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.
37428 Commits
32 Branches
2.1GB
Tree: 0d40fbaef0
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '0d40fbaef0'
${ noResults }
FFmpeg/libavcodec/vc1dec.c

5905 lines
230KB
Raw Blame History 

  1. /*

  2.  * VC-1 and WMV3 decoder

  3.  * Copyright (c) 2011 Mashiat Sarker Shakkhar

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

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

  6.  *

  7.  * This file is part of FFmpeg.

  8.  *

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

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

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

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

  13.  *

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

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

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

  17.  * Lesser General Public License for more details.

  18.  *

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

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

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

  22.  */

  23. 

  24. /**

  25.  * @file

  26.  * VC-1 and WMV3 decoder

  27.  */

  28. 

  29. #include "internal.h"

  30. #include "dsputil.h"

  31. #include "avcodec.h"

  32. #include "mpegvideo.h"

  33. #include "h263.h"

  34. #include "vc1.h"

  35. #include "vc1data.h"

  36. #include "vc1acdata.h"

  37. #include "msmpeg4data.h"

  38. #include "unary.h"

  39. #include "simple_idct.h"

  40. #include "mathops.h"

  41. #include "vdpau_internal.h"

  42. 

  43. #undef NDEBUG

  44. #include <assert.h>

  45. 

  46. #define MB_INTRA_VLC_BITS 9

  47. #define DC_VLC_BITS 9

  48. #define AC_VLC_BITS 9

  49. 

  50. 

  51. static const uint16_t vlc_offs[] = {

  52.         0,   520,   552,   616,  1128,  1160,  1224,  1740,  1772,  1836,  1900,  2436,

  53.      2986,  3050,  3610,  4154,  4218,  4746,  5326,  5390,  5902,  6554,  7658,  8342,

  54.      9304,  9988, 10630, 11234, 12174, 13006, 13560, 14232, 14786, 15432, 16350, 17522,

  55.     20372, 21818, 22330, 22394, 23166, 23678, 23742, 24820, 25332, 25396, 26460, 26980,

  56.     27048, 27592, 27600, 27608, 27616, 27624, 28224, 28258, 28290, 28802, 28834, 28866,

  57.     29378, 29412, 29444, 29960, 29994, 30026, 30538, 30572, 30604, 31120, 31154, 31186,

  58.     31714, 31746, 31778, 32306, 32340, 32372

  59. };

  60. 

  61. // offset tables for interlaced picture MVDATA decoding

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

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

  64. 

  65. /**

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

  67.  * @param v The VC1Context to initialize

  68.  * @return Status

  69.  */

  70. int ff_vc1_init_common(VC1Context *v)

  71. {

  72.     static int done = 0;

  73.     int i = 0;

  74.     static VLC_TYPE vlc_table[32372][2];

  75. 

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

  77. 

  78.     /* VLC tables */

  79.     if (!done) {

  80.         INIT_VLC_STATIC(&ff_vc1_bfraction_vlc, VC1_BFRACTION_VLC_BITS, 23,

  81.                         ff_vc1_bfraction_bits, 1, 1,

  82.                         ff_vc1_bfraction_codes, 1, 1, 1 << VC1_BFRACTION_VLC_BITS);

  83.         INIT_VLC_STATIC(&ff_vc1_norm2_vlc, VC1_NORM2_VLC_BITS, 4,

  84.                         ff_vc1_norm2_bits, 1, 1,

  85.                         ff_vc1_norm2_codes, 1, 1, 1 << VC1_NORM2_VLC_BITS);

  86.         INIT_VLC_STATIC(&ff_vc1_norm6_vlc, VC1_NORM6_VLC_BITS, 64,

  87.                         ff_vc1_norm6_bits, 1, 1,

  88.                         ff_vc1_norm6_codes, 2, 2, 556);

  89.         INIT_VLC_STATIC(&ff_vc1_imode_vlc, VC1_IMODE_VLC_BITS, 7,

  90.                         ff_vc1_imode_bits, 1, 1,

  91.                         ff_vc1_imode_codes, 1, 1, 1 << VC1_IMODE_VLC_BITS);

  92.         for (i = 0; i < 3; i++) {

  93.             ff_vc1_ttmb_vlc[i].table           = &vlc_table[vlc_offs[i * 3 + 0]];

  94.             ff_vc1_ttmb_vlc[i].table_allocated = vlc_offs[i * 3 + 1] - vlc_offs[i * 3 + 0];

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

  96.                      ff_vc1_ttmb_bits[i], 1, 1,

  97.                      ff_vc1_ttmb_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);

  98.             ff_vc1_ttblk_vlc[i].table           = &vlc_table[vlc_offs[i * 3 + 1]];

  99.             ff_vc1_ttblk_vlc[i].table_allocated = vlc_offs[i * 3 + 2] - vlc_offs[i * 3 + 1];

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

  101.                      ff_vc1_ttblk_bits[i], 1, 1,

  102.                      ff_vc1_ttblk_codes[i], 1, 1, INIT_VLC_USE_NEW_STATIC);

  103.             ff_vc1_subblkpat_vlc[i].table           = &vlc_table[vlc_offs[i * 3 + 2]];

  104.             ff_vc1_subblkpat_vlc[i].table_allocated = vlc_offs[i * 3 + 3] - vlc_offs[i * 3 + 2];

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

  106.                      ff_vc1_subblkpat_bits[i], 1, 1,

  107.                      ff_vc1_subblkpat_codes[i], 1, 1, INIT_VLC_USE_NEW_STATIC);

  108.         }

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

  110.             ff_vc1_4mv_block_pattern_vlc[i].table           = &vlc_table[vlc_offs[i * 3 + 9]];

  111.             ff_vc1_4mv_block_pattern_vlc[i].table_allocated = vlc_offs[i * 3 + 10] - vlc_offs[i * 3 + 9];

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

  113.                      ff_vc1_4mv_block_pattern_bits[i], 1, 1,

  114.                      ff_vc1_4mv_block_pattern_codes[i], 1, 1, INIT_VLC_USE_NEW_STATIC);

  115.             ff_vc1_cbpcy_p_vlc[i].table           = &vlc_table[vlc_offs[i * 3 + 10]];

  116.             ff_vc1_cbpcy_p_vlc[i].table_allocated = vlc_offs[i * 3 + 11] - vlc_offs[i * 3 + 10];

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

  118.                      ff_vc1_cbpcy_p_bits[i], 1, 1,

  119.                      ff_vc1_cbpcy_p_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);

  120.             ff_vc1_mv_diff_vlc[i].table           = &vlc_table[vlc_offs[i * 3 + 11]];

  121.             ff_vc1_mv_diff_vlc[i].table_allocated = vlc_offs[i * 3 + 12] - vlc_offs[i * 3 + 11];

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

  123.                      ff_vc1_mv_diff_bits[i], 1, 1,

  124.                      ff_vc1_mv_diff_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);

  125.         }

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

  127.             ff_vc1_ac_coeff_table[i].table           = &vlc_table[vlc_offs[i * 2 + 21]];

  128.             ff_vc1_ac_coeff_table[i].table_allocated = vlc_offs[i * 2 + 22] - vlc_offs[i * 2 + 21];

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

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

  131.                      &vc1_ac_tables[i][0][0], 8, 4, INIT_VLC_USE_NEW_STATIC);

  132.             /* initialize interlaced MVDATA tables (2-Ref) */

  133.             ff_vc1_2ref_mvdata_vlc[i].table           = &vlc_table[vlc_offs[i * 2 + 22]];

  134.             ff_vc1_2ref_mvdata_vlc[i].table_allocated = vlc_offs[i * 2 + 23] - vlc_offs[i * 2 + 22];

  135.             init_vlc(&ff_vc1_2ref_mvdata_vlc[i], VC1_2REF_MVDATA_VLC_BITS, 126,

  136.                      ff_vc1_2ref_mvdata_bits[i], 1, 1,

  137.                      ff_vc1_2ref_mvdata_codes[i], 4, 4, INIT_VLC_USE_NEW_STATIC);

  138.         }

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

  140.             /* initialize 4MV MBMODE VLC tables for interlaced frame P picture */

  141.             ff_vc1_intfr_4mv_mbmode_vlc[i].table           = &vlc_table[vlc_offs[i * 3 + 37]];

  142.             ff_vc1_intfr_4mv_mbmode_vlc[i].table_allocated = vlc_offs[i * 3 + 38] - vlc_offs[i * 3 + 37];

  143.             init_vlc(&ff_vc1_intfr_4mv_mbmode_vlc[i], VC1_INTFR_4MV_MBMODE_VLC_BITS, 15,

  144.                      ff_vc1_intfr_4mv_mbmode_bits[i], 1, 1,

  145.                      ff_vc1_intfr_4mv_mbmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);

  146.             /* initialize NON-4MV MBMODE VLC tables for the same */

  147.             ff_vc1_intfr_non4mv_mbmode_vlc[i].table           = &vlc_table[vlc_offs[i * 3 + 38]];

  148.             ff_vc1_intfr_non4mv_mbmode_vlc[i].table_allocated = vlc_offs[i * 3 + 39] - vlc_offs[i * 3 + 38];

  149.             init_vlc(&ff_vc1_intfr_non4mv_mbmode_vlc[i], VC1_INTFR_NON4MV_MBMODE_VLC_BITS, 9,

  150.                      ff_vc1_intfr_non4mv_mbmode_bits[i], 1, 1,

  151.                      ff_vc1_intfr_non4mv_mbmode_codes[i], 1, 1, INIT_VLC_USE_NEW_STATIC);

