pos + v->mb_off] = MB_TYPE_16x16;

  4269.         for (i = 0; i < 6; i++) v->mb_type[0][s->block_index[i]] = 0;

  4270.         if (v->fmb_is_raw)

  4271.             fwd = v->forward_mb_plane[mb_pos] = get_bits1(gb);

  4272.         else

  4273.             fwd = v->forward_mb_plane[mb_pos];

  4274.         if (idx_mbmode <= 5) { // 1-MV

  4275.             dmv_x[0]     = dmv_x[1] = dmv_y[0] = dmv_y[1] = 0;

  4276.             pred_flag[0] = pred_flag[1] = 0;

  4277.             if (fwd)

  4278.                 bmvtype = BMV_TYPE_FORWARD;

  4279.             else {

  4280.                 bmvtype = decode012(gb);

  4281.                 switch (bmvtype) {

  4282.                 case 0:

  4283.                     bmvtype = BMV_TYPE_BACKWARD;

  4284.                     break;

  4285.                 case 1:

  4286.                     bmvtype = BMV_TYPE_DIRECT;

  4287.                     break;

  4288.                 case 2:

  4289.                     bmvtype   = BMV_TYPE_INTERPOLATED;

  4290.                     interpmvp = get_bits1(gb);

  4291.                 }

  4292.             }

  4293.             v->bmvtype = bmvtype;

  4294.             if (bmvtype != BMV_TYPE_DIRECT && idx_mbmode & 1) {

  4295.                 get_mvdata_interlaced(v, &dmv_x[bmvtype == BMV_TYPE_BACKWARD], &dmv_y[bmvtype == BMV_TYPE_BACKWARD], &pred_flag[bmvtype == BMV_TYPE_BACKWARD]);

  4296.             }

  4297.             if (bmvtype == BMV_TYPE_INTERPOLATED && interpmvp) {

  4298.                 get_mvdata_interlaced(v, &dmv_x[1], &dmv_y[1], &pred_flag[1]);

  4299.             }

  4300.             if (bmvtype == BMV_TYPE_DIRECT) {

  4301.                 dmv_x[0] = dmv_y[0] = pred_flag[0] = 0;

  4302.                 dmv_x[1] = dmv_y[1] = pred_flag[0] = 0;

  4303.             }

  4304.             vc1_pred_b_mv_intfi(v, 0, dmv_x, dmv_y, 1, pred_flag);

  4305.             vc1_b_mc(v, dmv_x, dmv_y, (bmvtype == BMV_TYPE_DIRECT), bmvtype);

  4306.             mb_has_coeffs = !(idx_mbmode & 2);

  4307.         } else { // 4-MV

  4308.             if (fwd)

  4309.                 bmvtype = BMV_TYPE_FORWARD;

  4310.             v->bmvtype  = bmvtype;

  4311.             v->fourmvbp = get_vlc2(gb, v->fourmvbp_vlc->table, VC1_4MV_BLOCK_PATTERN_VLC_BITS, 1);

  4312.             for (i = 0; i < 6; i++) {

  4313.                 if (i < 4) {

  4314.                     dmv_x[0] = dmv_y[0] = pred_flag[0] = 0;

  4315.                     dmv_x[1] = dmv_y[1] = pred_flag[1] = 0;

  4316.                     val = ((v->fourmvbp >> (3 - i)) & 1);

  4317.                     if (val) {

  4318.                         get_mvdata_interlaced(v, &dmv_x[bmvtype == BMV_TYPE_BACKWARD],

  4319.                                                  &dmv_y[bmvtype == BMV_TYPE_BACKWARD],

  4320.                                              &pred_flag[bmvtype == BMV_TYPE_BACKWARD]);

  4321.                     }

  4322.                     vc1_pred_b_mv_intfi(v, i, dmv_x, dmv_y, 0, pred_flag);

  4323.                     vc1_mc_4mv_luma(v, i, bmvtype == BMV_TYPE_BACKWARD, 0);

  4324.                 } else if (i == 4)

  4325.                     vc1_mc_4mv_chroma(v, bmvtype == BMV_TYPE_BACKWARD);

  4326.             }

  4327.             mb_has_coeffs = idx_mbmode & 1;

  4328.         }

  4329.         if (mb_has_coeffs)

  4330.             cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  4331.         if (cbp) {

  4332.             GET_MQUANT();

  4333.         }

  4334.         s->current_picture.qscale_table[mb_pos] = mquant;

  4335.         if (!v->ttmbf && cbp) {

  4336.             ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  4337.         }

  4338.         dst_idx = 0;

  4339.         for (i = 0; i < 6; i++) {

  4340.             s->dc_val[0][s->block_index[i]] = 0;

  4341.             dst_idx += i >> 2;

  4342.             val = ((cbp >> (5 - i)) & 1);

  4343.             off = (i & 4) ? 0 : (i & 1) * 8 + (i & 2) * 4 * s->linesize;

  4344.             if (val) {

  4345.                 vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,

  4346.                                    first_block, s->dest[dst_idx] + off,

  4347.                                    (i & 4) ? s->uvlinesize : s->linesize,

  4348.                                    (i & 4) && (s->flags & CODEC_FLAG_GRAY), NULL);

  4349.                 if (!v->ttmbf && ttmb < 8)

  4350.                     ttmb = -1;

  4351.                 first_block = 0;

  4352.             }

  4353.         }

  4354.     }

  4355. }

  4356. 

  4357. /** Decode one B-frame MB (in interlaced frame B picture)

  4358.  */

  4359. static int vc1_decode_b_mb_intfr(VC1Context *v)

  4360. {

  4361.     MpegEncContext *s = &v->s;

  4362.     GetBitContext *gb = &s->gb;

  4363.     int i, j;

  4364.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  4365.     int cbp = 0; /* cbp decoding stuff */

  4366.     int mqdiff, mquant; /* MB quantization */

  4367.     int ttmb = v->ttfrm; /* MB Transform type */

  4368.     int mvsw = 0; /* motion vector switch */

  4369.     int mb_has_coeffs = 1; /* last_flag */

  4370.     int dmv_x, dmv_y; /* Differential MV components */

  4371.     int val; /* temp value */

  4372.     int first_block = 1;

  4373.     int dst_idx, off;

  4374.     int skipped, direct, twomv = 0;

  4375.     int block_cbp = 0, pat, block_tt = 0;

  4376.     int idx_mbmode = 0, mvbp;

  4377.     int stride_y, fieldtx;

  4378.     int bmvtype = BMV_TYPE_BACKWARD;

  4379.     int dir, dir2;

  4380. 

  4381.     mquant = v->pq; /* Lossy initialization */

  4382.     s->mb_intra = 0;

  4383.     if (v->skip_is_raw)

  4384.         skipped = get_bits1(gb);

  4385.     else

  4386.         skipped = v->s.mbskip_table[mb_pos];

  4387. 

  4388.     if (!skipped) {

  4389.         idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_INTFR_NON4MV_MBMODE_VLC_BITS, 2);

  4390.         if (ff_vc1_mbmode_intfrp[0][idx_mbmode][0] == MV_PMODE_INTFR_2MV_FIELD) {

  4391.             twomv = 1;

  4392.             v->blk_mv_type[s->block_index[0]] = 1;

  4393.             v->blk_mv_type[s->block_index[1]] = 1;

  4394.             v->blk_mv_type[s->block_index[2]] = 1;

  4395.             v->blk_mv_type[s->block_index[3]] = 1;

  4396.         } else {

  4397.             v->blk_mv_type[s->block_index[0]] = 0;

  4398.             v->blk_mv_type[s->block_index[1]] = 0;

  4399.             v->blk_mv_type[s->block_index[2]] = 0;

  4400.             v->blk_mv_type[s->block_index[3]] = 0;

  4401.         }

  4402.     }

  4403. 

  4404.     if (v->dmb_is_raw)

  4405.         direct = get_bits1(gb);

  4406.     else

  4407.         direct = v->direct_mb_plane[mb_pos];

  4408. 

  4409.     if (direct) {

  4410.         s->mv[0][0][0] = s->current_picture.motion_val[0][s->block_index[0]][0] = scale_mv(s->next_picture.motion_val[1][s->block_index[0]][0], v->bfraction, 0, s->quarter_sample);

  4411.         s->mv[0][0][1] = s->current_picture.motion_val[0][s->block_index[0]][1] = scale_mv(s->next_picture.motion_val[1][s->block_index[0]][1], v->bfraction, 0, s->quarter_sample);

  4412.         s->mv[1][0][0] = s->current_picture.motion_val[1][s->block_index[0]][0] = scale_mv(s->next_picture.motion_val[1][s->block_index[0]][0], v->bfraction, 1, s->quarter_sample);

  4413.         s->mv[1][0][1] = s->current_picture.motion_val[1][s->block_index[0]][1] = scale_mv(s->next_picture.motion_val[1][s->block_index[0]][1], v->bfraction, 1, s->quarter_sample);

  4414. 

  4415.         if (twomv) {

  4416.             s->mv[0][2][0] = s->current_picture.motion_val[0][s->block_index[2]][0] = scale_mv(s->next_picture.motion_val[1][s->block_index[2]][0], v->bfraction, 0, s->quarter_sample);

  4417.             s->mv[0][2][1] = s->current_picture.motion_val[0][s->block_index[2]][1] = scale_mv(s->next_picture.motion_val[1][s->block_index[2]][1], v->bfraction, 0, s->quarter_sample);

  4418.             s->mv[1][2][0] = s->current_picture.motion_val[1][s->block_index[2]][0] = scale_mv(s->next_picture.motion_val[1][s->block_index[2]][0], v->bfraction, 1, s->quarter_sample);

  4419.             s->mv[1][2][1] = s->current_picture.motion_val[1][s->block_index[2]][1] = scale_mv(s->next_picture.motion_val[1][s->block_index[2]][1], v->bfraction, 1, s->quarter_sample);

  4420. 

  4421.             for (i = 1; i < 4; i += 2) {

  4422.                 s->mv[0][i][0] = s->current_picture.motion_val[0][s->block_index[i]][0] = s->mv[0][i-1][0];

  4423.                 s->mv[0][i][1] = s->current_picture.motion_val[0][s->block_index[i]][1] = s->mv[0][i-1][1];

  4424.                 s->mv[1][i][0] = s->current_picture.motion_val[1][s->block_index[i]][0] = s->mv[1][i-1][0];

  4425.                 s->mv[1][i][1] = s->current_picture.motion_val[1][s->block_index[i]][1] = s->mv[1][i-1][1];

  4426.             }

  4427.         } else {

  4428.             for (i = 1; i < 4; i++) {

  4429.                 s->mv[0][i][0] = s->current_picture.motion_val[0][s->block_index[i]][0] = s->mv[0][0][0];

  4430.                 s->mv[0][i][1] = s->current_picture.motion_val[0][s->block_index[i]][1] = s->mv[0][0][1];

  4431.                 s->mv[1][i][0] = s->current_picture.motion_val[1][s->block_index[i]][0] = s->mv[1][0][0];

  4432.                 s->mv[1][i][1] = s->current_picture.motion_val[1][s->block_index[i]][1] = s->mv[1][0][1];

  4433.             }

  4434.         }

  4435.     }

  4436. 

  4437.     if (ff_vc1_mbmode_intfrp[0][idx_mbmode][0] == MV_PMODE_INTFR_INTRA) { // intra MB

  4438.         for (i = 0; i < 4; i++) {

  4439.             s->mv[0][i][0] = s->current_picture.motion_val[0][s->block_index[i]][0] = 0;

  4440.             s->mv[0][i][1] = s->current_picture.motion_val[0][s->block_index[i]][1] = 0;

  4441.             s->mv[1][i][0] = s->current_picture.motion_val[1][s->block_index[i]][0] = 0;

  4442.             s->mv[1][i][1] = s->current_picture.motion_val[1][s->block_index[i]][1] = 0;

  4443.         }

  4444.         s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;

  4445.         s->mb_intra = v->is_intra[s->mb_x] = 1;

  4446.         for (i = 0; i < 6; i++)

  4447.             v->mb_type[0][s->block_index[i]] = 1;

  4448.         fieldtx = v->fieldtx_plane[mb_pos] = get_bits1(gb);

  4449.         mb_has_coeffs = get_bits1(gb);

  4450.         if (mb_has_coeffs)

  4451.             cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  4452.         v->s.ac_pred = v->acpred_plane[mb_pos] = get_bits1(gb);

  4453.         GET_MQUANT();

  4454.         s->current_picture.qscale_table[mb_pos] = mquant;

  4455.         /* Set DC scale - y and c use the same (not sure if necessary here) */

  4456.         s->y_dc_scale = s->y_dc_scale_table[mquant];

  4457.         s->c_dc_scale = s->c_dc_scale_table[mquant];

  4458.         dst_idx = 0;

  4459.         for (i = 0; i < 6; i++) {

  4460.             s->dc_val[0][s->block_index[i]] = 0;

  4461.             dst_idx += i >> 2;

  4462.             val = ((cbp >> (5 - i)) & 1);

  4463.             v->mb_type[0][s->block_index[i]] = s->mb_intra;

  4464.             v->a_avail = v->c_avail = 0;

  4465.             if (i == 2 || i == 3 || !s->first_slice_line)

  4466.                 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  4467.             if (i == 1 || i == 3 || s->mb_x)

  4468.                 v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  4469. 

  4470.             vc1_decode_intra_block(v, s->block[i], i, val, mquant,

  4471.                                    (i & 4) ? v->codingset2 : v->codingset);

  4472.             if (i > 3 && (s->flags & CODEC_FLAG_GRAY))

  4473.                 continue;

  4474.             v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);

  4475.             if (i < 4) {

  4476.                 stride_y = s->linesize << fieldtx;

  4477.                 off = (fieldtx) ? ((i & 1) * 8) + ((i & 2) >> 1) * s->linesize : (i & 1) * 8 + 4 * (i & 2) * s->linesize;

  4478.             } else {

  4479.                 stride_y = s->uvlinesize;

  4480.                 off = 0;

  4481.             }

  4482.             s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, stride_y);

  4483.         }

  4484.     } else {

  4485.         s->mb_intra = v->is_intra[s->mb_x] = 0;

  4486.         if (!direct) {

  4487.             if (skipped || !s->mb_intra) {

  4488.                 bmvtype = decode012(gb);

  4489.                 switch (bmvtype) {

  4490.                 case 0:

  4491.                     bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_BACKWARD : BMV_TYPE_FORWARD;

  4492.                     break;

  4493.                 case 1:

  4494.                     bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_FORWARD : BMV_TYPE_BACKWARD;

  4495.                     break;

  4496.                 case 2:

  4497.                     bmvtype  = BMV_TYPE_INTERPOLATED;

  4498.                 }

  4499.             }

  4500. 

  4501.             if (twomv && bmvtype != BMV_TYPE_INTERPOLATED)

  4502.                 mvsw = get_bits1(gb);

  4503.         }

  4504. 

  4505.         if (!skipped) { // inter MB

  4506.             mb_has_coeffs = ff_vc1_mbmode_intfrp[0][idx_mbmode][3];

  4507.             if (mb_has_coeffs)

  4508.                 cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  4509.             if (!direct) {

  4510.                 if (bmvtype == BMV_TYPE_INTERPOLATED & twomv) {

  4511.                     v->fourmvbp = get_vlc2(gb, v->fourmvbp_vlc->table, VC1_4MV_BLOCK_PATTERN_VLC_BITS, 1);

  4512.                 } else if (bmvtype == BMV_TYPE_INTERPOLATED | twomv) {

  4513.                     v->twomvbp = get_vlc2(gb, v->twomvbp_vlc->table, VC1_2MV_BLOCK_PATTERN_VLC_BITS, 1);

  4514.                 }

  4515.             }

  4516. 

  4517.             for (i = 0; i < 6; i++)

  4518.                 v->mb_type[0][s->block_index[i]] = 0;

  4519.             fieldtx = v->fieldtx_plane[mb_pos] = ff_vc1_mbmode_intfrp[0][idx_mbmode][1];

  4520.             /* for all motion vector read MVDATA and motion compensate each block */

  4521.             dst_idx = 0;

  4522.             if (direct) {

  4523.                 if (twomv) {

  4524.                     for (i = 0; i < 4; i++) {

  4525.                         vc1_mc_4mv_luma(v, i, 0, 0);

  4526.                         vc1_mc_4mv_luma(v, i, 1, 1);

  4527.                     }

  4528.                     vc1_mc_4mv_chroma4(v);

  4529.                 } else {

  4530.                     vc1_mc_1mv(v, 0);

  4531.                     vc1_interp_mc(v);

  4532.                 }

  4533.             } else if (twomv && bmvtype == BMV_TYPE_INTERPOLATED) {

  4534.                 mvbp = v->fourmvbp;

  4535.                 for (i = 0; i < 4; i++) {

  4536.                     dir = i==1 || i==3;

  4537.                     dmv_x = dmv_y = 0;

  4538.                     val = ((mvbp >> (3 - i)) & 1);

  4539.                     if (val)

  4540.                         get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);

  4541.                     j = i > 1 ? 2 : 0;

  4542.                     vc1_pred_mv_intfr(v, j, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0], dir);

  4543.                     vc1_mc_4mv_luma(v, j, dir, dir);

  4544.                     vc1_mc_4mv_luma(v, j+1, dir, dir);

  4545.                 }

  4546. 

  4547.                 vc1_mc_4mv_chroma4(v);

  4548.             } else if (bmvtype == BMV_TYPE_INTERPOLATED) {

  4549.                 mvbp = v->twomvbp;

  4550.                 dmv_x = dmv_y = 0;

  4551.                 if (mvbp & 2)

  4552.                     get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);

  4553. 

  4554.                 vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], 0);

  4555.                 vc1_mc_1mv(v, 0);

  4556. 

  4557.                 dmv_x = dmv_y = 0;

  4558.                 if (mvbp & 1)

  4559.                     get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);

  4560. 

  4561.                 vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], 1);

  4562.                 vc1_interp_mc(v);

  4563.             } else if (twomv) {

  4564.                 dir = bmvtype == BMV_TYPE_BACKWARD;

  4565.                 dir2 = dir;

  4566.                 if (mvsw)

  4567.                     dir2 = !dir;

  4568.                 mvbp = v->twomvbp;

  4569.                 dmv_x = dmv_y = 0;

  4570.                 if (mvbp & 2)

  4571.                     get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);

  4572.                 vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0], dir);

  4573. 

  4574.                 dmv_x = dmv_y = 0;

  4575.                 if (mvbp & 1)

  4576.                     get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);

  4577.                 vc1_pred_mv_intfr(v, 2, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0], dir2);

  4578. 

  4579.                 if (mvsw) {

  4580.                     for (i = 0; i < 2; i++) {

  4581.                         s->mv[dir][i+2][0] = s->mv[dir][i][0] = s->current_picture.motion_val[dir][s->block_index[i+2]][0] = s->current_picture.motion_val[dir][s->block_index[i]][0];

  4582.                         s->mv[dir][i+2][1] = s->mv[dir][i][1] = s->current_picture.motion_val[dir][s->block_index[i+2]][1] = s->current_picture.motion_val[dir][s->block_index[i]][1];

  4583.                         s->mv[dir2][i+2][0] = s->mv[dir2][i][0] = s->current_picture.motion_val[dir2][s->block_index[i]][0] = s->current_picture.motion_val[dir2][s->block_index[i+2]][0];

  4584.                         s->mv[dir2][i+2][1] = s->mv[dir2][i][1] = s->current_picture.motion_val[dir2][s->block_index[i]][1] = s->current_picture.motion_val[dir2][s->block_index[i+2]][1];

  4585.                     }

  4586.                 } else {

  4587.                     vc1_pred_mv_intfr(v, 0, 0, 0, 2, v->range_x, v->range_y, v->mb_type[0], !dir);

  4588.                     vc1_pred_mv_intfr(v, 2, 0, 0, 2, v->range_x, v->range_y, v->mb_type[0], !dir);

  4589.                 }

  4590. 

  4591.                 vc1_mc_4mv_luma(v, 0, dir, 0);

  4592.                 vc1_mc_4mv_luma(v, 1, dir, 0);

  4593.                 vc1_mc_4mv_luma(v, 2, dir2, 0);

  4594.                 vc1_mc_4mv_luma(v, 3, dir2, 0);

  4595.                 vc1_mc_4mv_chroma4(v);

  4596.             } else {

  4597.                 dir = bmvtype == BMV_TYPE_BACKWARD;

  4598. 

  4599.                 mvbp = ff_vc1_mbmode_intfrp[0][idx_mbmode][2];

  4600.                 dmv_x = dmv_y = 0;

  4601.                 if (mvbp)

  4602.                     get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);

  4603. 

  4604.                 vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], dir);

  4605.                 v->blk_mv_type[s->block_index[0]] = 1;

  4606.                 v->blk_mv_type[s->block_index[1]] = 1;

  4607.                 v->blk_mv_type[s->block_index[2]] = 1;

  4608.                 v->blk_mv_type[s->block_index[3]] = 1;

  4609.                 vc1_pred_mv_intfr(v, 0, 0, 0, 2, v->range_x, v->range_y, 0, !dir);

  4610.                 for (i = 0; i < 2; i++) {

  4611.                     s->mv[!dir][i+2][0] = s->mv[!dir][i][0] = s->current_picture.motion_val[!dir][s->block_index[i+2]][0] = s->current_picture.motion_val[!dir][s->block_index[i]][0];

  4612.                     s->mv[!dir][i+2][1] = s->mv[!dir][i][1] = s->current_picture.motion_val[!dir][s->block_index[i+2]][1] = s->current_picture.motion_val[!dir][s->block_index[i]][1];

  4613.                 }

  4614.                 vc1_mc_1mv(v, dir);

  4615.             }

  4616. 

  4617.             if (cbp)

  4618.                 GET_MQUANT();  // p. 227

  4619.             s->current_picture.qscale_table[mb_pos] = mquant;

  4620.             if (!v->ttmbf && cbp)

  4621.                 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  4622.             for (i = 0; i < 6; i++) {

  4623.                 s->dc_val[0][s->block_index[i]] = 0;

  4624.                 dst_idx += i >> 2;

  4625.                 val = ((cbp >> (5 - i)) & 1);

  4626.                 if (!fieldtx)

  4627.                     off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  4628.                 else

  4629.                     off = (i & 4) ? 0 : ((i & 1) * 8 + ((i > 1) * s->linesize));

  4630.                 if (val) {

  4631.                     pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,

  4632.                                              first_block, s->dest[dst_idx] + off,

  4633.                                              (i & 4) ? s->uvlinesize : (s->linesize << fieldtx),

  4634.                                              (i & 4) && (s->flags & CODEC_FLAG_GRAY), &block_tt);

  4635.                     block_cbp |= pat << (i << 2);

  4636.                     if (!v->ttmbf && ttmb < 8)

  4637.                         ttmb = -1;

  4638.                     first_block = 0;

  4639.                 }

  4640.             }

  4641. 

  4642.         } else { // skipped

  4643.             dir = 0;

  4644.             for (i = 0; i < 6; i++) {

  4645.                 v->mb_type[0][s->block_index[i]] = 0;

  4646.                 s->dc_val[0][s->block_index[i]] = 0;

  4647.             }

  4648.             s->current_picture.mb_type[mb_pos]      = MB_TYPE_SKIP;

  4649.             s->current_picture.qscale_table[mb_pos] = 0;

  4650.             v->blk_mv_type[s->block_index[0]] = 0;

  4651.             v->blk_mv_type[s->block_index[1]] = 0;

  4652.             v->blk_mv_type[s->block_index[2]] = 0;

  4653.             v->blk_mv_type[s->block_index[3]] = 0;

  4654. 

  4655.             if (!direct) {

  4656.                 if (bmvtype == BMV_TYPE_INTERPOLATED) {

  4657.                     vc1_pred_mv_intfr(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0], 0);

  4658.                     vc1_pred_mv_intfr(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0], 1);

  4659.                 } else {

  4660.                     dir = bmvtype == BMV_TYPE_BACKWARD;

  4661.                     vc1_pred_mv_intfr(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0], dir);

  4662.                     if (mvsw) {

  4663.                         int dir2 = dir;

  4664.                         if (mvsw)

  4665.                             dir2 = !dir;

  4666.                         for (i = 0; i < 2; i++) {

  4667.                             s->mv[dir][i+2][0] = s->mv[dir][i][0] = s->current_picture.motion_val[dir][s->block_index[i+2]][0] = s->current_picture.motion_val[dir][s->block_index[i]][0];

  4668.                             s->mv[dir][i+2][1] = s->mv[dir][i][1] = s->current_picture.motion_val[dir][s->block_index[i+2]][1] = s->current_picture.motion_val[dir][s->block_index[i]][1];

  4669.                             s->mv[dir2][i+2][0] = s->mv[dir2][i][0] = s->current_picture.motion_val[dir2][s->block_index[i]][0] = s->current_picture.motion_val[dir2][s->block_index[i+2]][0];

  4670.                             s->mv[dir2][i+2][1] = s->mv[dir2][i][1] = s->current_picture.motion_val[dir2][s->block_index[i]][1] = s->current_picture.motion_val[dir2][s->block_index[i+2]][1];

  4671.                         }

  4672.                     } else {

  4673.                         v->blk_mv_type[s->block_index[0]] = 1;

  4674.                         v->blk_mv_type[s->block_index[1]] = 1;

  4675.                         v->blk_mv_type[s->block_index[2]] = 1;

  4676.                         v->blk_mv_type[s->block_index[3]] = 1;

  4677.                         vc1_pred_mv_intfr(v, 0, 0, 0, 2, v->range_x, v->range_y, 0, !dir);

  4678.                         for (i = 0; i < 2; i++) {

  4679.                             s->mv[!dir][i+2][0] = s->mv[!dir][i][0] = s->current_picture.motion_val[!dir][s->block_index[i+2]][0] = s->current_picture.motion_val[!dir][s->block_index[i]][0];

  4680.                             s->mv[!dir][i+2][1] = s->mv[!dir][i][1] = s->current_picture.motion_val[!dir][s->block_index[i+2]][1] = s->current_picture.motion_val[!dir][s->block_index[i]][1];

  4681.                         }

  4682.                     }

  4683.                 }

  4684.             }

  4685. 