  152.             /* initialize interlaced MVDATA tables (1-Ref) */

  153.             ff_vc1_1ref_mvdata_vlc[i].table           = &vlc_table[vlc_offs[i * 3 + 39]];

  154.             ff_vc1_1ref_mvdata_vlc[i].table_allocated = vlc_offs[i * 3 + 40] - vlc_offs[i * 3 + 39];

  155.             init_vlc(&ff_vc1_1ref_mvdata_vlc[i], VC1_1REF_MVDATA_VLC_BITS, 72,

  156.                      ff_vc1_1ref_mvdata_bits[i], 1, 1,

  157.                      ff_vc1_1ref_mvdata_codes[i], 4, 4, INIT_VLC_USE_NEW_STATIC);

  158.         }

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

  160.             /* Initialize 2MV Block pattern VLC tables */

  161.             ff_vc1_2mv_block_pattern_vlc[i].table           = &vlc_table[vlc_offs[i + 49]];

  162.             ff_vc1_2mv_block_pattern_vlc[i].table_allocated = vlc_offs[i + 50] - vlc_offs[i + 49];

  163.             init_vlc(&ff_vc1_2mv_block_pattern_vlc[i], VC1_2MV_BLOCK_PATTERN_VLC_BITS, 4,

  164.                      ff_vc1_2mv_block_pattern_bits[i], 1, 1,

  165.                      ff_vc1_2mv_block_pattern_codes[i], 1, 1, INIT_VLC_USE_NEW_STATIC);

  166.         }

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

  168.             /* Initialize interlaced CBPCY VLC tables (Table 124 - Table 131) */

  169.             ff_vc1_icbpcy_vlc[i].table           = &vlc_table[vlc_offs[i * 3 + 53]];

  170.             ff_vc1_icbpcy_vlc[i].table_allocated = vlc_offs[i * 3 + 54] - vlc_offs[i * 3 + 53];

  171.             init_vlc(&ff_vc1_icbpcy_vlc[i], VC1_ICBPCY_VLC_BITS, 63,

  172.                      ff_vc1_icbpcy_p_bits[i], 1, 1,

  173.                      ff_vc1_icbpcy_p_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);

  174.             /* Initialize interlaced field picture MBMODE VLC tables */

  175.             ff_vc1_if_mmv_mbmode_vlc[i].table           = &vlc_table[vlc_offs[i * 3 + 54]];

  176.             ff_vc1_if_mmv_mbmode_vlc[i].table_allocated = vlc_offs[i * 3 + 55] - vlc_offs[i * 3 + 54];

  177.             init_vlc(&ff_vc1_if_mmv_mbmode_vlc[i], VC1_IF_MMV_MBMODE_VLC_BITS, 8,

  178.                      ff_vc1_if_mmv_mbmode_bits[i], 1, 1,

  179.                      ff_vc1_if_mmv_mbmode_codes[i], 1, 1, INIT_VLC_USE_NEW_STATIC);

  180.             ff_vc1_if_1mv_mbmode_vlc[i].table           = &vlc_table[vlc_offs[i * 3 + 55]];

  181.             ff_vc1_if_1mv_mbmode_vlc[i].table_allocated = vlc_offs[i * 3 + 56] - vlc_offs[i * 3 + 55];

  182.             init_vlc(&ff_vc1_if_1mv_mbmode_vlc[i], VC1_IF_1MV_MBMODE_VLC_BITS, 6,

  183.                      ff_vc1_if_1mv_mbmode_bits[i], 1, 1,

  184.                      ff_vc1_if_1mv_mbmode_codes[i], 1, 1, INIT_VLC_USE_NEW_STATIC);

  185.         }

  186.         done = 1;

  187.     }

  188. 

  189.     /* Other defaults */

  190.     v->pq      = -1;

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

  192. 

  193.     return 0;

  194. }

  195. 

  196. /***********************************************************************/

  197. /**

  198.  * @name VC-1 Bitplane decoding

  199.  * @see 8.7, p56

  200.  * @{

  201.  */

  202. 

  203. /**

  204.  * Imode types

  205.  * @{

  206.  */

  207. enum Imode {

  208.     IMODE_RAW,

  209.     IMODE_NORM2,

  210.     IMODE_DIFF2,

  211.     IMODE_NORM6,

  212.     IMODE_DIFF6,

  213.     IMODE_ROWSKIP,

  214.     IMODE_COLSKIP

  215. };

  216. /** @} */ //imode defines

  217. 

  218. 

  219. /** @} */ //Bitplane group

  220. 

  221. static void vc1_put_signed_blocks_clamped(VC1Context *v)

  222. {

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

  224.     int topleft_mb_pos, top_mb_pos;

  225.     int stride_y, fieldtx;

  226.     int v_dist;

  227. 

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

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

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

  231.      * present as well.

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

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

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

  235.     if (!s->first_slice_line) {

  236.         if (s->mb_x) {

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

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

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

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

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

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

  243.                                              stride_y);

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

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

  246.                                              stride_y);

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

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

  249.                                              stride_y);

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

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

  252.                                              stride_y);

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

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

  255.                                              s->uvlinesize);

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

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

  258.                                              s->uvlinesize);

  259.         }

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

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

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

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

  264.             v_dist     = fieldtx ? 15 : 8;

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

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

  267.                                              stride_y);

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

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

  270.                                              stride_y);

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

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

  273.                                              stride_y);

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

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

  276.                                              stride_y);

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

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

  279.                                              s->uvlinesize);

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

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

  282.                                              s->uvlinesize);

  283.         }

  284.     }

  285. 

  286. #define inc_blk_idx(idx) do { \

  287.         idx++; \

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

  289.             idx = 0; \

  290.     } while (0)

  291. 

  292.     inc_blk_idx(v->topleft_blk_idx);

  293.     inc_blk_idx(v->top_blk_idx);

  294.     inc_blk_idx(v->left_blk_idx);

  295.     inc_blk_idx(v->cur_blk_idx);

  296. }

  297. 

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

  299. {

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

  301.     int j;

  302.     if (!s->first_slice_line) {

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

  304.         if (s->mb_x)

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

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

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

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

  309.             if (s->mb_x)

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

  311.         }

  312.     }

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

  314. 

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

  316.         if (s->mb_x) {

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

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

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

  320.         }

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

  322.     }

  323. }

  324. 

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

  326. {

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

  328.     int j;

  329. 

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

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

  332.     if (!s->first_slice_line) {

  333.         if (s->mb_x) {

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

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

  336. 

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

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

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

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

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

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

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

  344.                     }

  345.                 }

  346.             }

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

  348.         }

  349. 

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

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

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

  353. 

  354.                 if (s->mb_x)

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

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

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

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

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

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

  361.                     }

  362.                 }

  363.             }

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

  365.         }

  366. 

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

  368.             if (s->mb_x) {

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

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

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

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

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

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

  375.                     }

  376.                 }

  377.             }

  378. 

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

  380.                 if (s->mb_x)

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

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

  383.                 if (s->mb_x) {

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

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

  386.                     }

  387.                 }

  388.             }

  389.         }

  390.     }

  391. }

  392. 

  393. static void vc1_smooth_overlap_filter_iblk(VC1Context *v)

  394. {

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

  396.     int mb_pos;

  397. 

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

  399.         return;

  400. 

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

  402. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  422.             }

  423.         }

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

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

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

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

  428. 

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

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

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

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

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

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

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

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

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

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

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

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

  441.                 }

  442.             }

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

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

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

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

  447.         }

  448.     }

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

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

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

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

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

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

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

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

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

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

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

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

  461.             }

  462.         }

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

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

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

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

  467.     }

  468. }

  469. 

  470. /** Do motion compensation over 1 macroblock

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

  472.  */

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

  474. {

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

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

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

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

  479.     int off, off_uv;

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

  481. 

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

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

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

  485.         return;

  486. 

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

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

  489. 

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

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

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

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

  494.     }

  495. 

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

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

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

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

  500. 

  501.     if (v->field_mode &&

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

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

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

  505.     }

  506. 

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

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

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

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

  511.     }

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

  513.         if (!dir) {

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

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

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

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

  518.             } else {

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

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

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

  522.             }

  523.         } else {

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

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

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

  527.         }

  528.     } else {

  529.         if (!dir) {

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

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

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

  533.         } else {

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

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

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

  537.         }

  538.     }

  539. 

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

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

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

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

  544. 

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

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

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

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

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

  550.     } else {

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

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

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

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

  555.     }

  556. 

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

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

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

  560. 

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

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

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

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

  565.     }

  566. 

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

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

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

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

  571.     }

  572. 

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

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

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

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

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

  578. 

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

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

  581.                                 17 + s->mspel * 2, 17 + s->mspel * 2,

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

  583.                                 s->h_edge_pos, v_edge_pos);

  584.         srcY = s->edge_emu_buffer;

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

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

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

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

  589.         srcU = uvbuf;

  590.         srcV = uvbuf + 16;

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

  592.         if (v->rangeredfrm) {

  593.             int i, j;

  594.             uint8_t *src, *src2;

  595. 

  596.             src = srcY;

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

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

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

  600.                 src += s->linesize;

  601.             }

  602.             src  = srcU;

  603.             src2 = srcV;

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

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

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

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

  608.                 }

  609.                 src  += s->uvlinesize;

  610.                 src2 += s->uvlinesize;

  611.             }

  612.         }

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

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

  615.             int i, j;

  616.             uint8_t *src, *src2;

  617. 