  4686.             vc1_mc_1mv(v, dir);

  4687.             if (direct || bmvtype == BMV_TYPE_INTERPOLATED) {

  4688.                 vc1_interp_mc(v);

  4689.             }

  4690.         }

  4691.     }

  4692.     if (s->mb_x == s->mb_width - 1)

  4693.         memmove(v->is_intra_base, v->is_intra, sizeof(v->is_intra_base[0]) * s->mb_stride);

  4694.     v->cbp[s->mb_x]      = block_cbp;

  4695.     v->ttblk[s->mb_x]    = block_tt;

  4696.     return 0;

  4697. }

  4698. 

  4699. /** Decode blocks of I-frame

  4700.  */

  4701. static void vc1_decode_i_blocks(VC1Context *v)

  4702. {

  4703.     int k, j;

  4704.     MpegEncContext *s = &v->s;

  4705.     int cbp, val;

  4706.     uint8_t *coded_val;

  4707.     int mb_pos;

  4708. 

  4709.     /* select codingmode used for VLC tables selection */

  4710.     switch (v->y_ac_table_index) {

  4711.     case 0:

  4712.         v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;

  4713.         break;

  4714.     case 1:

  4715.         v->codingset = CS_HIGH_MOT_INTRA;

  4716.         break;

  4717.     case 2:

  4718.         v->codingset = CS_MID_RATE_INTRA;

  4719.         break;

  4720.     }

  4721. 

  4722.     switch (v->c_ac_table_index) {

  4723.     case 0:

  4724.         v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;

  4725.         break;

  4726.     case 1:

  4727.         v->codingset2 = CS_HIGH_MOT_INTER;

  4728.         break;

  4729.     case 2:

  4730.         v->codingset2 = CS_MID_RATE_INTER;

  4731.         break;

  4732.     }

  4733. 

  4734.     /* Set DC scale - y and c use the same */

  4735.     s->y_dc_scale = s->y_dc_scale_table[v->pq];

  4736.     s->c_dc_scale = s->c_dc_scale_table[v->pq];

  4737. 

  4738.     //do frame decode

  4739.     s->mb_x = s->mb_y = 0;

  4740.     s->mb_intra         = 1;

  4741.     s->first_slice_line = 1;

  4742.     for (s->mb_y = 0; s->mb_y < s->end_mb_y; s->mb_y++) {

  4743.         s->mb_x = 0;

  4744.         init_block_index(v);

  4745.         for (; s->mb_x < v->end_mb_x; s->mb_x++) {

  4746.             uint8_t *dst[6];

  4747.             ff_update_block_index(s);

  4748.             dst[0] = s->dest[0];

  4749.             dst[1] = dst[0] + 8;

  4750.             dst[2] = s->dest[0] + s->linesize * 8;

  4751.             dst[3] = dst[2] + 8;

  4752.             dst[4] = s->dest[1];

  4753.             dst[5] = s->dest[2];

  4754.             s->dsp.clear_blocks(s->block[0]);

  4755.             mb_pos = s->mb_x + s->mb_y * s->mb_width;

  4756.             s->current_picture.mb_type[mb_pos]                     = MB_TYPE_INTRA;

  4757.             s->current_picture.qscale_table[mb_pos]                = v->pq;

  4758.             s->current_picture.motion_val[1][s->block_index[0]][0] = 0;

  4759.             s->current_picture.motion_val[1][s->block_index[0]][1] = 0;

  4760. 

  4761.             // do actual MB decoding and displaying

  4762.             cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);

  4763.             v->s.ac_pred = get_bits1(&v->s.gb);

  4764. 

  4765.             for (k = 0; k < 6; k++) {

  4766.                 val = ((cbp >> (5 - k)) & 1);

  4767. 

  4768.                 if (k < 4) {

  4769.                     int pred   = vc1_coded_block_pred(&v->s, k, &coded_val);

  4770.                     val        = val ^ pred;

  4771.                     *coded_val = val;

  4772.                 }

  4773.                 cbp |= val << (5 - k);

  4774. 

  4775.                 vc1_decode_i_block(v, s->block[k], k, val, (k < 4) ? v->codingset : v->codingset2);

  4776. 

  4777.                 if (k > 3 && (s->flags & CODEC_FLAG_GRAY))

  4778.                     continue;

  4779.                 v->vc1dsp.vc1_inv_trans_8x8(s->block[k]);

  4780.                 if (v->pq >= 9 && v->overlap) {

  4781.                     if (v->rangeredfrm)

  4782.                         for (j = 0; j < 64; j++)

  4783.                             s->block[k][j] <<= 1;

  4784.                     s->dsp.put_signed_pixels_clamped(s->block[k], dst[k], k & 4 ? s->uvlinesize : s->linesize);

  4785.                 } else {

  4786.                     if (v->rangeredfrm)

  4787.                         for (j = 0; j < 64; j++)

  4788.                             s->block[k][j] = (s->block[k][j] - 64) << 1;

  4789.                     s->dsp.put_pixels_clamped(s->block[k], dst[k], k & 4 ? s->uvlinesize : s->linesize);

  4790.                 }

  4791.             }

  4792. 

  4793.             if (v->pq >= 9 && v->overlap) {

  4794.                 if (s->mb_x) {

  4795.                     v->vc1dsp.vc1_h_overlap(s->dest[0], s->linesize);

  4796.                     v->vc1dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);

  4797.                     if (!(s->flags & CODEC_FLAG_GRAY)) {

  4798.                         v->vc1dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);

  4799.                         v->vc1dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);

  4800.                     }

  4801.                 }

  4802.                 v->vc1dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);

  4803.                 v->vc1dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);

  4804.                 if (!s->first_slice_line) {

  4805.                     v->vc1dsp.vc1_v_overlap(s->dest[0], s->linesize);

  4806.                     v->vc1dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);

  4807.                     if (!(s->flags & CODEC_FLAG_GRAY)) {

  4808.                         v->vc1dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);

  4809.                         v->vc1dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);

  4810.                     }

  4811.                 }

  4812.                 v->vc1dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);

  4813.                 v->vc1dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);

  4814.             }

  4815.             if (v->s.loop_filter) vc1_loop_filter_iblk(v, v->pq);

  4816. 

  4817.             if (get_bits_count(&s->gb) > v->bits) {

  4818.                 ff_er_add_slice(&s->er, 0, 0, s->mb_x, s->mb_y, ER_MB_ERROR);

  4819.                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n",

  4820.                        get_bits_count(&s->gb), v->bits);

  4821.                 return;

  4822.             }

  4823.         }

  4824.         if (!v->s.loop_filter)

  4825.             ff_mpeg_draw_horiz_band(s, s->mb_y * 16, 16);

  4826.         else if (s->mb_y)

  4827.             ff_mpeg_draw_horiz_band(s, (s->mb_y - 1) * 16, 16);

  4828. 

  4829.         s->first_slice_line = 0;

  4830.     }

  4831.     if (v->s.loop_filter)

  4832.         ff_mpeg_draw_horiz_band(s, (s->end_mb_y - 1) * 16, 16);

  4833. 

  4834.     /* This is intentionally mb_height and not end_mb_y - unlike in advanced

  4835.      * profile, these only differ are when decoding MSS2 rectangles. */

  4836.     ff_er_add_slice(&s->er, 0, 0, s->mb_width - 1, s->mb_height - 1, ER_MB_END);

  4837. }

  4838. 

  4839. /** Decode blocks of I-frame for advanced profile

  4840.  */

  4841. static void vc1_decode_i_blocks_adv(VC1Context *v)

  4842. {

  4843.     int k;

  4844.     MpegEncContext *s = &v->s;

  4845.     int cbp, val;

  4846.     uint8_t *coded_val;

  4847.     int mb_pos;

  4848.     int mquant = v->pq;

  4849.     int mqdiff;

  4850.     GetBitContext *gb = &s->gb;

  4851. 

  4852.     /* select codingmode used for VLC tables selection */

  4853.     switch (v->y_ac_table_index) {

  4854.     case 0:

  4855.         v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;

  4856.         break;

  4857.     case 1:

  4858.         v->codingset = CS_HIGH_MOT_INTRA;

  4859.         break;

  4860.     case 2:

  4861.         v->codingset = CS_MID_RATE_INTRA;

  4862.         break;

  4863.     }

  4864. 

  4865.     switch (v->c_ac_table_index) {

  4866.     case 0:

  4867.         v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;

  4868.         break;

  4869.     case 1:

  4870.         v->codingset2 = CS_HIGH_MOT_INTER;

  4871.         break;

  4872.     case 2:

  4873.         v->codingset2 = CS_MID_RATE_INTER;

  4874.         break;

  4875.     }

  4876. 

  4877.     // do frame decode

  4878.     s->mb_x             = s->mb_y = 0;

  4879.     s->mb_intra         = 1;

  4880.     s->first_slice_line = 1;

  4881.     s->mb_y             = s->start_mb_y;

  4882.     if (s->start_mb_y) {

  4883.         s->mb_x = 0;

  4884.         init_block_index(v);

  4885.         memset(&s->coded_block[s->block_index[0] - s->b8_stride], 0,

  4886.                (1 + s->b8_stride) * sizeof(*s->coded_block));

  4887.     }

  4888.     for (; s->mb_y < s->end_mb_y; s->mb_y++) {

  4889.         s->mb_x = 0;

  4890.         init_block_index(v);

  4891.         for (;s->mb_x < s->mb_width; s->mb_x++) {

  4892.             int16_t (*block)[64] = v->block[v->cur_blk_idx];

  4893.             ff_update_block_index(s);

  4894.             s->dsp.clear_blocks(block[0]);

  4895.             mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  4896.             s->current_picture.mb_type[mb_pos + v->mb_off]                         = MB_TYPE_INTRA;

  4897.             s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][0] = 0;

  4898.             s->current_picture.motion_val[1][s->block_index[0] + v->blocks_off][1] = 0;

  4899. 

  4900.             // do actual MB decoding and displaying

  4901.             if (v->fieldtx_is_raw)

  4902.                 v->fieldtx_plane[mb_pos] = get_bits1(&v->s.gb);

  4903.             cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);

  4904.             if ( v->acpred_is_raw)

  4905.                 v->s.ac_pred = get_bits1(&v->s.gb);

  4906.             else

  4907.                 v->s.ac_pred = v->acpred_plane[mb_pos];

  4908. 

  4909.             if (v->condover == CONDOVER_SELECT && v->overflg_is_raw)

  4910.                 v->over_flags_plane[mb_pos] = get_bits1(&v->s.gb);

  4911. 

  4912.             GET_MQUANT();

  4913. 

  4914.             s->current_picture.qscale_table[mb_pos] = mquant;

  4915.             /* Set DC scale - y and c use the same */

  4916.             s->y_dc_scale = s->y_dc_scale_table[mquant];

  4917.             s->c_dc_scale = s->c_dc_scale_table[mquant];

  4918. 

  4919.             for (k = 0; k < 6; k++) {

  4920.                 val = ((cbp >> (5 - k)) & 1);

  4921. 

  4922.                 if (k < 4) {

  4923.                     int pred   = vc1_coded_block_pred(&v->s, k, &coded_val);

  4924.                     val        = val ^ pred;

  4925.                     *coded_val = val;

  4926.                 }

  4927.                 cbp |= val << (5 - k);

  4928. 

  4929.                 v->a_avail = !s->first_slice_line || (k == 2 || k == 3);

  4930.                 v->c_avail = !!s->mb_x || (k == 1 || k == 3);

  4931. 

  4932.                 vc1_decode_i_block_adv(v, block[k], k, val,

  4933.                                        (k < 4) ? v->codingset : v->codingset2, mquant);

  4934. 

  4935.                 if (k > 3 && (s->flags & CODEC_FLAG_GRAY))

  4936.                     continue;

  4937.                 v->vc1dsp.vc1_inv_trans_8x8(block[k]);

  4938.             }

  4939. 

  4940.             vc1_smooth_overlap_filter_iblk(v);

  4941.             vc1_put_signed_blocks_clamped(v);

  4942.             if (v->s.loop_filter) vc1_loop_filter_iblk_delayed(v, v->pq);

  4943. 

  4944.             if (get_bits_count(&s->gb) > v->bits) {

  4945.                 // TODO: may need modification to handle slice coding

  4946.                 ff_er_add_slice(&s->er, 0, s->start_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR);

  4947.                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n",

  4948.                        get_bits_count(&s->gb), v->bits);

  4949.                 return;

  4950.             }

  4951.         }

  4952.         if (!v->s.loop_filter)

  4953.             ff_mpeg_draw_horiz_band(s, s->mb_y * 16, 16);

  4954.         else if (s->mb_y)

  4955.             ff_mpeg_draw_horiz_band(s, (s->mb_y-1) * 16, 16);

  4956.         s->first_slice_line = 0;

  4957.     }

  4958. 

  4959.     /* raw bottom MB row */

  4960.     s->mb_x = 0;

  4961.     init_block_index(v);

  4962. 

  4963.     for (;s->mb_x < s->mb_width; s->mb_x++) {

  4964.         ff_update_block_index(s);

  4965.         vc1_put_signed_blocks_clamped(v);

  4966.         if (v->s.loop_filter)

  4967.             vc1_loop_filter_iblk_delayed(v, v->pq);

  4968.     }

  4969.     if (v->s.loop_filter)

  4970.         ff_mpeg_draw_horiz_band(s, (s->end_mb_y-1)*16, 16);

  4971.     ff_er_add_slice(&s->er, 0, s->start_mb_y << v->field_mode, s->mb_width - 1,

  4972.                     (s->end_mb_y << v->field_mode) - 1, ER_MB_END);

  4973. }

  4974. 

  4975. static void vc1_decode_p_blocks(VC1Context *v)

  4976. {

  4977.     MpegEncContext *s = &v->s;

  4978.     int apply_loop_filter;

  4979. 

  4980.     /* select codingmode used for VLC tables selection */

  4981.     switch (v->c_ac_table_index) {

  4982.     case 0:

  4983.         v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;

  4984.         break;

  4985.     case 1:

  4986.         v->codingset = CS_HIGH_MOT_INTRA;

  4987.         break;

  4988.     case 2:

  4989.         v->codingset = CS_MID_RATE_INTRA;

  4990.         break;

  4991.     }

  4992. 

  4993.     switch (v->c_ac_table_index) {

  4994.     case 0:

  4995.         v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;

  4996.         break;

  4997.     case 1:

  4998.         v->codingset2 = CS_HIGH_MOT_INTER;

  4999.         break;

  5000.     case 2:

  5001.         v->codingset2 = CS_MID_RATE_INTER;

  5002.         break;

  5003.     }

  5004. 

  5005.     apply_loop_filter   = s->loop_filter && !(s->avctx->skip_loop_filter >= AVDISCARD_NONKEY) &&

  5006.                           v->fcm == PROGRESSIVE;

  5007.     s->first_slice_line = 1;

  5008.     memset(v->cbp_base, 0, sizeof(v->cbp_base[0])*2*s->mb_stride);

  5009.     for (s->mb_y = s->start_mb_y; s->mb_y < s->end_mb_y; s->mb_y++) {

  5010.         s->mb_x = 0;

  5011.         init_block_index(v);

  5012.         for (; s->mb_x < s->mb_width; s->mb_x++) {

  5013.             ff_update_block_index(s);

  5014. 

  5015.             if (v->fcm == ILACE_FIELD)

  5016.                 vc1_decode_p_mb_intfi(v);

  5017.             else if (v->fcm == ILACE_FRAME)

  5018.                 vc1_decode_p_mb_intfr(v);

  5019.             else vc1_decode_p_mb(v);

  5020.             if (s->mb_y != s->start_mb_y && apply_loop_filter)

  5021.                 vc1_apply_p_loop_filter(v);

  5022.             if (get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {

  5023.                 // TODO: may need modification to handle slice coding

  5024.                 ff_er_add_slice(&s->er, 0, s->start_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR);

  5025.                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i at %ix%i\n",

  5026.                        get_bits_count(&s->gb), v->bits, s->mb_x, s->mb_y);

  5027.                 return;

  5028.             }

  5029.         }

  5030.         memmove(v->cbp_base,      v->cbp,      sizeof(v->cbp_base[0])      * s->mb_stride);

  5031.         memmove(v->ttblk_base,    v->ttblk,    sizeof(v->ttblk_base[0])    * s->mb_stride);

  5032.         memmove(v->is_intra_base, v->is_intra, sizeof(v->is_intra_base[0]) * s->mb_stride);

  5033.         memmove(v->luma_mv_base,  v->luma_mv,  sizeof(v->luma_mv_base[0])  * s->mb_stride);

  5034.         if (s->mb_y != s->start_mb_y) ff_mpeg_draw_horiz_band(s, (s->mb_y - 1) * 16, 16);

  5035.         s->first_slice_line = 0;

  5036.     }

  5037.     if (apply_loop_filter) {

  5038.         s->mb_x = 0;

  5039.         init_block_index(v);

  5040.         for (; s->mb_x < s->mb_width; s->mb_x++) {

  5041.             ff_update_block_index(s);

  5042.             vc1_apply_p_loop_filter(v);

  5043.         }

  5044.     }

  5045.     if (s->end_mb_y >= s->start_mb_y)

  5046.         ff_mpeg_draw_horiz_band(s, (s->end_mb_y - 1) * 16, 16);

  5047.     ff_er_add_slice(&s->er, 0, s->start_mb_y << v->field_mode, s->mb_width - 1,

  5048.                     (s->end_mb_y << v->field_mode) - 1, ER_MB_END);

  5049. }

  5050. 

  5051. static void vc1_decode_b_blocks(VC1Context *v)

  5052. {

  5053.     MpegEncContext *s = &v->s;

  5054. 

  5055.     /* select codingmode used for VLC tables selection */

  5056.     switch (v->c_ac_table_index) {

  5057.     case 0:

  5058.         v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;

  5059.         break;

  5060.     case 1:

  5061.         v->codingset = CS_HIGH_MOT_INTRA;

  5062.         break;

  5063.     case 2:

  5064.         v->codingset = CS_MID_RATE_INTRA;

  5065.         break;

  5066.     }

  5067. 

  5068.     switch (v->c_ac_table_index) {

  5069.     case 0:

  5070.         v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;

  5071.         break;

  5072.     case 1:

  5073.         v->codingset2 = CS_HIGH_MOT_INTER;

  5074.         break;

  5075.     case 2:

  5076.         v->codingset2 = CS_MID_RATE_INTER;

  5077.         break;

  5078.     }

  5079. 

  5080.     s->first_slice_line = 1;

  5081.     for (s->mb_y = s->start_mb_y; s->mb_y < s->end_mb_y; s->mb_y++) {

  5082.         s->mb_x = 0;

  5083.         init_block_index(v);

  5084.         for (; s->mb_x < s->mb_width; s->mb_x++) {

  5085.             ff_update_block_index(s);

  5086. 

  5087.             if (v->fcm == ILACE_FIELD)

  5088.                 vc1_decode_b_mb_intfi(v);

  5089.             else if (v->fcm == ILACE_FRAME)

  5090.                 vc1_decode_b_mb_intfr(v);

  5091.             else

  5092.                 vc1_decode_b_mb(v);

  5093.             if (get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {

  5094.                 // TODO: may need modification to handle slice coding

  5095.                 ff_er_add_slice(&s->er, 0, s->start_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR);

  5096.                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i at %ix%i\n",

  5097.                        get_bits_count(&s->gb), v->bits, s->mb_x, s->mb_y);

  5098.                 return;

  5099.             }

  5100.             if (v->s.loop_filter) vc1_loop_filter_iblk(v, v->pq);

  5101.         }

  5102.         if (!v->s.loop_filter)

  5103.             ff_mpeg_draw_horiz_band(s, s->mb_y * 16, 16);

  5104.         else if (s->mb_y)

  5105.             ff_mpeg_draw_horiz_band(s, (s->mb_y - 1) * 16, 16);

  5106.         s->first_slice_line = 0;

  5107.     }

  5108.     if (v->s.loop_filter)

  5109.         ff_mpeg_draw_horiz_band(s, (s->end_mb_y - 1) * 16, 16);

  5110.     ff_er_add_slice(&s->er, 0, s->start_mb_y << v->field_mode, s->mb_width - 1,

  5111.                     (s->end_mb_y << v->field_mode) - 1, ER_MB_END);

  5112. }

  5113. 

  5114. static void vc1_decode_skip_blocks(VC1Context *v)

  5115. {

  5116.     MpegEncContext *s = &v->s;

  5117. 

  5118.     ff_er_add_slice(&s->er, 0, s->start_mb_y, s->mb_width - 1, s->end_mb_y - 1, ER_MB_END);

  5119.     s->first_slice_line = 1;

  5120.     for (s->mb_y = s->start_mb_y; s->mb_y < s->end_mb_y; s->mb_y++) {

  5121.         s->mb_x = 0;

  5122.         init_block_index(v);

  5123.         ff_update_block_index(s);

  5124.         memcpy(s->dest[0], s->last_picture.f.data[0] + s->mb_y * 16 * s->linesize,   s->linesize   * 16);

  5125.         memcpy(s->dest[1], s->last_picture.f.data[1] + s->mb_y *  8 * s->uvlinesize, s->uvlinesize *  8);

  5126.         memcpy(s->dest[2], s->last_picture.f.data[2] + s->mb_y *  8 * s->uvlinesize, s->uvlinesize *  8);

  5127.         ff_mpeg_draw_horiz_band(s, s->mb_y * 16, 16);

  5128.         s->first_slice_line = 0;

  5129.     }

  5130.     s->pict_type = AV_PICTURE_TYPE_P;

  5131. }

  5132. 

  5133. void ff_vc1_decode_blocks(VC1Context *v)

  5134. {

  5135. 

  5136.     v->s.esc3_level_length = 0;

  5137.     if (v->x8_type) {

  5138.         ff_intrax8_decode_picture(&v->x8, 2*v->pq + v->halfpq, v->pq * !v->pquantizer);

  5139.     } else {

  5140.         v->cur_blk_idx     =  0;

  5141.         v->left_blk_idx    = -1;

  5142.         v->topleft_blk_idx =  1;

  5143.         v->top_blk_idx     =  2;

  5144.         switch (v->s.pict_type) {

  5145.         case AV_PICTURE_TYPE_I:

  5146.             if (v->profile == PROFILE_ADVANCED)

  5147.                 vc1_decode_i_blocks_adv(v);

  5148.             else

  5149.                 vc1_decode_i_blocks(v);

  5150.             break;

  5151.         case AV_PICTURE_TYPE_P:

  5152.             if (v->p_frame_skipped)

  5153.                 vc1_decode_skip_blocks(v);

  5154.             else

  5155.                 vc1_decode_p_blocks(v);

  5156.             break;

  5157.         case AV_PICTURE_TYPE_B:

  5158.             if (v->bi_type) {

  5159.                 if (v->profile == PROFILE_ADVANCED)

  5160.                     vc1_decode_i_blocks_adv(v);

  5161.                 else

  5162.                     vc1_decode_i_blocks(v);

  5163.             } else

  5164.                 vc1_decode_b_blocks(v);

  5165.             break;

  5166.         }

  5167.     }

  5168. }

  5169. 

  5170. #if CONFIG_WMV3IMAGE_DECODER || CONFIG_VC1IMAGE_DECODER

  5171. 

  5172. typedef struct {

  5173.     /**

  5174.      * Transform coefficients for both sprites in 16.16 fixed point format,

  5175.      * in the order they appear in the bitstream:

  5176.      *  x scale

  5177.      *  rotation 1 (unused)

  5178.      *  x offset

  5179.      *  rotation 2 (unused)

  5180.      *  y scale

  5181.      *  y offset

  5182.      *  alpha

  5183.      */

  5184.     int coefs[2][7];

  5185. 

  5186.     int effect_type, effect_flag;

  5187.     int effect_pcount1, effect_pcount2;   ///< amount of effect parameters stored in effect_params

  5188.     int effect_params1[15], effect_params2[10]; ///< effect parameters in 16.16 fixed point format

  5189. } SpriteData;

  5190. 

  5191. static inline int get_fp_val(GetBitContext* gb)

  5192. {

  5193.     return (get_bits_long(gb, 30) - (1 << 29)) << 1;

  5194. }

  5195. 

  5196. static void vc1_sprite_parse_transform(GetBitContext* gb, int c[7])

  5197. {

  5198.     c[1] = c[3] = 0;

  5199. 