  618.             src = srcY;

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

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

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

  622.                 src += s->linesize;

  623.             }

  624.             src  = srcU;

  625.             src2 = srcV;

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

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

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

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

  630.                 }

  631.                 src  += s->uvlinesize;

  632.                 src2 += s->uvlinesize;

  633.             }

  634.         }

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

  636.     }

  637. 

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

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

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

  641.     } else {

  642.         off    = 0;

  643.         off_uv = 0;

  644.     }

  645.     if (s->mspel) {

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

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

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

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

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

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

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

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

  654.         if (!v->rnd)

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

  656.         else

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

  658.     }

  659. 

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

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

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

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

  664.     if (!v->rnd) {

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

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

  667.     } else {

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

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

  670.     }

  671. }

  672. 

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

  674. {

  675.     if (a < b) {

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

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

  678.     } else {

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

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

  681.     }

  682. }

  683. 

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

  685.  */

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

  687. {

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

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

  690.     uint8_t *srcY;

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

  692.     int off;

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

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

  695. 

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

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

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

  699.         return;

  700. 

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

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

  703. 

  704.     if (!dir) {

  705.         if (v->field_mode) {

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

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

  708.             else

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

  710.         } else

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

  712.     } else

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

  714. 

  715.     if (v->field_mode) {

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

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

  718.     }

  719. 

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

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

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

  723.         int tx, ty;

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

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

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

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

  728.             opp_count  += f;

  729.             same_count += 1 - f;

  730.         }

  731.         f = opp_count > same_count;

  732.         switch (f ? opp_count : same_count) {

  733.         case 4:

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

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

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

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

  738.             break;

  739.         case 3:

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

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

  742.             break;

  743.         case 2:

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

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

  746.             break;

  747.         }

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

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

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

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

  752.     }

  753. 

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

  755.         int qx, qy;

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

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

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

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

  760. 

  761.         if (qx < -17)

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

  763.         else if (qx > width)

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

  765.         if (qy < -18)

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

  767.         else if (qy > height + 1)

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

  769.     }

  770. 

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

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

  773.     else

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

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

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

  777. 

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

  779.     if (!fieldmv)

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

  781.     else

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

  783. 

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

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

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

  787.     } else {

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

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

  790.             if (src_y & 1)

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

  792.             else

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

  794.         } else {

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

  796.         }

  797.     }

  798. 

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

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

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

  802. 

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

  804.         v_edge_pos--;

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

  806.         src_y--;

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

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

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

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

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

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

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

  814.                                 9 + s->mspel * 2, (9 + s->mspel * 2) << fieldmv,

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

  816.                                 s->h_edge_pos, v_edge_pos);

  817.         srcY = s->edge_emu_buffer;

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

  819.         if (v->rangeredfrm) {

  820.             int i, j;

  821.             uint8_t *src;

  822. 

  823.             src = srcY;

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

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

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

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

  828.             }

  829.         }

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

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

  832.             int i, j;

  833.             uint8_t *src;

  834. 

  835.             src = srcY;

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

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

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

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

  840.             }

  841.         }

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

  843.     }

  844. 

  845.     if (s->mspel) {

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

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

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

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

  850.         if (!v->rnd)

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

  852.         else

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

  854.     }

  855. }

  856. 

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

  858. {

  859.     int idx, i;

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

  861. 

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

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

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

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

  866.     if (!idx) {

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

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

  869.         return 4;

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

  871.         switch (idx) {

  872.         case 0x1:

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

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

  875.             return 3;

  876.         case 0x2:

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

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

  879.             return 3;

  880.         case 0x4:

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

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

  883.             return 3;

  884.         case 0x8:

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

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

  887.             return 3;

  888.         }

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

  890.         int t1 = 0, t2 = 0;

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

  892.             if (!a[i]) {

  893.                 t1 = i;

  894.                 break;

  895.             }

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

  897.             if (!a[i]) {

  898.                 t2 = i;

  899.                 break;

  900.             }

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

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

  903.         return 2;

  904.     } else {

  905.         return 0;

  906.     }

  907.     return -1;

  908. }

  909. 

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

  911.  */

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

  913. {

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

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

  916.     uint8_t *srcU, *srcV;

  917.     int uvmx, uvmy, uvsrc_x, uvsrc_y;

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

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

  920.     int valid_count;

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

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

  923. 

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

  925.         return;

  926.     if (s->flags & CODEC_FLAG_GRAY)

  927.         return;

  928. 

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

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

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

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

  933.         if (v->field_mode)

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

  935.     }

  936. 

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

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

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

  940.         if (!valid_count) {

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

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

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

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

  945.         }

  946.     } else {

  947.         int dominant = 0;

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

  949.             dominant = 1;

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

  951.         if (dominant)

  952.             chroma_ref_type = !v->cur_field_type;

  953.     }

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

  955.         return;

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

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

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

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

  960. 

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

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

  963. 

  964.     if (v->fastuvmc) {

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

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

  967.     }

  968.     // Field conversion bias

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

  970.         uvmy += 2 - 4 * chroma_ref_type;

  971. 

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

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

  974. 

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

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

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

  978.     } else {

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

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

  981.     }

  982. 

  983.     if (!dir) {

  984.         if (v->field_mode) {

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

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

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

  988.             } else {

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

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

  991.             }

  992.         } else {

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

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

  995.         }

  996.     } else {

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

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

  999.     }

  1000. 

  1001.     if (v->field_mode) {

  1002.         if (chroma_ref_type) {

  1003.             srcU += s->current_picture_ptr->f.linesize[1];

  1004.             srcV += s->current_picture_ptr->f.linesize[2];

  1005.         }

  1006.         off = v->second_field ? s->current_picture_ptr->f.linesize[1] : 0;

  1007.     }

  1008. 

  1009.     if (v->rangeredfrm || (v->mv_mode == MV_PMODE_INTENSITY_COMP)

  1010.         || s->h_edge_pos < 18 || v_edge_pos < 18

  1011.         || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 9

  1012.         || (unsigned)uvsrc_y > (v_edge_pos    >> 1) - 9) {

  1013.         s->dsp.emulated_edge_mc(s->edge_emu_buffer     , srcU, s->uvlinesize,

  1014.                                 8 + 1, 8 + 1, uvsrc_x, uvsrc_y,

  1015.                                 s->h_edge_pos >> 1, v_edge_pos >> 1);

  1016.         s->dsp.emulated_edge_mc(s->edge_emu_buffer + 16, srcV, s->uvlinesize,

  1017.                                 8 + 1, 8 + 1, uvsrc_x, uvsrc_y,

  1018.                                 s->h_edge_pos >> 1, v_edge_pos >> 1);

  1019.         srcU = s->edge_emu_buffer;

  1020.         srcV = s->edge_emu_buffer + 16;

  1021. 

  1022.         /* if we deal with range reduction we need to scale source blocks */

  1023.         if (v->rangeredfrm) {

  1024.             int i, j;

  1025.             uint8_t *src, *src2;

  1026. 

  1027.             src  = srcU;

  1028.             src2 = srcV;

  1029.             for (j = 0; j < 9; j++) {

  1030.                 for (i = 0; i < 9; i++) {

  1031.                     src[i]  = ((src[i]  - 128) >> 1) + 128;

  1032.                     src2[i] = ((src2[i] - 128) >> 1) + 128;

  1033.                 }

  1034.                 src  += s->uvlinesize;

  1035.                 src2 += s->uvlinesize;

  1036.             }

  1037.         }

  1038.         /* if we deal with intensity compensation we need to scale source blocks */

  1039.         if (v->mv_mode == MV_PMODE_INTENSITY_COMP) {

  1040.             int i, j;

  1041.             uint8_t *src, *src2;

  1042. 

  1043.             src  = srcU;

  1044.             src2 = srcV;

  1045.             for (j = 0; j < 9; j++) {

  1046.                 for (i = 0; i < 9; i++) {

  1047.                     src[i]  = v->lutuv[src[i]];

  1048.                     src2[i] = v->lutuv[src2[i]];

  1049.                 }

  1050.                 src  += s->uvlinesize;

  1051.                 src2 += s->uvlinesize;

  1052.             }

  1053.         }

  1054.     }

  1055. 

  1056.     /* Chroma MC always uses qpel bilinear */

  1057.     uvmx = (uvmx & 3) << 1;

  1058.     uvmy = (uvmy & 3) << 1;

  1059.     if (!v->rnd) {

  1060.         dsp->put_h264_chroma_pixels_tab[0](s->dest[1] + off, srcU, s->uvlinesize, 8, uvmx, uvmy);

  1061.         dsp->put_h264_chroma_pixels_tab[0](s->dest[2] + off, srcV, s->uvlinesize, 8, uvmx, uvmy);

  1062.     } else {

  1063.         v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1] + off, srcU, s->uvlinesize, 8, uvmx, uvmy);

  1064.         v->vc1dsp.put_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2] + off, srcV, s->uvlinesize, 8, uvmx, uvmy);

  1065.     }

  1066. }

  1067. 