  5200.     switch (get_bits(gb, 2)) {

  5201.     case 0:

  5202.         c[0] = 1 << 16;

  5203.         c[2] = get_fp_val(gb);

  5204.         c[4] = 1 << 16;

  5205.         break;

  5206.     case 1:

  5207.         c[0] = c[4] = get_fp_val(gb);

  5208.         c[2] = get_fp_val(gb);

  5209.         break;

  5210.     case 2:

  5211.         c[0] = get_fp_val(gb);

  5212.         c[2] = get_fp_val(gb);

  5213.         c[4] = get_fp_val(gb);

  5214.         break;

  5215.     case 3:

  5216.         c[0] = get_fp_val(gb);

  5217.         c[1] = get_fp_val(gb);

  5218.         c[2] = get_fp_val(gb);

  5219.         c[3] = get_fp_val(gb);

  5220.         c[4] = get_fp_val(gb);

  5221.         break;

  5222.     }

  5223.     c[5] = get_fp_val(gb);

  5224.     if (get_bits1(gb))

  5225.         c[6] = get_fp_val(gb);

  5226.     else

  5227.         c[6] = 1 << 16;

  5228. }

  5229. 

  5230. static void vc1_parse_sprites(VC1Context *v, GetBitContext* gb, SpriteData* sd)

  5231. {

  5232.     AVCodecContext *avctx = v->s.avctx;

  5233.     int sprite, i;

  5234. 

  5235.     for (sprite = 0; sprite <= v->two_sprites; sprite++) {

  5236.         vc1_sprite_parse_transform(gb, sd->coefs[sprite]);

  5237.         if (sd->coefs[sprite][1] || sd->coefs[sprite][3])

  5238.             avpriv_request_sample(avctx, "Non-zero rotation coefficients");

  5239.         av_log(avctx, AV_LOG_DEBUG, sprite ? "S2:" : "S1:");

  5240.         for (i = 0; i < 7; i++)

  5241.             av_log(avctx, AV_LOG_DEBUG, " %d.%.3d",

  5242.                    sd->coefs[sprite][i] / (1<<16),

  5243.                    (abs(sd->coefs[sprite][i]) & 0xFFFF) * 1000 / (1 << 16));

  5244.         av_log(avctx, AV_LOG_DEBUG, "\n");

  5245.     }

  5246. 

  5247.     skip_bits(gb, 2);

  5248.     if (sd->effect_type = get_bits_long(gb, 30)) {

  5249.         switch (sd->effect_pcount1 = get_bits(gb, 4)) {

  5250.         case 7:

  5251.             vc1_sprite_parse_transform(gb, sd->effect_params1);

  5252.             break;

  5253.         case 14:

  5254.             vc1_sprite_parse_transform(gb, sd->effect_params1);

  5255.             vc1_sprite_parse_transform(gb, sd->effect_params1 + 7);

  5256.             break;

  5257.         default:

  5258.             for (i = 0; i < sd->effect_pcount1; i++)

  5259.                 sd->effect_params1[i] = get_fp_val(gb);

  5260.         }

  5261.         if (sd->effect_type != 13 || sd->effect_params1[0] != sd->coefs[0][6]) {

  5262.             // effect 13 is simple alpha blending and matches the opacity above

  5263.             av_log(avctx, AV_LOG_DEBUG, "Effect: %d; params: ", sd->effect_type);

  5264.             for (i = 0; i < sd->effect_pcount1; i++)

  5265.                 av_log(avctx, AV_LOG_DEBUG, " %d.%.2d",

  5266.                        sd->effect_params1[i] / (1 << 16),

  5267.                        (abs(sd->effect_params1[i]) & 0xFFFF) * 1000 / (1 << 16));

  5268.             av_log(avctx, AV_LOG_DEBUG, "\n");

  5269.         }

  5270. 

  5271.         sd->effect_pcount2 = get_bits(gb, 16);

  5272.         if (sd->effect_pcount2 > 10) {

  5273.             av_log(avctx, AV_LOG_ERROR, "Too many effect parameters\n");

  5274.             return;

  5275.         } else if (sd->effect_pcount2) {

  5276.             i = -1;

  5277.             av_log(avctx, AV_LOG_DEBUG, "Effect params 2: ");

  5278.             while (++i < sd->effect_pcount2) {

  5279.                 sd->effect_params2[i] = get_fp_val(gb);

  5280.                 av_log(avctx, AV_LOG_DEBUG, " %d.%.2d",

  5281.                        sd->effect_params2[i] / (1 << 16),

  5282.                        (abs(sd->effect_params2[i]) & 0xFFFF) * 1000 / (1 << 16));

  5283.             }

  5284.             av_log(avctx, AV_LOG_DEBUG, "\n");

  5285.         }

  5286.     }

  5287.     if (sd->effect_flag = get_bits1(gb))

  5288.         av_log(avctx, AV_LOG_DEBUG, "Effect flag set\n");

  5289. 

  5290.     if (get_bits_count(gb) >= gb->size_in_bits +

  5291.        (avctx->codec_id == AV_CODEC_ID_WMV3IMAGE ? 64 : 0))

  5292.         av_log(avctx, AV_LOG_ERROR, "Buffer overrun\n");

  5293.     if (get_bits_count(gb) < gb->size_in_bits - 8)

  5294.         av_log(avctx, AV_LOG_WARNING, "Buffer not fully read\n");

  5295. }

  5296. 

  5297. static void vc1_draw_sprites(VC1Context *v, SpriteData* sd)

  5298. {

  5299.     int i, plane, row, sprite;

  5300.     int sr_cache[2][2] = { { -1, -1 }, { -1, -1 } };

  5301.     uint8_t* src_h[2][2];

  5302.     int xoff[2], xadv[2], yoff[2], yadv[2], alpha;

  5303.     int ysub[2];

  5304.     MpegEncContext *s = &v->s;

  5305. 

  5306.     for (i = 0; i < 2; i++) {

  5307.         xoff[i] = av_clip(sd->coefs[i][2], 0, v->sprite_width-1 << 16);

  5308.         xadv[i] = sd->coefs[i][0];

  5309.         if (xadv[i] != 1<<16 || (v->sprite_width << 16) - (v->output_width << 16) - xoff[i])

  5310.             xadv[i] = av_clip(xadv[i], 0, ((v->sprite_width<<16) - xoff[i] - 1) / v->output_width);

  5311. 

  5312.         yoff[i] = av_clip(sd->coefs[i][5], 0, v->sprite_height-1 << 16);

  5313.         yadv[i] = av_clip(sd->coefs[i][4], 0, ((v->sprite_height << 16) - yoff[i]) / v->output_height);

  5314.     }

  5315.     alpha = av_clip(sd->coefs[1][6], 0, (1<<16) - 1);

  5316. 

  5317.     for (plane = 0; plane < (s->flags&CODEC_FLAG_GRAY ? 1 : 3); plane++) {

  5318.         int width = v->output_width>>!!plane;

  5319. 

  5320.         for (row = 0; row < v->output_height>>!!plane; row++) {

  5321.             uint8_t *dst = v->sprite_output_frame.data[plane] +

  5322.                            v->sprite_output_frame.linesize[plane] * row;

  5323. 

  5324.             for (sprite = 0; sprite <= v->two_sprites; sprite++) {

  5325.                 uint8_t *iplane = s->current_picture.f.data[plane];

  5326.                 int      iline  = s->current_picture.f.linesize[plane];

  5327.                 int      ycoord = yoff[sprite] + yadv[sprite] * row;

  5328.                 int      yline  = ycoord >> 16;

  5329.                 int      next_line;

  5330.                 ysub[sprite] = ycoord & 0xFFFF;

  5331.                 if (sprite) {

  5332.                     iplane = s->last_picture.f.data[plane];

  5333.                     iline  = s->last_picture.f.linesize[plane];

  5334.                 }

  5335.                 next_line = FFMIN(yline + 1, (v->sprite_height >> !!plane) - 1) * iline;

  5336.                 if (!(xoff[sprite] & 0xFFFF) && xadv[sprite] == 1 << 16) {

  5337.                         src_h[sprite][0] = iplane + (xoff[sprite] >> 16) +  yline      * iline;

  5338.                     if (ysub[sprite])

  5339.                         src_h[sprite][1] = iplane + (xoff[sprite] >> 16) + next_line;

  5340.                 } else {

  5341.                     if (sr_cache[sprite][0] != yline) {

  5342.                         if (sr_cache[sprite][1] == yline) {

  5343.                             FFSWAP(uint8_t*, v->sr_rows[sprite][0], v->sr_rows[sprite][1]);

  5344.                             FFSWAP(int,        sr_cache[sprite][0],   sr_cache[sprite][1]);

  5345.                         } else {

  5346.                             v->vc1dsp.sprite_h(v->sr_rows[sprite][0], iplane + yline * iline, xoff[sprite], xadv[sprite], width);

  5347.                             sr_cache[sprite][0] = yline;

  5348.                         }

  5349.                     }

  5350.                     if (ysub[sprite] && sr_cache[sprite][1] != yline + 1) {

  5351.                         v->vc1dsp.sprite_h(v->sr_rows[sprite][1],

  5352.                                            iplane + next_line, xoff[sprite],

  5353.                                            xadv[sprite], width);

  5354.                         sr_cache[sprite][1] = yline + 1;

  5355.                     }

  5356.                     src_h[sprite][0] = v->sr_rows[sprite][0];

  5357.                     src_h[sprite][1] = v->sr_rows[sprite][1];

  5358.                 }

  5359.             }

  5360. 

  5361.             if (!v->two_sprites) {

  5362.                 if (ysub[0]) {

  5363.                     v->vc1dsp.sprite_v_single(dst, src_h[0][0], src_h[0][1], ysub[0], width);

  5364.                 } else {

  5365.                     memcpy(dst, src_h[0][0], width);

  5366.                 }

  5367.             } else {

  5368.                 if (ysub[0] && ysub[1]) {

  5369.                     v->vc1dsp.sprite_v_double_twoscale(dst, src_h[0][0], src_h[0][1], ysub[0],

  5370.                                                        src_h[1][0], src_h[1][1], ysub[1], alpha, width);

  5371.                 } else if (ysub[0]) {

  5372.                     v->vc1dsp.sprite_v_double_onescale(dst, src_h[0][0], src_h[0][1], ysub[0],

  5373.                                                        src_h[1][0], alpha, width);

  5374.                 } else if (ysub[1]) {

  5375.                     v->vc1dsp.sprite_v_double_onescale(dst, src_h[1][0], src_h[1][1], ysub[1],

  5376.                                                        src_h[0][0], (1<<16)-1-alpha, width);

  5377.                 } else {

  5378.                     v->vc1dsp.sprite_v_double_noscale(dst, src_h[0][0], src_h[1][0], alpha, width);

  5379.                 }

  5380.             }

  5381.         }

  5382. 

  5383.         if (!plane) {

  5384.             for (i = 0; i < 2; i++) {

  5385.                 xoff[i] >>= 1;

  5386.                 yoff[i] >>= 1;

  5387.             }

  5388.         }

  5389. 

  5390.     }

  5391. }

  5392. 

  5393. 

  5394. static int vc1_decode_sprites(VC1Context *v, GetBitContext* gb)

  5395. {

  5396.     MpegEncContext *s     = &v->s;

  5397.     AVCodecContext *avctx = s->avctx;

  5398.     SpriteData sd;

  5399. 

  5400.     vc1_parse_sprites(v, gb, &sd);

  5401. 

  5402.     if (!s->current_picture.f.data[0]) {

  5403.         av_log(avctx, AV_LOG_ERROR, "Got no sprites\n");

  5404.         return -1;

  5405.     }

  5406. 

  5407.     if (v->two_sprites && (!s->last_picture_ptr || !s->last_picture.f.data[0])) {

  5408.         av_log(avctx, AV_LOG_WARNING, "Need two sprites, only got one\n");

  5409.         v->two_sprites = 0;

  5410.     }

  5411. 

  5412.     av_frame_unref(&v->sprite_output_frame);

  5413.     if (ff_get_buffer(avctx, &v->sprite_output_frame, 0) < 0) {

  5414.         av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");

  5415.         return -1;

  5416.     }

  5417. 

  5418.     vc1_draw_sprites(v, &sd);

  5419. 

  5420.     return 0;

  5421. }

  5422. 

  5423. static void vc1_sprite_flush(AVCodecContext *avctx)

  5424. {

  5425.     VC1Context *v     = avctx->priv_data;

  5426.     MpegEncContext *s = &v->s;

  5427.     AVFrame *f = &s->current_picture.f;

  5428.     int plane, i;

  5429. 

  5430.     /* Windows Media Image codecs have a convergence interval of two keyframes.

  5431.        Since we can't enforce it, clear to black the missing sprite. This is

  5432.        wrong but it looks better than doing nothing. */

  5433. 

  5434.     if (f->data[0])

  5435.         for (plane = 0; plane < (s->flags&CODEC_FLAG_GRAY ? 1 : 3); plane++)

  5436.             for (i = 0; i < v->sprite_height>>!!plane; i++)

  5437.                 memset(f->data[plane] + i * f->linesize[plane],

  5438.                        plane ? 128 : 0, f->linesize[plane]);

  5439. }

  5440. 

  5441. #endif

  5442. 

  5443. av_cold int ff_vc1_decode_init_alloc_tables(VC1Context *v)

  5444. {

  5445.     MpegEncContext *s = &v->s;

  5446.     int i;

  5447. 

  5448.     /* Allocate mb bitplanes */

  5449.     v->mv_type_mb_plane = av_malloc (s->mb_stride * s->mb_height);

  5450.     v->direct_mb_plane  = av_malloc (s->mb_stride * s->mb_height);

  5451.     v->forward_mb_plane = av_malloc (s->mb_stride * s->mb_height);

  5452.     v->fieldtx_plane    = av_mallocz(s->mb_stride * s->mb_height);

  5453.     v->acpred_plane     = av_malloc (s->mb_stride * s->mb_height);

  5454.     v->over_flags_plane = av_malloc (s->mb_stride * s->mb_height);

  5455. 

  5456.     v->n_allocated_blks = s->mb_width + 2;

  5457.     v->block            = av_malloc(sizeof(*v->block) * v->n_allocated_blks);

  5458.     v->cbp_base         = av_malloc(sizeof(v->cbp_base[0]) * 2 * s->mb_stride);

  5459.     v->cbp              = v->cbp_base + s->mb_stride;

  5460.     v->ttblk_base       = av_malloc(sizeof(v->ttblk_base[0]) * 2 * s->mb_stride);

  5461.     v->ttblk            = v->ttblk_base + s->mb_stride;

  5462.     v->is_intra_base    = av_mallocz(sizeof(v->is_intra_base[0]) * 2 * s->mb_stride);

  5463.     v->is_intra         = v->is_intra_base + s->mb_stride;

  5464.     v->luma_mv_base     = av_malloc(sizeof(v->luma_mv_base[0]) * 2 * s->mb_stride);

  5465.     v->luma_mv          = v->luma_mv_base + s->mb_stride;

  5466. 

  5467.     /* allocate block type info in that way so it could be used with s->block_index[] */

  5468.     v->mb_type_base = av_malloc(s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);

  5469.     v->mb_type[0]   = v->mb_type_base + s->b8_stride + 1;

  5470.     v->mb_type[1]   = v->mb_type_base + s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride + 1;

  5471.     v->mb_type[2]   = v->mb_type[1] + s->mb_stride * (s->mb_height + 1);

  5472. 

  5473.     /* allocate memory to store block level MV info */

  5474.     v->blk_mv_type_base = av_mallocz(     s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);

  5475.     v->blk_mv_type      = v->blk_mv_type_base + s->b8_stride + 1;

  5476.     v->mv_f_base        = av_mallocz(2 * (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2));

  5477.     v->mv_f[0]          = v->mv_f_base + s->b8_stride + 1;

  5478.     v->mv_f[1]          = v->mv_f[0] + (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);

  5479.     v->mv_f_next_base   = av_mallocz(2 * (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2));

  5480.     v->mv_f_next[0]     = v->mv_f_next_base + s->b8_stride + 1;

  5481.     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);

  5482. 

  5483.     /* Init coded blocks info */

  5484.     if (v->profile == PROFILE_ADVANCED) {

  5485. //        if (alloc_bitplane(&v->over_flags_plane, s->mb_width, s->mb_height) < 0)

  5486. //            return -1;

  5487. //        if (alloc_bitplane(&v->ac_pred_plane, s->mb_width, s->mb_height) < 0)

  5488. //            return -1;

  5489.     }

  5490. 

  5491.     ff_intrax8_common_init(&v->x8,s);

  5492. 

  5493.     if (s->avctx->codec_id == AV_CODEC_ID_WMV3IMAGE || s->avctx->codec_id == AV_CODEC_ID_VC1IMAGE) {

  5494.         for (i = 0; i < 4; i++)

  5495.             if (!(v->sr_rows[i >> 1][i & 1] = av_malloc(v->output_width))) return -1;

  5496.     }

  5497. 

  5498.     if (!v->mv_type_mb_plane || !v->direct_mb_plane || !v->acpred_plane || !v->over_flags_plane ||

  5499.         !v->block || !v->cbp_base || !v->ttblk_base || !v->is_intra_base || !v->luma_mv_base ||

  5500.         !v->mb_type_base)

  5501.             return -1;

  5502. 

  5503.     return 0;

  5504. }

  5505. 

  5506. av_cold void ff_vc1_init_transposed_scantables(VC1Context *v)

  5507. {

  5508.     int i;

  5509.     for (i = 0; i < 64; i++) {

  5510. #define transpose(x) ((x >> 3) | ((x & 7) << 3))

  5511.         v->zz_8x8[0][i] = transpose(ff_wmv1_scantable[0][i]);

  5512.         v->zz_8x8[1][i] = transpose(ff_wmv1_scantable[1][i]);

  5513.         v->zz_8x8[2][i] = transpose(ff_wmv1_scantable[2][i]);

  5514.         v->zz_8x8[3][i] = transpose(ff_wmv1_scantable[3][i]);

  5515.         v->zzi_8x8[i]   = transpose(ff_vc1_adv_interlaced_8x8_zz[i]);

  5516.     }

  5517.     v->left_blk_sh = 0;

  5518.     v->top_blk_sh  = 3;

  5519. }

  5520. 

  5521. /** Initialize a VC1/WMV3 decoder

  5522.  * @todo TODO: Handle VC-1 IDUs (Transport level?)

  5523.  * @todo TODO: Decypher remaining bits in extra_data

  5524.  */

  5525. static av_cold int vc1_decode_init(AVCodecContext *avctx)

  5526. {

  5527.     VC1Context *v = avctx->priv_data;

  5528.     MpegEncContext *s = &v->s;

  5529.     GetBitContext gb;

  5530. 

  5531.     /* save the container output size for WMImage */

  5532.     v->output_width  = avctx->width;

  5533.     v->output_height = avctx->height;

  5534. 

  5535.     if (!avctx->extradata_size || !avctx->extradata)

  5536.         return -1;

  5537.     if (!(avctx->flags & CODEC_FLAG_GRAY))

  5538.         avctx->pix_fmt = avctx->get_format(avctx, avctx->codec->pix_fmts);

  5539.     else

  5540.         avctx->pix_fmt = AV_PIX_FMT_GRAY8;

  5541.     avctx->hwaccel = ff_find_hwaccel(avctx->codec->id, avctx->pix_fmt);

  5542.     v->s.avctx = avctx;

  5543.     avctx->flags |= CODEC_FLAG_EMU_EDGE;

  5544.     v->s.flags   |= CODEC_FLAG_EMU_EDGE;

  5545. 

  5546.     if (ff_vc1_init_common(v) < 0)

  5547.         return -1;

  5548.     ff_h264chroma_init(&v->h264chroma, 8);

  5549.     ff_vc1dsp_init(&v->vc1dsp);

  5550. 

  5551.     if (avctx->codec_id == AV_CODEC_ID_WMV3 || avctx->codec_id == AV_CODEC_ID_WMV3IMAGE) {

  5552.         int count = 0;

  5553. 

  5554.         // looks like WMV3 has a sequence header stored in the extradata

  5555.         // advanced sequence header may be before the first frame

  5556.         // the last byte of the extradata is a version number, 1 for the

  5557.         // samples we can decode

  5558. 

  5559.         init_get_bits(&gb, avctx->extradata, avctx->extradata_size*8);

  5560. 

  5561.         if (ff_vc1_decode_sequence_header(avctx, v, &gb) < 0)

  5562.           return -1;

  5563. 

  5564.         count = avctx->extradata_size*8 - get_bits_count(&gb);

  5565.         if (count > 0) {

  5566.             av_log(avctx, AV_LOG_INFO, "Extra data: %i bits left, value: %X\n",

  5567.                    count, get_bits(&gb, count));

  5568.         } else if (count < 0) {

  5569.             av_log(avctx, AV_LOG_INFO, "Read %i bits in overflow\n", -count);

  5570.         }

  5571.     } else { // VC1/WVC1/WVP2

  5572.         const uint8_t *start = avctx->extradata;

  5573.         uint8_t *end = avctx->extradata + avctx->extradata_size;

  5574.         const uint8_t *next;

  5575.         int size, buf2_size;

  5576.         uint8_t *buf2 = NULL;

  5577.         int seq_initialized = 0, ep_initialized = 0;

  5578. 

  5579.         if (avctx->extradata_size < 16) {

  5580.             av_log(avctx, AV_LOG_ERROR, "Extradata size too small: %i\n", avctx->extradata_size);

  5581.             return -1;

  5582.         }

  5583. 

  5584.         buf2  = av_mallocz(avctx->extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);

  5585.         start = find_next_marker(start, end); // in WVC1 extradata first byte is its size, but can be 0 in mkv

  5586.         next  = start;

  5587.         for (; next < end; start = next) {

  5588.             next = find_next_marker(start + 4, end);

  5589.             size = next - start - 4;

  5590.             if (size <= 0)

  5591.                 continue;

  5592.             buf2_size = vc1_unescape_buffer(start + 4, size, buf2);

  5593.             init_get_bits(&gb, buf2, buf2_size * 8);

  5594.             switch (AV_RB32(start)) {

  5595.             case VC1_CODE_SEQHDR:

  5596.                 if (ff_vc1_decode_sequence_header(avctx, v, &gb) < 0) {

  5597.                     av_free(buf2);

  5598.                     return -1;

  5599.                 }

  5600.                 seq_initialized = 1;

  5601.                 break;

  5602.             case VC1_CODE_ENTRYPOINT:

  5603.                 if (ff_vc1_decode_entry_point(avctx, v, &gb) < 0) {

  5604.                     av_free(buf2);

  5605.                     return -1;

  5606.                 }

  5607.                 ep_initialized = 1;

  5608.                 break;

  5609.             }

  5610.         }

  5611.         av_free(buf2);

  5612.         if (!seq_initialized || !ep_initialized) {

  5613.             av_log(avctx, AV_LOG_ERROR, "Incomplete extradata\n");

  5614.             return -1;

  5615.         }

  5616.         v->res_sprite = (avctx->codec_id == AV_CODEC_ID_VC1IMAGE);

  5617.     }

  5618. 

  5619.     avctx->profile = v->profile;

  5620.     if (v->profile == PROFILE_ADVANCED)

  5621.         avctx->level = v->level;

  5622. 

  5623.     avctx->has_b_frames = !!avctx->max_b_frames;

  5624. 

  5625.     s->mb_width  = (avctx->coded_width  + 15) >> 4;

  5626.     s->mb_height = (avctx->coded_height + 15) >> 4;

  5627. 

  5628.     if (v->profile == PROFILE_ADVANCED || v->res_fasttx) {

  5629.         ff_vc1_init_transposed_scantables(v);

  5630.     } else {

  5631.         memcpy(v->zz_8x8, ff_wmv1_scantable, 4*64);

  5632.         v->left_blk_sh = 3;

  5633.         v->top_blk_sh  = 0;

  5634.     }

  5635. 

  5636.     if (avctx->codec_id == AV_CODEC_ID_WMV3IMAGE || avctx->codec_id == AV_CODEC_ID_VC1IMAGE) {

  5637.         v->sprite_width  = avctx->coded_width;

  5638.         v->sprite_height = avctx->coded_height;

  5639. 

  5640.         avctx->coded_width  = avctx->width  = v->output_width;

  5641.         avctx->coded_height = avctx->height = v->output_height;

  5642. 

  5643.         // prevent 16.16 overflows

  5644.         if (v->sprite_width  > 1 << 14 ||

  5645.             v->sprite_height > 1 << 14 ||

  5646.             v->output_width  > 1 << 14 ||

  5647.             v->output_height > 1 << 14) return -1;

  5648.     }

  5649.     return 0;

  5650. }

  5651. 

  5652. /** Close a VC1/WMV3 decoder

  5653.  * @warning Initial try at using MpegEncContext stuff

  5654.  */

  5655. av_cold int ff_vc1_decode_end(AVCodecContext *avctx)

  5656. {

  5657.     VC1Context *v = avctx->priv_data;

  5658.     int i;

  5659. 

  5660.     av_frame_unref(&v->sprite_output_frame);

  5661. 

  5662.     for (i = 0; i < 4; i++)

  5663.         av_freep(&v->sr_rows[i >> 1][i & 1]);

  5664.     av_freep(&v->hrd_rate);

  5665.     av_freep(&v->hrd_buffer);

  5666.     ff_MPV_common_end(&v->s);

  5667.     av_freep(&v->mv_type_mb_plane);

  5668.     av_freep(&v->direct_mb_plane);

  5669.     av_freep(&v->forward_mb_plane);

  5670.     av_freep(&v->fieldtx_plane);

  5671.     av_freep(&v->acpred_plane);

  5672.     av_freep(&v->over_flags_plane);

  5673.     av_freep(&v->mb_type_base);

  5674.     av_freep(&v->blk_mv_type_base);

  5675.     av_freep(&v->mv_f_base);

  5676.     av_freep(&v->mv_f_next_base);

  5677.     av_freep(&v->block);

  5678.     av_freep(&v->cbp_base);

  5679.     av_freep(&v->ttblk_base);

  5680.     av_freep(&v->is_intra_base); // FIXME use v->mb_type[]

  5681.     av_freep(&v->luma_mv_base);

  5682.     ff_intrax8_common_end(&v->x8);

  5683.     return 0;

  5684. }

  5685. 