  1068. /** Do motion compensation for 4-MV field chroma macroblock (both U and V)

  1069.  */

  1070. static void vc1_mc_4mv_chroma4(VC1Context *v)

  1071. {

  1072.     MpegEncContext *s = &v->s;

  1073.     DSPContext *dsp = &v->s.dsp;

  1074.     uint8_t *srcU, *srcV;

  1075.     int uvsrc_x, uvsrc_y;

  1076.     int uvmx_field[4], uvmy_field[4];

  1077.     int i, off, tx, ty;

  1078.     int fieldmv = v->blk_mv_type[s->block_index[0]];

  1079.     static const int s_rndtblfield[16] = { 0, 0, 1, 2, 4, 4, 5, 6, 2, 2, 3, 8, 6, 6, 7, 12 };

  1080.     int v_dist = fieldmv ? 1 : 4; // vertical offset for lower sub-blocks

  1081.     int v_edge_pos = s->v_edge_pos >> 1;

  1082. 

  1083.     if (!v->s.last_picture.f.data[0])

  1084.         return;

  1085.     if (s->flags & CODEC_FLAG_GRAY)

  1086.         return;

  1087. 

  1088.     for (i = 0; i < 4; i++) {

  1089.         tx = s->mv[0][i][0];

  1090.         uvmx_field[i] = (tx + ((tx & 3) == 3)) >> 1;

  1091.         ty = s->mv[0][i][1];

  1092.         if (fieldmv)

  1093.             uvmy_field[i] = (ty >> 4) * 8 + s_rndtblfield[ty & 0xF];

  1094.         else

  1095.             uvmy_field[i] = (ty + ((ty & 3) == 3)) >> 1;

  1096.     }

  1097. 

  1098.     for (i = 0; i < 4; i++) {

  1099.         off = (i & 1) * 4 + ((i & 2) ? v_dist * s->uvlinesize : 0);

  1100.         uvsrc_x = s->mb_x * 8 +  (i & 1) * 4           + (uvmx_field[i] >> 2);

  1101.         uvsrc_y = s->mb_y * 8 + ((i & 2) ? v_dist : 0) + (uvmy_field[i] >> 2);

  1102.         // FIXME: implement proper pull-back (see vc1cropmv.c, vc1CROPMV_ChromaPullBack())

  1103.         uvsrc_x = av_clip(uvsrc_x, -8, s->avctx->coded_width  >> 1);

  1104.         uvsrc_y = av_clip(uvsrc_y, -8, s->avctx->coded_height >> 1);

  1105.         srcU = s->last_picture.f.data[1] + uvsrc_y * s->uvlinesize + uvsrc_x;

  1106.         srcV = s->last_picture.f.data[2] + uvsrc_y * s->uvlinesize + uvsrc_x;

  1107.         uvmx_field[i] = (uvmx_field[i] & 3) << 1;

  1108.         uvmy_field[i] = (uvmy_field[i] & 3) << 1;

  1109. 

  1110.         if (fieldmv && !(uvsrc_y & 1))

  1111.             v_edge_pos--;

  1112.         if (fieldmv && (uvsrc_y & 1) && uvsrc_y < 2)

  1113.             uvsrc_y--;

  1114.         if ((v->mv_mode == MV_PMODE_INTENSITY_COMP)

  1115.             || s->h_edge_pos < 10 || v_edge_pos < (5 << fieldmv)

  1116.             || (unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 5

  1117.             || (unsigned)uvsrc_y > v_edge_pos - (5 << fieldmv)) {

  1118.             s->dsp.emulated_edge_mc(s->edge_emu_buffer, srcU, s->uvlinesize,

  1119.                                     5, (5 << fieldmv), uvsrc_x, uvsrc_y,

  1120.                                     s->h_edge_pos >> 1, v_edge_pos);

  1121.             s->dsp.emulated_edge_mc(s->edge_emu_buffer + 16, srcV, s->uvlinesize,

  1122.                                     5, (5 << fieldmv), uvsrc_x, uvsrc_y,

  1123.                                     s->h_edge_pos >> 1, v_edge_pos);

  1124.             srcU = s->edge_emu_buffer;

  1125.             srcV = s->edge_emu_buffer + 16;

  1126. 

  1127.             /* if we deal with intensity compensation we need to scale source blocks */

  1128.             if (v->mv_mode == MV_PMODE_INTENSITY_COMP) {

  1129.                 int i, j;

  1130.                 uint8_t *src, *src2;

  1131. 

  1132.                 src  = srcU;

  1133.                 src2 = srcV;

  1134.                 for (j = 0; j < 5; j++) {

  1135.                     for (i = 0; i < 5; i++) {

  1136.                         src[i]  = v->lutuv[src[i]];

  1137.                         src2[i] = v->lutuv[src2[i]];

  1138.                     }

  1139.                     src  += s->uvlinesize << 1;

  1140.                     src2 += s->uvlinesize << 1;

  1141.                 }

  1142.             }

  1143.         }

  1144.         if (!v->rnd) {

  1145.             dsp->put_h264_chroma_pixels_tab[1](s->dest[1] + off, srcU, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);

  1146.             dsp->put_h264_chroma_pixels_tab[1](s->dest[2] + off, srcV, s->uvlinesize << fieldmv, 4, uvmx_field[i], uvmy_field[i]);

  1147.         } else {

  1148.             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]);

  1149.             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]);

  1150.         }

  1151.     }

  1152. }

  1153. 

  1154. /***********************************************************************/

  1155. /**

  1156.  * @name VC-1 Block-level functions

  1157.  * @see 7.1.4, p91 and 8.1.1.7, p(1)04

  1158.  * @{

  1159.  */

  1160. 

  1161. /**

  1162.  * @def GET_MQUANT

  1163.  * @brief Get macroblock-level quantizer scale

  1164.  */

  1165. #define GET_MQUANT()                                           \

  1166.     if (v->dquantfrm) {                                        \

  1167.         int edges = 0;                                         \

  1168.         if (v->dqprofile == DQPROFILE_ALL_MBS) {               \

  1169.             if (v->dqbilevel) {                                \

  1170.                 mquant = (get_bits1(gb)) ? v->altpq : v->pq;   \

  1171.             } else {                                           \

  1172.                 mqdiff = get_bits(gb, 3);                      \

  1173.                 if (mqdiff != 7)                               \

  1174.                     mquant = v->pq + mqdiff;                   \

  1175.                 else                                           \

  1176.                     mquant = get_bits(gb, 5);                  \

  1177.             }                                                  \

  1178.         }                                                      \

  1179.         if (v->dqprofile == DQPROFILE_SINGLE_EDGE)             \

  1180.             edges = 1 << v->dqsbedge;                          \

  1181.         else if (v->dqprofile == DQPROFILE_DOUBLE_EDGES)       \

  1182.             edges = (3 << v->dqsbedge) % 15;                   \

  1183.         else if (v->dqprofile == DQPROFILE_FOUR_EDGES)         \

  1184.             edges = 15;                                        \

  1185.         if ((edges&1) && !s->mb_x)                             \

  1186.             mquant = v->altpq;                                 \

  1187.         if ((edges&2) && s->first_slice_line)                  \

  1188.             mquant = v->altpq;                                 \

  1189.         if ((edges&4) && s->mb_x == (s->mb_width - 1))         \

  1190.             mquant = v->altpq;                                 \

  1191.         if ((edges&8) && s->mb_y == (s->mb_height - 1))        \

  1192.             mquant = v->altpq;                                 \

  1193.     }

  1194. 

  1195. /**

  1196.  * @def GET_MVDATA(_dmv_x, _dmv_y)

  1197.  * @brief Get MV differentials

  1198.  * @see MVDATA decoding from 8.3.5.2, p(1)20

  1199.  * @param _dmv_x Horizontal differential for decoded MV

  1200.  * @param _dmv_y Vertical differential for decoded MV

  1201.  */

  1202. #define GET_MVDATA(_dmv_x, _dmv_y)                                      \

  1203.     index = 1 + get_vlc2(gb, ff_vc1_mv_diff_vlc[s->mv_table_index].table, \

  1204.                          VC1_MV_DIFF_VLC_BITS, 2);                      \

  1205.     if (index > 36) {                                                   \

  1206.         mb_has_coeffs = 1;                                              \

  1207.         index -= 37;                                                    \

  1208.     } else                                                              \

  1209.         mb_has_coeffs = 0;                                              \

  1210.     s->mb_intra = 0;                                                    \

  1211.     if (!index) {                                                       \

  1212.         _dmv_x = _dmv_y = 0;                                            \

  1213.     } else if (index == 35) {                                           \

  1214.         _dmv_x = get_bits(gb, v->k_x - 1 + s->quarter_sample);          \

  1215.         _dmv_y = get_bits(gb, v->k_y - 1 + s->quarter_sample);          \

  1216.     } else if (index == 36) {                                           \

  1217.         _dmv_x = 0;                                                     \

  1218.         _dmv_y = 0;                                                     \

  1219.         s->mb_intra = 1;                                                \

  1220.     } else {                                                            \

  1221.         index1 = index % 6;                                             \

  1222.         if (!s->quarter_sample && index1 == 5) val = 1;                 \

  1223.         else                                   val = 0;                 \

  1224.         if (size_table[index1] - val > 0)                               \

  1225.             val = get_bits(gb, size_table[index1] - val);               \

  1226.         else                                   val = 0;                 \

  1227.         sign = 0 - (val&1);                                             \

  1228.         _dmv_x = (sign ^ ((val>>1) + offset_table[index1])) - sign;     \

  1229.                                                                         \

  1230.         index1 = index / 6;                                             \

  1231.         if (!s->quarter_sample && index1 == 5) val = 1;                 \

  1232.         else                                   val = 0;                 \

  1233.         if (size_table[index1] - val > 0)                               \

  1234.             val = get_bits(gb, size_table[index1] - val);               \

  1235.         else                                   val = 0;                 \

  1236.         sign = 0 - (val & 1);                                           \

  1237.         _dmv_y = (sign ^ ((val >> 1) + offset_table[index1])) - sign;   \

  1238.     }

  1239. 