  5686. 

  5687. /** Decode a VC1/WMV3 frame

  5688.  * @todo TODO: Handle VC-1 IDUs (Transport level?)

  5689.  */

  5690. static int vc1_decode_frame(AVCodecContext *avctx, void *data,

  5691.                             int *got_frame, AVPacket *avpkt)

  5692. {

  5693.     const uint8_t *buf = avpkt->data;

  5694.     int buf_size = avpkt->size, n_slices = 0, i, ret;

  5695.     VC1Context *v = avctx->priv_data;

  5696.     MpegEncContext *s = &v->s;

  5697.     AVFrame *pict = data;

  5698.     uint8_t *buf2 = NULL;

  5699.     const uint8_t *buf_start = buf;

  5700.     int mb_height, n_slices1;

  5701.     struct {

  5702.         uint8_t *buf;

  5703.         GetBitContext gb;

  5704.         int mby_start;

  5705.     } *slices = NULL, *tmp;

  5706. 

  5707.     /* no supplementary picture */

  5708.     if (buf_size == 0 || (buf_size == 4 && AV_RB32(buf) == VC1_CODE_ENDOFSEQ)) {

  5709.         /* special case for last picture */

  5710.         if (s->low_delay == 0 && s->next_picture_ptr) {

  5711.             if ((ret = av_frame_ref(pict, &s->next_picture_ptr->f)) < 0)

  5712.                 return ret;

  5713.             s->next_picture_ptr = NULL;

  5714. 

  5715.             *got_frame = 1;

  5716.         }

  5717. 

  5718.         return 0;

  5719.     }

  5720. 

  5721.     if (s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU) {

  5722.         if (v->profile < PROFILE_ADVANCED)

  5723.             avctx->pix_fmt = AV_PIX_FMT_VDPAU_WMV3;

  5724.         else

  5725.             avctx->pix_fmt = AV_PIX_FMT_VDPAU_VC1;

  5726.     }

  5727. 

  5728.     //for advanced profile we may need to parse and unescape data

  5729.     if (avctx->codec_id == AV_CODEC_ID_VC1 || avctx->codec_id == AV_CODEC_ID_VC1IMAGE) {

  5730.         int buf_size2 = 0;

  5731.         buf2 = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);

  5732. 

  5733.         if (IS_MARKER(AV_RB32(buf))) { /* frame starts with marker and needs to be parsed */

  5734.             const uint8_t *start, *end, *next;

  5735.             int size;

  5736. 

  5737.             next = buf;

  5738.             for (start = buf, end = buf + buf_size; next < end; start = next) {

  5739.                 next = find_next_marker(start + 4, end);

  5740.                 size = next - start - 4;

  5741.                 if (size <= 0) continue;

  5742.                 switch (AV_RB32(start)) {

  5743.                 case VC1_CODE_FRAME:

  5744.                     if (avctx->hwaccel ||

  5745.                         s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU)

  5746.                         buf_start = start;

  5747.                     buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);

  5748.                     break;

  5749.                 case VC1_CODE_FIELD: {

  5750.                     int buf_size3;

  5751.                     tmp = av_realloc(slices, sizeof(*slices) * (n_slices+1));

  5752.                     if (!tmp)

  5753.                         goto err;

  5754.                     slices = tmp;

  5755.                     slices[n_slices].buf = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);

  5756.                     if (!slices[n_slices].buf)

  5757.                         goto err;

  5758.                     buf_size3 = vc1_unescape_buffer(start + 4, size,

  5759.                                                     slices[n_slices].buf);

  5760.                     init_get_bits(&slices[n_slices].gb, slices[n_slices].buf,

  5761.                                   buf_size3 << 3);

  5762.                     /* assuming that the field marker is at the exact middle,

  5763.                        hope it's correct */

  5764.                     slices[n_slices].mby_start = s->mb_height >> 1;

  5765.                     n_slices1 = n_slices - 1; // index of the last slice of the first field

  5766.                     n_slices++;

  5767.                     break;

  5768.                 }

  5769.                 case VC1_CODE_ENTRYPOINT: /* it should be before frame data */

  5770.                     buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);

  5771.                     init_get_bits(&s->gb, buf2, buf_size2 * 8);

  5772.                     ff_vc1_decode_entry_point(avctx, v, &s->gb);

  5773.                     break;

  5774.                 case VC1_CODE_SLICE: {

  5775.                     int buf_size3;

  5776.                     tmp = av_realloc(slices, sizeof(*slices) * (n_slices+1));

  5777.                     if (!tmp)

  5778.                         goto err;

  5779.                     slices = tmp;

  5780.                     slices[n_slices].buf = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);

  5781.                     if (!slices[n_slices].buf)

  5782.                         goto err;

  5783.                     buf_size3 = vc1_unescape_buffer(start + 4, size,

  5784.                                                     slices[n_slices].buf);

  5785.                     init_get_bits(&slices[n_slices].gb, slices[n_slices].buf,

  5786.                                   buf_size3 << 3);

  5787.                     slices[n_slices].mby_start = get_bits(&slices[n_slices].gb, 9);

  5788.                     n_slices++;

  5789.                     break;

  5790.                 }

  5791.                 }

  5792.             }

  5793.         } else if (v->interlace && ((buf[0] & 0xC0) == 0xC0)) { /* WVC1 interlaced stores both fields divided by marker */

  5794.             const uint8_t *divider;

  5795.             int buf_size3;

  5796. 

  5797.             divider = find_next_marker(buf, buf + buf_size);

  5798.             if ((divider == (buf + buf_size)) || AV_RB32(divider) != VC1_CODE_FIELD) {

  5799.                 av_log(avctx, AV_LOG_ERROR, "Error in WVC1 interlaced frame\n");

  5800.                 goto err;

  5801.             } else { // found field marker, unescape second field

  5802.                 tmp = av_realloc(slices, sizeof(*slices) * (n_slices+1));

  5803.                 if (!tmp)

  5804.                     goto err;

  5805.                 slices = tmp;

  5806.                 slices[n_slices].buf = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);

  5807.                 if (!slices[n_slices].buf)

  5808.                     goto err;

  5809.                 buf_size3 = vc1_unescape_buffer(divider + 4, buf + buf_size - divider - 4, slices[n_slices].buf);

  5810.                 init_get_bits(&slices[n_slices].gb, slices[n_slices].buf,

  5811.                               buf_size3 << 3);

  5812.                 slices[n_slices].mby_start = s->mb_height >> 1;

  5813.                 n_slices1 = n_slices - 1;

  5814.                 n_slices++;

  5815.             }

  5816.             buf_size2 = vc1_unescape_buffer(buf, divider - buf, buf2);

  5817.         } else {

  5818.             buf_size2 = vc1_unescape_buffer(buf, buf_size, buf2);

  5819.         }

  5820.         init_get_bits(&s->gb, buf2, buf_size2*8);

  5821.     } else

  5822.         init_get_bits(&s->gb, buf, buf_size*8);

  5823. 

  5824.     if (v->res_sprite) {

  5825.         v->new_sprite  = !get_bits1(&s->gb);

  5826.         v->two_sprites =  get_bits1(&s->gb);

  5827.         /* res_sprite means a Windows Media Image stream, AV_CODEC_ID_*IMAGE means

  5828.            we're using the sprite compositor. These are intentionally kept separate

  5829.            so you can get the raw sprites by using the wmv3 decoder for WMVP or

  5830.            the vc1 one for WVP2 */

  5831.         if (avctx->codec_id == AV_CODEC_ID_WMV3IMAGE || avctx->codec_id == AV_CODEC_ID_VC1IMAGE) {

  5832.             if (v->new_sprite) {

  5833.                 // switch AVCodecContext parameters to those of the sprites

  5834.                 avctx->width  = avctx->coded_width  = v->sprite_width;

  5835.                 avctx->height = avctx->coded_height = v->sprite_height;

  5836.             } else {

  5837.                 goto image;

  5838.             }

  5839.         }

  5840.     }

  5841. 

  5842.     if (s->context_initialized &&

  5843.         (s->width  != avctx->coded_width ||

  5844.          s->height != avctx->coded_height)) {

  5845.         ff_vc1_decode_end(avctx);

  5846.     }

  5847. 

  5848.     if (!s->context_initialized) {

  5849.         if (ff_msmpeg4_decode_init(avctx) < 0 || ff_vc1_decode_init_alloc_tables(v) < 0)

  5850.             goto err;

  5851. 

  5852.         s->low_delay = !avctx->has_b_frames || v->res_sprite;

  5853. 

  5854.         if (v->profile == PROFILE_ADVANCED) {

  5855.             s->h_edge_pos = avctx->coded_width;

  5856.             s->v_edge_pos = avctx->coded_height;

  5857.         }

  5858.     }

  5859. 

  5860.     /* We need to set current_picture_ptr before reading the header,

  5861.      * otherwise we cannot store anything in there. */

  5862.     if (s->current_picture_ptr == NULL || s->current_picture_ptr->f.data[0]) {

  5863.         int i = ff_find_unused_picture(s, 0);

  5864.         if (i < 0)

  5865.             goto err;

  5866.         s->current_picture_ptr = &s->picture[i];

  5867.     }

  5868. 

  5869.     // do parse frame header

  5870.     v->pic_header_flag = 0;

  5871.     v->first_pic_header_flag = 1;

  5872.     if (v->profile < PROFILE_ADVANCED) {

  5873.         if (ff_vc1_parse_frame_header(v, &s->gb) < 0) {

  5874.             goto err;

  5875.         }

  5876.     } else {

  5877.         if (ff_vc1_parse_frame_header_adv(v, &s->gb) < 0) {

  5878.             goto err;

  5879.         }

  5880.     }

  5881.     v->first_pic_header_flag = 0;

  5882. 

  5883.     if ((avctx->codec_id == AV_CODEC_ID_WMV3IMAGE || avctx->codec_id == AV_CODEC_ID_VC1IMAGE)

  5884.         && s->pict_type != AV_PICTURE_TYPE_I) {

  5885.         av_log(v->s.avctx, AV_LOG_ERROR, "Sprite decoder: expected I-frame\n");

  5886.         goto err;

  5887.     }

  5888. 

  5889.     // process pulldown flags

  5890.     s->current_picture_ptr->f.repeat_pict = 0;

  5891.     // Pulldown flags are only valid when 'broadcast' has been set.

  5892.     // So ticks_per_frame will be 2

  5893.     if (v->rff) {

  5894.         // repeat field

  5895.         s->current_picture_ptr->f.repeat_pict = 1;

  5896.     } else if (v->rptfrm) {

  5897.         // repeat frames

  5898.         s->current_picture_ptr->f.repeat_pict = v->rptfrm * 2;

  5899.     }

  5900. 

  5901.     // for skipping the frame

  5902.     s->current_picture.f.pict_type = s->pict_type;

  5903.     s->current_picture.f.key_frame = s->pict_type == AV_PICTURE_TYPE_I;

  5904. 

  5905.     /* skip B-frames if we don't have reference frames */

  5906.     if (s->last_picture_ptr == NULL && (s->pict_type == AV_PICTURE_TYPE_B || s->droppable)) {

  5907.         goto err;

  5908.     }

  5909.     if ((avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type == AV_PICTURE_TYPE_B) ||

  5910.         (avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type != AV_PICTURE_TYPE_I) ||

  5911.          avctx->skip_frame >= AVDISCARD_ALL) {

  5912.         goto end;

  5913.     }

  5914. 

  5915.     if (s->next_p_frame_damaged) {

  5916.         if (s->pict_type == AV_PICTURE_TYPE_B)

  5917.             goto end;

  5918.         else

  5919.             s->next_p_frame_damaged = 0;

  5920.     }

  5921. 

  5922.     if (ff_MPV_frame_start(s, avctx) < 0) {

  5923.         goto err;

  5924.     }

  5925. 

  5926.     s->me.qpel_put = s->dsp.put_qpel_pixels_tab;

  5927.     s->me.qpel_avg = s->dsp.avg_qpel_pixels_tab;

  5928. 

  5929.     if ((CONFIG_VC1_VDPAU_DECODER)

  5930.         &&s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU)

  5931.         ff_vdpau_vc1_decode_picture(s, buf_start, (buf + buf_size) - buf_start);

  5932.     else if (avctx->hwaccel) {

  5933.         if (avctx->hwaccel->start_frame(avctx, buf, buf_size) < 0)

  5934.             goto err;

  5935.         if (avctx->hwaccel->decode_slice(avctx, buf_start, (buf + buf_size) - buf_start) < 0)

  5936.             goto err;

  5937.         if (avctx->hwaccel->end_frame(avctx) < 0)

  5938.             goto err;

  5939.     } else {

  5940.         ff_mpeg_er_frame_start(s);

  5941. 

  5942.         v->bits = buf_size * 8;

  5943.         v->end_mb_x = s->mb_width;

  5944.         if (v->field_mode) {

  5945.             s->current_picture.f.linesize[0] <<= 1;

  5946.             s->current_picture.f.linesize[1] <<= 1;

  5947.             s->current_picture.f.linesize[2] <<= 1;

  5948.             s->linesize                      <<= 1;

  5949.             s->uvlinesize                    <<= 1;

  5950.         }

  5951.         mb_height = s->mb_height >> v->field_mode;

  5952.         for (i = 0; i <= n_slices; i++) {

  5953.             if (i > 0 &&  slices[i - 1].mby_start >= mb_height) {

  5954.                 if (v->field_mode <= 0) {

  5955.                     av_log(v->s.avctx, AV_LOG_ERROR, "Slice %d starts beyond "

  5956.                            "picture boundary (%d >= %d)\n", i,

  5957.                            slices[i - 1].mby_start, mb_height);

  5958.                     continue;

  5959.                 }

  5960.                 v->second_field = 1;

  5961.                 v->blocks_off   = s->mb_width  * s->mb_height << 1;

  5962.                 v->mb_off       = s->mb_stride * s->mb_height >> 1;

  5963.             } else {

  5964.                 v->second_field = 0;

  5965.                 v->blocks_off   = 0;

  5966.                 v->mb_off       = 0;

  5967.             }

  5968.             if (i) {

  5969.                 v->pic_header_flag = 0;

  5970.                 if (v->field_mode && i == n_slices1 + 2) {

  5971.                     if (ff_vc1_parse_frame_header_adv(v, &s->gb) < 0) {

  5972.                         av_log(v->s.avctx, AV_LOG_ERROR, "Field header damaged\n");

  5973.                         continue;

  5974.                     }

  5975.                 } else if (get_bits1(&s->gb)) {

  5976.                     v->pic_header_flag = 1;

  5977.                     if (ff_vc1_parse_frame_header_adv(v, &s->gb) < 0) {

  5978.                         av_log(v->s.avctx, AV_LOG_ERROR, "Slice header damaged\n");

  5979.                         continue;

  5980.                     }

  5981.                 }

  5982.             }

  5983.             s->start_mb_y = (i == 0) ? 0 : FFMAX(0, slices[i-1].mby_start % mb_height);

  5984.             if (!v->field_mode || v->second_field)

  5985.                 s->end_mb_y = (i == n_slices     ) ? mb_height : FFMIN(mb_height, slices[i].mby_start % mb_height);

  5986.             else

  5987.                 s->end_mb_y = (i <= n_slices1 + 1) ? mb_height : FFMIN(mb_height, slices[i].mby_start % mb_height);

  5988.             ff_vc1_decode_blocks(v);

  5989.             if (i != n_slices)

  5990.                 s->gb = slices[i].gb;

  5991.         }

  5992.         if (v->field_mode) {

  5993.             v->second_field = 0;

  5994.             s->current_picture.f.linesize[0] >>= 1;

  5995.             s->current_picture.f.linesize[1] >>= 1;

  5996.             s->current_picture.f.linesize[2] >>= 1;

  5997.             s->linesize                      >>= 1;

  5998.             s->uvlinesize                    >>= 1;

  5999.             if (v->s.pict_type != AV_PICTURE_TYPE_BI && v->s.pict_type != AV_PICTURE_TYPE_B) {

  6000.                 FFSWAP(uint8_t *, v->mv_f_next[0], v->mv_f[0]);

  6001.                 FFSWAP(uint8_t *, v->mv_f_next[1], v->mv_f[1]);

  6002.             }

  6003.         }

  6004.         av_dlog(s->avctx, "Consumed %i/%i bits\n",

  6005.                 get_bits_count(&s->gb), s->gb.size_in_bits);

  6006. //  if (get_bits_count(&s->gb) > buf_size * 8)

  6007. //      return -1;

  6008.         if (!v->field_mode)

  6009.             ff_er_frame_end(&s->er);

  6010.     }

  6011. 

  6012.     ff_MPV_frame_end(s);

  6013. 

  6014.     if (avctx->codec_id == AV_CODEC_ID_WMV3IMAGE || avctx->codec_id == AV_CODEC_ID_VC1IMAGE) {

  6015. image:

  6016.         avctx->width  = avctx->coded_width  = v->output_width;

  6017.         avctx->height = avctx->coded_height = v->output_height;

  6018.         if (avctx->skip_frame >= AVDISCARD_NONREF)

  6019.             goto end;

  6020. #if CONFIG_WMV3IMAGE_DECODER || CONFIG_VC1IMAGE_DECODER

  6021.         if (vc1_decode_sprites(v, &s->gb))

  6022.             goto err;

  6023. #endif

  6024.         if ((ret = av_frame_ref(pict, &v->sprite_output_frame)) < 0)

  6025.             goto err;

  6026.         *got_frame = 1;

  6027.     } else {

  6028.         if (s->pict_type == AV_PICTURE_TYPE_B || s->low_delay) {

  6029.             if ((ret = av_frame_ref(pict, &s->current_picture_ptr->f)) < 0)

  6030.                 goto err;

  6031.             ff_print_debug_info(s, s->current_picture_ptr);

  6032.         } else if (s->last_picture_ptr != NULL) {

  6033.             if ((ret = av_frame_ref(pict, &s->last_picture_ptr->f)) < 0)

  6034.                 goto err;

  6035.             ff_print_debug_info(s, s->last_picture_ptr);

  6036.         }

  6037.         if (s->last_picture_ptr || s->low_delay) {

  6038.             *got_frame = 1;

  6039.         }

  6040.     }

  6041. 

  6042. end:

  6043.     av_free(buf2);

  6044.     for (i = 0; i < n_slices; i++)

  6045.         av_free(slices[i].buf);

  6046.     av_free(slices);

  6047.     return buf_size;

  6048. 

  6049. err:

  6050.     av_free(buf2);

  6051.     for (i = 0; i < n_slices; i++)

  6052.         av_free(slices[i].buf);

  6053.     av_free(slices);

  6054.     return -1;

  6055. }

  6056. 

  6057. 

  6058. static const AVProfile profiles[] = {

  6059.     { FF_PROFILE_VC1_SIMPLE,   "Simple"   },

  6060.     { FF_PROFILE_VC1_MAIN,     "Main"     },

  6061.     { FF_PROFILE_VC1_COMPLEX,  "Complex"  },

  6062.     { FF_PROFILE_VC1_ADVANCED, "Advanced" },

  6063.     { FF_PROFILE_UNKNOWN },

  6064. };

  6065. 

  6066. static const enum AVPixelFormat vc1_hwaccel_pixfmt_list_420[] = {

  6067. #if CONFIG_DXVA2

  6068.     AV_PIX_FMT_DXVA2_VLD,

  6069. #endif

  6070. #if CONFIG_VAAPI

  6071.     AV_PIX_FMT_VAAPI_VLD,

  6072. #endif

  6073. #if CONFIG_VDPAU

  6074.     AV_PIX_FMT_VDPAU,

  6075. #endif

  6076.     AV_PIX_FMT_YUV420P,

  6077.     AV_PIX_FMT_NONE

  6078. };

  6079. 

  6080. AVCodec ff_vc1_decoder = {

  6081.     .name           = "vc1",

  6082.     .type           = AVMEDIA_TYPE_VIDEO,

  6083.     .id             = AV_CODEC_ID_VC1,

  6084.     .priv_data_size = sizeof(VC1Context),

  6085.     .init           = vc1_decode_init,

  6086.     .close          = ff_vc1_decode_end,

  6087.     .decode         = vc1_decode_frame,

  6088.     .flush          = ff_mpeg_flush,

  6089.     .capabilities   = CODEC_CAP_DR1 | CODEC_CAP_DELAY,

  6090.     .long_name      = NULL_IF_CONFIG_SMALL("SMPTE VC-1"),

  6091.     .pix_fmts       = vc1_hwaccel_pixfmt_list_420,

  6092.     .profiles       = NULL_IF_CONFIG_SMALL(profiles)

  6093. };

  6094. 

  6095. #if CONFIG_WMV3_DECODER

  6096. AVCodec ff_wmv3_decoder = {

  6097.     .name           = "wmv3",

  6098.     .type           = AVMEDIA_TYPE_VIDEO,

  6099.     .id             = AV_CODEC_ID_WMV3,

  6100.     .priv_data_size = sizeof(VC1Context),

  6101.     .init           = vc1_decode_init,

  6102.     .close          = ff_vc1_decode_end,

  6103.     .decode         = vc1_decode_frame,

  6104.     .flush          = ff_mpeg_flush,

  6105.     .capabilities   = CODEC_CAP_DR1 | CODEC_CAP_DELAY,

  6106.     .long_name      = NULL_IF_CONFIG_SMALL("Windows Media Video 9"),

  6107.     .pix_fmts       = vc1_hwaccel_pixfmt_list_420,

  6108.     .profiles       = NULL_IF_CONFIG_SMALL(profiles)

  6109. };

  6110. #endif

  6111. 

  6112. #if CONFIG_WMV3_VDPAU_DECODER

  6113. AVCodec ff_wmv3_vdpau_decoder = {

  6114.     .name           = "wmv3_vdpau",

  6115.     .type           = AVMEDIA_TYPE_VIDEO,

  6116.     .id             = AV_CODEC_ID_WMV3,

  6117.     .priv_data_size = sizeof(VC1Context),

  6118.     .init           = vc1_decode_init,

  6119.     .close          = ff_vc1_decode_end,

  6120.     .decode         = vc1_decode_frame,

  6121.     .capabilities   = CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_HWACCEL_VDPAU,

  6122.     .long_name      = NULL_IF_CONFIG_SMALL("Windows Media Video 9 VDPAU"),

  6123.     .pix_fmts       = (const enum AVPixelFormat[]){ AV_PIX_FMT_VDPAU_WMV3, AV_PIX_FMT_NONE },

  6124.     .profiles       = NULL_IF_CONFIG_SMALL(profiles)

  6125. };

  6126. #endif

  6127. 

  6128. #if CONFIG_VC1_VDPAU_DECODER

  6129. AVCodec ff_vc1_vdpau_decoder = {

  6130.     .name           = "vc1_vdpau",

  6131.     .type           = AVMEDIA_TYPE_VIDEO,

  6132.     .id             = AV_CODEC_ID_VC1,

  6133.     .priv_data_size = sizeof(VC1Context),

  6134.     .init           = vc1_decode_init,

  6135.     .close          = ff_vc1_decode_end,

  6136.     .decode         = vc1_decode_frame,

  6137.     .capabilities   = CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_HWACCEL_VDPAU,

  6138.     .long_name      = NULL_IF_CONFIG_SMALL("SMPTE VC-1 VDPAU"),

  6139.     .pix_fmts       = (const enum AVPixelFormat[]){ AV_PIX_FMT_VDPAU_VC1, AV_PIX_FMT_NONE },

  6140.     .profiles       = NULL_IF_CONFIG_SMALL(profiles)

  6141. };

  6142. #endif

  6143. 

  6144. #if CONFIG_WMV3IMAGE_DECODER

  6145. AVCodec ff_wmv3image_decoder = {

  6146.     .name           = "wmv3image",

  6147.     .type           = AVMEDIA_TYPE_VIDEO,

  6148.     .id             = AV_CODEC_ID_WMV3IMAGE,

  6149.     .priv_data_size = sizeof(VC1Context),

  6150.     .init           = vc1_decode_init,

  6151.     .close          = ff_vc1_decode_end,

  6152.     .decode         = vc1_decode_frame,

  6153.     .capabilities   = CODEC_CAP_DR1,

  6154.     .flush          = vc1_sprite_flush,

  6155.     .long_name      = NULL_IF_CONFIG_SMALL("Windows Media Video 9 Image"),

  6156.     .pix_fmts       = ff_pixfmt_list_420

  6157. };

  6158. #endif

  6159. 

  6160. #if CONFIG_VC1IMAGE_DECODER

  6161. AVCodec ff_vc1image_decoder = {

  6162.     .name           = "vc1image",

  6163.     .type           = AVMEDIA_TYPE_VIDEO,

  6164.     .id             = AV_CODEC_ID_VC1IMAGE,

  6165.     .priv_data_size = sizeof(VC1Context),

  6166.     .init           = vc1_decode_init,

  6167.     .close          = ff_vc1_decode_end,

  6168.     .decode         = vc1_decode_frame,

  6169.     .capabilities   = CODEC_CAP_DR1,

  6170.     .flush          = vc1_sprite_flush,

  6171.     .long_name      = NULL_IF_CONFIG_SMALL("Windows Media Video 9 Image v2"),

  6172.     .pix_fmts       = ff_pixfmt_list_420

  6173. };

  6174. #endif