  1240. static av_always_inline void get_mvdata_interlaced(VC1Context *v, int *dmv_x,

  1241.                                                    int *dmv_y, int *pred_flag)

  1242. {

  1243.     int index, index1;

  1244.     int extend_x = 0, extend_y = 0;

  1245.     GetBitContext *gb = &v->s.gb;

  1246.     int bits, esc;

  1247.     int val, sign;

  1248.     const int* offs_tab;

  1249. 

  1250.     if (v->numref) {

  1251.         bits = VC1_2REF_MVDATA_VLC_BITS;

  1252.         esc  = 125;

  1253.     } else {

  1254.         bits = VC1_1REF_MVDATA_VLC_BITS;

  1255.         esc  = 71;

  1256.     }

  1257.     switch (v->dmvrange) {

  1258.     case 1:

  1259.         extend_x = 1;

  1260.         break;

  1261.     case 2:

  1262.         extend_y = 1;

  1263.         break;

  1264.     case 3:

  1265.         extend_x = extend_y = 1;

  1266.         break;

  1267.     }

  1268.     index = get_vlc2(gb, v->imv_vlc->table, bits, 3);

  1269.     if (index == esc) {

  1270.         *dmv_x = get_bits(gb, v->k_x);

  1271.         *dmv_y = get_bits(gb, v->k_y);

  1272.         if (v->numref) {

  1273.             *pred_flag = *dmv_y & 1;

  1274.             *dmv_y     = (*dmv_y + *pred_flag) >> 1;

  1275.         }

  1276.     }

  1277.     else {

  1278.         if (extend_x)

  1279.             offs_tab = offset_table2;

  1280.         else

  1281.             offs_tab = offset_table1;

  1282.         index1 = (index + 1) % 9;

  1283.         if (index1 != 0) {

  1284.             val    = get_bits(gb, index1 + extend_x);

  1285.             sign   = 0 -(val & 1);

  1286.             *dmv_x = (sign ^ ((val >> 1) + offs_tab[index1])) - sign;

  1287.         } else

  1288.             *dmv_x = 0;

  1289.         if (extend_y)

  1290.             offs_tab = offset_table2;

  1291.         else

  1292.             offs_tab = offset_table1;

  1293.         index1 = (index + 1) / 9;

  1294.         if (index1 > v->numref) {

  1295.             val    = get_bits(gb, (index1 + (extend_y << v->numref)) >> v->numref);

  1296.             sign   = 0 - (val & 1);

  1297.             *dmv_y = (sign ^ ((val >> 1) + offs_tab[index1 >> v->numref])) - sign;

  1298.         } else

  1299.             *dmv_y = 0;

  1300.         if (v->numref)

  1301.             *pred_flag = index1 & 1;

  1302.     }

  1303. }

  1304. 

  1305. static av_always_inline int scaleforsame_x(VC1Context *v, int n /* MV */, int dir)

  1306. {

  1307.     int scaledvalue, refdist;

  1308.     int scalesame1, scalesame2;

  1309.     int scalezone1_x, zone1offset_x;

  1310.     int table_index = dir ^ v->second_field;

  1311. 

  1312.     if (v->s.pict_type != AV_PICTURE_TYPE_B)

  1313.         refdist = v->refdist;

  1314.     else

  1315.         refdist = dir ? v->brfd : v->frfd;

  1316.     if (refdist > 3)

  1317.         refdist = 3;

  1318.     scalesame1    = vc1_field_mvpred_scales[table_index][1][refdist];

  1319.     scalesame2    = vc1_field_mvpred_scales[table_index][2][refdist];

  1320.     scalezone1_x  = vc1_field_mvpred_scales[table_index][3][refdist];

  1321.     zone1offset_x = vc1_field_mvpred_scales[table_index][5][refdist];

  1322. 

  1323.     if (FFABS(n) > 255)

  1324.         scaledvalue = n;

  1325.     else {

  1326.         if (FFABS(n) < scalezone1_x)

  1327.             scaledvalue = (n * scalesame1) >> 8;

  1328.         else {

  1329.             if (n < 0)

  1330.                 scaledvalue = ((n * scalesame2) >> 8) - zone1offset_x;

  1331.             else

  1332.                 scaledvalue = ((n * scalesame2) >> 8) + zone1offset_x;

  1333.         }

  1334.     }

  1335.     return av_clip(scaledvalue, -v->range_x, v->range_x - 1);

  1336. }

  1337. 

  1338. static av_always_inline int scaleforsame_y(VC1Context *v, int i, int n /* MV */, int dir)

  1339. {

  1340.     int scaledvalue, refdist;

  1341.     int scalesame1, scalesame2;

  1342.     int scalezone1_y, zone1offset_y;

  1343.     int table_index = dir ^ v->second_field;

  1344. 

  1345.     if (v->s.pict_type != AV_PICTURE_TYPE_B)

  1346.         refdist = v->refdist;

  1347.     else

  1348.         refdist = dir ? v->brfd : v->frfd;

  1349.     if (refdist > 3)

  1350.         refdist = 3;

  1351.     scalesame1    = vc1_field_mvpred_scales[table_index][1][refdist];

  1352.     scalesame2    = vc1_field_mvpred_scales[table_index][2][refdist];

  1353.     scalezone1_y  = vc1_field_mvpred_scales[table_index][4][refdist];

  1354.     zone1offset_y = vc1_field_mvpred_scales[table_index][6][refdist];

  1355. 

  1356.     if (FFABS(n) > 63)

  1357.         scaledvalue = n;

  1358.     else {

  1359.         if (FFABS(n) < scalezone1_y)

  1360.             scaledvalue = (n * scalesame1) >> 8;

  1361.         else {

  1362.             if (n < 0)

  1363.                 scaledvalue = ((n * scalesame2) >> 8) - zone1offset_y;

  1364.             else

  1365.                 scaledvalue = ((n * scalesame2) >> 8) + zone1offset_y;

  1366.         }

  1367.     }

  1368. 

  1369.     if (v->cur_field_type && !v->ref_field_type[dir])

  1370.         return av_clip(scaledvalue, -v->range_y / 2 + 1, v->range_y / 2);

  1371.     else

  1372.         return av_clip(scaledvalue, -v->range_y / 2, v->range_y / 2 - 1);

  1373. }

  1374. 

  1375. static av_always_inline int scaleforopp_x(VC1Context *v, int n /* MV */)

  1376. {

  1377.     int scalezone1_x, zone1offset_x;

  1378.     int scaleopp1, scaleopp2, brfd;

  1379.     int scaledvalue;

  1380. 

  1381.     brfd = FFMIN(v->brfd, 3);

  1382.     scalezone1_x  = vc1_b_field_mvpred_scales[3][brfd];

  1383.     zone1offset_x = vc1_b_field_mvpred_scales[5][brfd];

  1384.     scaleopp1     = vc1_b_field_mvpred_scales[1][brfd];

  1385.     scaleopp2     = vc1_b_field_mvpred_scales[2][brfd];

  1386. 

  1387.     if (FFABS(n) > 255)

  1388.         scaledvalue = n;

  1389.     else {

  1390.         if (FFABS(n) < scalezone1_x)

  1391.             scaledvalue = (n * scaleopp1) >> 8;

  1392.         else {

  1393.             if (n < 0)

  1394.                 scaledvalue = ((n * scaleopp2) >> 8) - zone1offset_x;

  1395.             else

  1396.                 scaledvalue = ((n * scaleopp2) >> 8) + zone1offset_x;

  1397.         }

  1398.     }

  1399.     return av_clip(scaledvalue, -v->range_x, v->range_x - 1);

  1400. }

  1401. 

  1402. static av_always_inline int scaleforopp_y(VC1Context *v, int n /* MV */, int dir)

  1403. {

  1404.     int scalezone1_y, zone1offset_y;

  1405.     int scaleopp1, scaleopp2, brfd;

  1406.     int scaledvalue;

  1407. 

  1408.     brfd = FFMIN(v->brfd, 3);

  1409.     scalezone1_y  = vc1_b_field_mvpred_scales[4][brfd];

  1410.     zone1offset_y = vc1_b_field_mvpred_scales[6][brfd];

  1411.     scaleopp1     = vc1_b_field_mvpred_scales[1][brfd];

  1412.     scaleopp2     = vc1_b_field_mvpred_scales[2][brfd];

  1413. 

  1414.     if (FFABS(n) > 63)

  1415.         scaledvalue = n;

  1416.     else {

  1417.         if (FFABS(n) < scalezone1_y)

  1418.             scaledvalue = (n * scaleopp1) >> 8;

  1419.         else {

  1420.             if (n < 0)

  1421.                 scaledvalue = ((n * scaleopp2) >> 8) - zone1offset_y;

  1422.             else

  1423.                 scaledvalue = ((n * scaleopp2) >> 8) + zone1offset_y;

  1424.         }

  1425.     }

  1426.     if (v->cur_field_type && !v->ref_field_type[dir]) {

  1427.         return av_clip(scaledvalue, -v->range_y / 2 + 1, v->range_y / 2);

  1428.     } else {

  1429.         return av_clip(scaledvalue, -v->range_y / 2, v->range_y / 2 - 1);

  1430.     }

  1431. }

  1432. 

  1433. static av_always_inline int scaleforsame(VC1Context *v, int i, int n /* MV */,

  1434.                                          int dim, int dir)

  1435. {

  1436.     int brfd, scalesame;

  1437.     int hpel = 1 - v->s.quarter_sample;

  1438. 

  1439.     n >>= hpel;

  1440.     if (v->s.pict_type != AV_PICTURE_TYPE_B || v->second_field || !dir) {

  1441.         if (dim)

  1442.             n = scaleforsame_y(v, i, n, dir) << hpel;

  1443.         else

  1444.             n = scaleforsame_x(v, n, dir) << hpel;

  1445.         return n;

  1446.     }

  1447.     brfd      = FFMIN(v->brfd, 3);

  1448.     scalesame = vc1_b_field_mvpred_scales[0][brfd];

  1449. 

  1450.     n = (n * scalesame >> 8) << hpel;

  1451.     return n;

  1452. }

  1453. 

  1454. static av_always_inline int scaleforopp(VC1Context *v, int n /* MV */,

  1455.                                         int dim, int dir)

  1456. {

  1457.     int refdist, scaleopp;

  1458.     int hpel = 1 - v->s.quarter_sample;

  1459. 

  1460.     n >>= hpel;

  1461.     if (v->s.pict_type == AV_PICTURE_TYPE_B && !v->second_field && dir == 1) {

  1462.         if (dim)

  1463.             n = scaleforopp_y(v, n, dir) << hpel;

  1464.         else

  1465.             n = scaleforopp_x(v, n) << hpel;

  1466.         return n;

  1467.     }

  1468.     if (v->s.pict_type != AV_PICTURE_TYPE_B)

  1469.         refdist = FFMIN(v->refdist, 3);

  1470.     else

  1471.         refdist = dir ? v->brfd : v->frfd;

  1472.     scaleopp = vc1_field_mvpred_scales[dir ^ v->second_field][0][refdist];

  1473. 

  1474.     n = (n * scaleopp >> 8) << hpel;

  1475.     return n;

  1476. }

  1477. 

  1478. /** Predict and set motion vector

  1479.  */

  1480. static inline void vc1_pred_mv(VC1Context *v, int n, int dmv_x, int dmv_y,

  1481.                                int mv1, int r_x, int r_y, uint8_t* is_intra,

  1482.                                int pred_flag, int dir)

  1483. {

  1484.     MpegEncContext *s = &v->s;

  1485.     int xy, wrap, off = 0;

  1486.     int16_t *A, *B, *C;

  1487.     int px, py;

  1488.     int sum;

  1489.     int mixedmv_pic, num_samefield = 0, num_oppfield = 0;

  1490.     int opposit, a_f, b_f, c_f;

  1491.     int16_t field_predA[2];

  1492.     int16_t field_predB[2];

  1493.     int16_t field_predC[2];

  1494.     int a_valid, b_valid, c_valid;

  1495.     int hybridmv_thresh, y_bias = 0;

  1496. 

  1497.     if (v->mv_mode == MV_PMODE_MIXED_MV ||

  1498.         ((v->mv_mode == MV_PMODE_INTENSITY_COMP) && (v->mv_mode2 == MV_PMODE_MIXED_MV)))

  1499.         mixedmv_pic = 1;

  1500.     else

  1501.         mixedmv_pic = 0;

  1502.     /* scale MV difference to be quad-pel */

  1503.     dmv_x <<= 1 - s->quarter_sample;

  1504.     dmv_y <<= 1 - s->quarter_sample;

  1505. 

  1506.     wrap = s->b8_stride;

  1507.     xy   = s->block_index[n];

  1508. 

  1509.     if (s->mb_intra) {

  1510.         s->mv[0][n][0] = s->current_picture.f.motion_val[0][xy + v->blocks_off][0] = 0;

  1511.         s->mv[0][n][1] = s->current_picture.f.motion_val[0][xy + v->blocks_off][1] = 0;

  1512.         s->current_picture.f.motion_val[1][xy + v->blocks_off][0] = 0;

  1513.         s->current_picture.f.motion_val[1][xy + v->blocks_off][1] = 0;

  1514.         if (mv1) { /* duplicate motion data for 1-MV block */

  1515.             s->current_picture.f.motion_val[0][xy + 1 + v->blocks_off][0]        = 0;

  1516.             s->current_picture.f.motion_val[0][xy + 1 + v->blocks_off][1]        = 0;

  1517.             s->current_picture.f.motion_val[0][xy + wrap + v->blocks_off][0]     = 0;

  1518.             s->current_picture.f.motion_val[0][xy + wrap + v->blocks_off][1]     = 0;

  1519.             s->current_picture.f.motion_val[0][xy + wrap + 1 + v->blocks_off][0] = 0;

  1520.             s->current_picture.f.motion_val[0][xy + wrap + 1 + v->blocks_off][1] = 0;

  1521.             v->luma_mv[s->mb_x][0] = v->luma_mv[s->mb_x][1] = 0;

  1522.             s->current_picture.f.motion_val[1][xy + 1 + v->blocks_off][0]        = 0;

  1523.             s->current_picture.f.motion_val[1][xy + 1 + v->blocks_off][1]        = 0;

  1524.             s->current_picture.f.motion_val[1][xy + wrap][0]                     = 0;

  1525.             s->current_picture.f.motion_val[1][xy + wrap + v->blocks_off][1]     = 0;

  1526.             s->current_picture.f.motion_val[1][xy + wrap + 1 + v->blocks_off][0] = 0;

  1527.             s->current_picture.f.motion_val[1][xy + wrap + 1 + v->blocks_off][1] = 0;

  1528.         }

  1529.         return;

  1530.     }

  1531. 

  1532.     C = s->current_picture.f.motion_val[dir][xy -    1 + v->blocks_off];

  1533.     A = s->current_picture.f.motion_val[dir][xy - wrap + v->blocks_off];

  1534.     if (mv1) {

  1535.         if (v->field_mode && mixedmv_pic)

  1536.             off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;

  1537.         else

  1538.             off = (s->mb_x == (s->mb_width - 1)) ? -1 : 2;

  1539.     } else {

  1540.         //in 4-MV mode different blocks have different B predictor position

  1541.         switch (n) {

  1542.         case 0:

  1543.             off = (s->mb_x > 0) ? -1 : 1;

  1544.             break;

  1545.         case 1:

  1546.             off = (s->mb_x == (s->mb_width - 1)) ? -1 : 1;

  1547.             break;

  1548.         case 2:

  1549.             off = 1;

  1550.             break;

  1551.         case 3:

  1552.             off = -1;

  1553.         }

  1554.     }

  1555.     B = s->current_picture.f.motion_val[dir][xy - wrap + off + v->blocks_off];

  1556. 

  1557.     a_valid = !s->first_slice_line || (n == 2 || n == 3);

  1558.     b_valid = a_valid && (s->mb_width > 1);

  1559.     c_valid = s->mb_x || (n == 1 || n == 3);

  1560.     if (v->field_mode) {

  1561.         a_valid = a_valid && !is_intra[xy - wrap];

  1562.         b_valid = b_valid && !is_intra[xy - wrap + off];

  1563.         c_valid = c_valid && !is_intra[xy - 1];

  1564.     }

  1565. 

  1566.     if (a_valid) {

  1567.         a_f = v->mv_f[dir][xy - wrap + v->blocks_off];

  1568.         num_oppfield  += a_f;

  1569.         num_samefield += 1 - a_f;

  1570.         field_predA[0] = A[0];

  1571.         field_predA[1] = A[1];

  1572.     } else {

  1573.         field_predA[0] = field_predA[1] = 0;

  1574.         a_f = 0;

  1575.     }

  1576.     if (b_valid) {

  1577.         b_f = v->mv_f[dir][xy - wrap + off + v->blocks_off];

  1578.         num_oppfield  += b_f;

  1579.         num_samefield += 1 - b_f;

  1580.         field_predB[0] = B[0];

  1581.         field_predB[1] = B[1];

  1582.     } else {

  1583.         field_predB[0] = field_predB[1] = 0;

  1584.         b_f = 0;

  1585.     }

  1586.     if (c_valid) {

  1587.         c_f = v->mv_f[dir][xy - 1 + v->blocks_off];

  1588.         num_oppfield  += c_f;

  1589.         num_samefield += 1 - c_f;

  1590.         field_predC[0] = C[0];

  1591.         field_predC[1] = C[1];

  1592.     } else {

  1593.         field_predC[0] = field_predC[1] = 0;

  1594.         c_f = 0;

  1595.     }

  1596. 

  1597.     if (v->field_mode) {

  1598.         if (num_samefield <= num_oppfield)

  1599.             opposit = 1 - pred_flag;

  1600.         else

  1601.             opposit = pred_flag;

  1602.     } else

  1603.         opposit = 0;

  1604.     if (opposit) {

  1605.         if (a_valid && !a_f) {

  1606.             field_predA[0] = scaleforopp(v, field_predA[0], 0, dir);

  1607.             field_predA[1] = scaleforopp(v, field_predA[1], 1, dir);

  1608.         }

  1609.         if (b_valid && !b_f) {

  1610.             field_predB[0] = scaleforopp(v, field_predB[0], 0, dir);

  1611.             field_predB[1] = scaleforopp(v, field_predB[1], 1, dir);

  1612.         }

  1613.         if (c_valid && !c_f) {

  1614.             field_predC[0] = scaleforopp(v, field_predC[0], 0, dir);

  1615.             field_predC[1] = scaleforopp(v, field_predC[1], 1, dir);

  1616.         }

  1617.         v->mv_f[dir][xy + v->blocks_off] = 1;

  1618.         v->ref_field_type[dir] = !v->cur_field_type;

  1619.     } else {

  1620.         if (a_valid && a_f) {

  1621.             field_predA[0] = scaleforsame(v, n, field_predA[0], 0, dir);

  1622.             field_predA[1] = scaleforsame(v, n, field_predA[1], 1, dir);

  1623.         }

  1624.         if (b_valid && b_f) {

  1625.             field_predB[0] = scaleforsame(v, n, field_predB[0], 0, dir);

  1626.             field_predB[1] = scaleforsame(v, n, field_predB[1], 1, dir);

  1627.         }

  1628.         if (c_valid && c_f) {

  1629.             field_predC[0] = scaleforsame(v, n, field_predC[0], 0, dir);

  1630.             field_predC[1] = scaleforsame(v, n, field_predC[1], 1, dir);

  1631.         }

  1632.         v->mv_f[dir][xy + v->blocks_off] = 0;

  1633.         v->ref_field_type[dir] = v->cur_field_type;

  1634.     }

  1635. 

  1636.     if (a_valid) {

  1637.         px = field_predA[0];

  1638.         py = field_predA[1];

  1639.     } else if (c_valid) {

  1640.         px = field_predC[0];

  1641.         py = field_predC[1];

  1642.     } else if (b_valid) {

  1643.         px = field_predB[0];

  1644.         py = field_predB[1];

  1645.     } else {

  1646.         px = 0;

  1647.         py = 0;

  1648.     }

  1649. 

  1650.     if (num_samefield + num_oppfield > 1) {

  1651.         px = mid_pred(field_predA[0], field_predB[0], field_predC[0]);

  1652.         py = mid_pred(field_predA[1], field_predB[1], field_predC[1]);

  1653.     }

  1654. 

  1655.     /* Pullback MV as specified in 8.3.5.3.4 */

  1656.     if (!v->field_mode) {

  1657.         int qx, qy, X, Y;

  1658.         qx = (s->mb_x << 6) + ((n == 1 || n == 3) ? 32 : 0);

  1659.         qy = (s->mb_y << 6) + ((n == 2 || n == 3) ? 32 : 0);

  1660.         X  = (s->mb_width  << 6) - 4;

  1661.         Y  = (s->mb_height << 6) - 4;

  1662.         if (mv1) {

  1663.             if (qx + px < -60) px = -60 - qx;

  1664.             if (qy + py < -60) py = -60 - qy;

  1665.         } else {

  1666.             if (qx + px < -28) px = -28 - qx;

  1667.             if (qy + py < -28) py = -28 - qy;

  1668.         }

  1669.         if (qx + px > X) px = X - qx;

  1670.         if (qy + py > Y) py = Y - qy;

  1671.     }

  1672. 

  1673.     if (!v->field_mode || s->pict_type != AV_PICTURE_TYPE_B) {

  1674.         /* Calculate hybrid prediction as specified in 8.3.5.3.5 (also 10.3.5.4.3.5) */

  1675.         hybridmv_thresh = 32;

  1676.         if (a_valid && c_valid) {

  1677.             if (is_intra[xy - wrap])

  1678.                 sum = FFABS(px) + FFABS(py);

  1679.             else

  1680.                 sum = FFABS(px - field_predA[0]) + FFABS(py - field_predA[1]);

  1681.             if (sum > hybridmv_thresh) {

  1682.                 if (get_bits1(&s->gb)) {     // read HYBRIDPRED bit

  1683.                     px = field_predA[0];

  1684.                     py = field_predA[1];

  1685.                 } else {

  1686.                     px = field_predC[0];

  1687.                     py = field_predC[1];

  1688.                 }

  1689.             } else {

  1690.                 if (is_intra[xy - 1])

  1691.                     sum = FFABS(px) + FFABS(py);

  1692.                 else

  1693.                     sum = FFABS(px - field_predC[0]) + FFABS(py - field_predC[1]);

  1694.                 if (sum > hybridmv_thresh) {

  1695.                     if (get_bits1(&s->gb)) {

  1696.                         px = field_predA[0];

  1697.                         py = field_predA[1];

  1698.                     } else {

  1699.                         px = field_predC[0];

  1700.                         py = field_predC[1];

  1701.                     }

  1702.                 }

  1703.             }

  1704.         }

  1705.     }

  1706. 

  1707.     if (v->field_mode && !s->quarter_sample) {

  1708.         r_x <<= 1;

  1709.         r_y <<= 1;

  1710.     }

  1711.     if (v->field_mode && v->numref)

  1712.         r_y >>= 1;

  1713.     if (v->field_mode && v->cur_field_type && v->ref_field_type[dir] == 0)

  1714.         y_bias = 1;

  1715.     /* store MV using signed modulus of MV range defined in 4.11 */

  1716.     s->mv[dir][n][0] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x;

  1717.     s->mv[dir][n][1] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][1] = ((py + dmv_y + r_y - y_bias) & ((r_y << 1) - 1)) - r_y + y_bias;

  1718.     if (mv1) { /* duplicate motion data for 1-MV block */

  1719.         s->current_picture.f.motion_val[dir][xy +    1 +     v->blocks_off][0] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][0];

  1720.         s->current_picture.f.motion_val[dir][xy +    1 +     v->blocks_off][1] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][1];

  1721.         s->current_picture.f.motion_val[dir][xy + wrap +     v->blocks_off][0] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][0];

  1722.         s->current_picture.f.motion_val[dir][xy + wrap +     v->blocks_off][1] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][1];

  1723.         s->current_picture.f.motion_val[dir][xy + wrap + 1 + v->blocks_off][0] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][0];

  1724.         s->current_picture.f.motion_val[dir][xy + wrap + 1 + v->blocks_off][1] = s->current_picture.f.motion_val[dir][xy + v->blocks_off][1];

  1725.         v->mv_f[dir][xy +    1 + v->blocks_off] = v->mv_f[dir][xy +            v->blocks_off];

  1726.         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];

  1727.     }

  1728. }

  1729. 

  1730. /** Predict and set motion vector for interlaced frame picture MBs

  1731.  */

  1732. static inline void vc1_pred_mv_intfr(VC1Context *v, int n, int dmv_x, int dmv_y,

  1733.                                      int mvn, int r_x, int r_y, uint8_t* is_intra)

  1734. {

  1735.     MpegEncContext *s = &v->s;

  1736.     int xy, wrap, off = 0;

  1737.     int A[2], B[2], C[2];

  1738.     int px, py;

  1739.     int a_valid = 0, b_valid = 0, c_valid = 0;

  1740.     int field_a, field_b, field_c; // 0: same, 1: opposit

  1741.     int total_valid, num_samefield, num_oppfield;

  1742.     int pos_c, pos_b, n_adj;

  1743. 

  1744.     wrap = s->b8_stride;

  1745.     xy = s->block_index[n];

  1746. 

  1747.     if (s->mb_intra) {

  1748.         s->mv[0][n][0] = s->current_picture.f.motion_val[0][xy][0] = 0;

  1749.         s->mv[0][n][1] = s->current_picture.f.motion_val[0][xy][1] = 0;

  1750.         s->current_picture.f.motion_val[1][xy][0] = 0;

  1751.         s->current_picture.f.motion_val[1][xy][1] = 0;

  1752.         if (mvn == 1) { /* duplicate motion data for 1-MV block */

  1753.             s->current_picture.f.motion_val[0][xy + 1][0]        = 0;

  1754.             s->current_picture.f.motion_val[0][xy + 1][1]        = 0;

  1755.             s->current_picture.f.motion_val[0][xy + wrap][0]     = 0;

  1756.             s->current_picture.f.motion_val[0][xy + wrap][1]     = 0;

  1757.             s->current_picture.f.motion_val[0][xy + wrap + 1][0] = 0;

  1758.             s->current_picture.f.motion_val[0][xy + wrap + 1][1] = 0;

  1759.             v->luma_mv[s->mb_x][0] = v->luma_mv[s->mb_x][1] = 0;

  1760.             s->current_picture.f.motion_val[1][xy + 1][0]        = 0;

  1761.             s->current_picture.f.motion_val[1][xy + 1][1]        = 0;

  1762.             s->current_picture.f.motion_val[1][xy + wrap][0]     = 0;

  1763.             s->current_picture.f.motion_val[1][xy + wrap][1]     = 0;

  1764.             s->current_picture.f.motion_val[1][xy + wrap + 1][0] = 0;

  1765.             s->current_picture.f.motion_val[1][xy + wrap + 1][1] = 0;

  1766.         }

  1767.         return;

  1768.     }

  1769. 

  1770.     off = ((n == 0) || (n == 1)) ? 1 : -1;

  1771.     /* predict A */

  1772.     if (s->mb_x || (n == 1) || (n == 3)) {

  1773.         if ((v->blk_mv_type[xy]) // current block (MB) has a field MV

  1774.             || (!v->blk_mv_type[xy] && !v->blk_mv_type[xy - 1])) { // or both have frame MV

  1775.             A[0] = s->current_picture.f.motion_val[0][xy - 1][0];

  1776.             A[1] = s->current_picture.f.motion_val[0][xy - 1][1];

  1777.             a_valid = 1;

  1778.         } else { // current block has frame mv and cand. has field MV (so average)

  1779.             A[0] = (s->current_picture.f.motion_val[0][xy - 1][0]

  1780.                     + s->current_picture.f.motion_val[0][xy - 1 + off * wrap][0] + 1) >> 1;

  1781.             A[1] = (s->current_picture.f.motion_val[0][xy - 1][1]

  1782.                     + s->current_picture.f.motion_val[0][xy - 1 + off * wrap][1] + 1) >> 1;

  1783.             a_valid = 1;

  1784.         }

  1785.         if (!(n & 1) && v->is_intra[s->mb_x - 1]) {

  1786.             a_valid = 0;

  1787.             A[0] = A[1] = 0;

  1788.         }

  1789.     } else

  1790.         A[0] = A[1] = 0;

  1791.     /* Predict B and C */

  1792.     B[0] = B[1] = C[0] = C[1] = 0;

  1793.     if (n == 0 || n == 1 || v->blk_mv_type[xy]) {

  1794.         if (!s->first_slice_line) {

  1795.             if (!v->is_intra[s->mb_x - s->mb_stride]) {

  1796.                 b_valid = 1;

  1797.                 n_adj   = n | 2;

  1798.                 pos_b   = s->block_index[n_adj] - 2 * wrap;

  1799.                 if (v->blk_mv_type[pos_b] && v->blk_mv_type[xy]) {

  1800.                     n_adj = (n & 2) | (n & 1);

  1801.                 }

  1802.                 B[0] = s->current_picture.f.motion_val[0][s->block_index[n_adj] - 2 * wrap][0];

  1803.                 B[1] = s->current_picture.f.motion_val[0][s->block_index[n_adj] - 2 * wrap][1];

  1804.                 if (v->blk_mv_type[pos_b] && !v->blk_mv_type[xy]) {

  1805.                     B[0] = (B[0] + s->current_picture.f.motion_val[0][s->block_index[n_adj ^ 2] - 2 * wrap][0] + 1) >> 1;

  1806.                     B[1] = (B[1] + s->current_picture.f.motion_val[0][s->block_index[n_adj ^ 2] - 2 * wrap][1] + 1) >> 1;

  1807.                 }

  1808.             }

  1809.             if (s->mb_width > 1) {

  1810.                 if (!v->is_intra[s->mb_x - s->mb_stride + 1]) {

  1811.                     c_valid = 1;

  1812.                     n_adj   = 2;

  1813.                     pos_c   = s->block_index[2] - 2 * wrap + 2;

  1814.                     if (v->blk_mv_type[pos_c] && v->blk_mv_type[xy]) {

  1815.                         n_adj = n & 2;

  1816.                     }

  1817.                     C[0] = s->current_picture.f.motion_val[0][s->block_index[n_adj] - 2 * wrap + 2][0];

  1818.                     C[1] = s->current_picture.f.motion_val[0][s->block_index[n_adj] - 2 * wrap + 2][1];

  1819.                     if (v->blk_mv_type[pos_c] && !v->blk_mv_type[xy]) {

  1820.                         C[0] = (1 + C[0] + (s->current_picture.f.motion_val[0][s->block_index[n_adj ^ 2] - 2 * wrap + 2][0])) >> 1;

  1821.                         C[1] = (1 + C[1] + (s->current_picture.f.motion_val[0][s->block_index[n_adj ^ 2] - 2 * wrap + 2][1])) >> 1;

  1822.                     }

  1823.                     if (s->mb_x == s->mb_width - 1) {

  1824.                         if (!v->is_intra[s->mb_x - s->mb_stride - 1]) {

  1825.                             c_valid = 1;

  1826.                             n_adj   = 3;

  1827.                             pos_c   = s->block_index[3] - 2 * wrap - 2;

  1828.                             if (v->blk_mv_type[pos_c] && v->blk_mv_type[xy]) {

  1829.                                 n_adj = n | 1;

  1830.                             }

  1831.                             C[0] = s->current_picture.f.motion_val[0][s->block_index[n_adj] - 2 * wrap - 2][0];

  1832.                             C[1] = s->current_picture.f.motion_val[0][s->block_index[n_adj] - 2 * wrap - 2][1];

  1833.                             if (v->blk_mv_type[pos_c] && !v->blk_mv_type[xy]) {

  1834.                                 C[0] = (1 + C[0] + s->current_picture.f.motion_val[0][s->block_index[1] - 2 * wrap - 2][0]) >> 1;

  1835.                                 C[1] = (1 + C[1] + s->current_picture.f.motion_val[0][s->block_index[1] - 2 * wrap - 2][1]) >> 1;

  1836.                             }

  1837.                         } else

  1838.                             c_valid = 0;

  1839.                     }

  1840.                 }

  1841.             }

  1842.         }

  1843.     } else {

  1844.         pos_b   = s->block_index[1];

  1845.         b_valid = 1;

  1846.         B[0]    = s->current_picture.f.motion_val[0][pos_b][0];

  1847.         B[1]    = s->current_picture.f.motion_val[0][pos_b][1];

  1848.         pos_c   = s->block_index[0];

  1849.         c_valid = 1;

  1850.         C[0]    = s->current_picture.f.motion_val[0][pos_c][0];

  1851.         C[1]    = s->current_picture.f.motion_val[0][pos_c][1];

  1852.     }

  1853. 

  1854.     total_valid = a_valid + b_valid + c_valid;

  1855.     // check if predictor A is out of bounds

  1856.     if (!s->mb_x && !(n == 1 || n == 3)) {

  1857.         A[0] = A[1] = 0;

  1858.     }

  1859.     // check if predictor B is out of bounds

  1860.     if ((s->first_slice_line && v->blk_mv_type[xy]) || (s->first_slice_line && !(n & 2))) {

  1861.         B[0] = B[1] = C[0] = C[1] = 0;

  1862.     }

  1863.     if (!v->blk_mv_type[xy]) {

  1864.         if (s->mb_width == 1) {

  1865.             px = B[0];

  1866.             py = B[1];

  1867.         } else {

  1868.             if (total_valid >= 2) {

  1869.                 px = mid_pred(A[0], B[0], C[0]);

  1870.                 py = mid_pred(A[1], B[1], C[1]);

  1871.             } else if (total_valid) {

  1872.                 if (a_valid) { px = A[0]; py = A[1]; }

  1873.                 if (b_valid) { px = B[0]; py = B[1]; }

  1874.                 if (c_valid) { px = C[0]; py = C[1]; }

  1875.             } else

  1876.                 px = py = 0;

  1877.         }

  1878.     } else {

  1879.         if (a_valid)

  1880.             field_a = (A[1] & 4) ? 1 : 0;

  1881.         else

  1882.             field_a = 0;

  1883.         if (b_valid)

  1884.             field_b = (B[1] & 4) ? 1 : 0;

  1885.         else

  1886.             field_b = 0;

  1887.         if (c_valid)

  1888.             field_c = (C[1] & 4) ? 1 : 0;

  1889.         else

  1890.             field_c = 0;

  1891. 

  1892.         num_oppfield  = field_a + field_b + field_c;

  1893.         num_samefield = total_valid - num_oppfield;

  1894.         if (total_valid == 3) {

  1895.             if ((num_samefield == 3) || (num_oppfield == 3)) {

  1896.                 px = mid_pred(A[0], B[0], C[0]);

  1897.                 py = mid_pred(A[1], B[1], C[1]);

  1898.             } else if (num_samefield >= num_oppfield) {

  1899.                 /* take one MV from same field set depending on priority

  1900.                 the check for B may not be necessary */

  1901.                 px = !field_a ? A[0] : B[0];

  1902.                 py = !field_a ? A[1] : B[1];

  1903.             } else {

  1904.                 px =  field_a ? A[0] : B[0];

  1905.                 py =  field_a ? A[1] : B[1];

  1906.             }

  1907.         } else if (total_valid == 2) {

  1908.             if (num_samefield >= num_oppfield) {

  1909.                 if (!field_a && a_valid) {

  1910.                     px = A[0];

  1911.                     py = A[1];

  1912.                 } else if (!field_b && b_valid) {

  1913.                     px = B[0];

  1914.                     py = B[1];

  1915.                 } else if (c_valid) {

  1916.                     px = C[0];

  1917.                     py = C[1];

  1918.                 } else px = py = 0;

  1919.             } else {

  1920.                 if (field_a && a_valid) {

  1921.                     px = A[0];

  1922.                     py = A[1];

  1923.                 } else if (field_b && b_valid) {

  1924.                     px = B[0];

  1925.                     py = B[1];

  1926.                 } else if (c_valid) {

  1927.                     px = C[0];

  1928.                     py = C[1];

  1929.                 }

  1930.             }

  1931.         } else if (total_valid == 1) {

  1932.             px = (a_valid) ? A[0] : ((b_valid) ? B[0] : C[0]);

  1933.             py = (a_valid) ? A[1] : ((b_valid) ? B[1] : C[1]);

  1934.         } else

  1935.             px = py = 0;

  1936.     }

  1937. 

  1938.     /* store MV using signed modulus of MV range defined in 4.11 */

  1939.     s->mv[0][n][0] = s->current_picture.f.motion_val[0][xy][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x;

  1940.     s->mv[0][n][1] = s->current_picture.f.motion_val[0][xy][1] = ((py + dmv_y + r_y) & ((r_y << 1) - 1)) - r_y;

  1941.     if (mvn == 1) { /* duplicate motion data for 1-MV block */

  1942.         s->current_picture.f.motion_val[0][xy +    1    ][0] = s->current_picture.f.motion_val[0][xy][0];

  1943.         s->current_picture.f.motion_val[0][xy +    1    ][1] = s->current_picture.f.motion_val[0][xy][1];

  1944.         s->current_picture.f.motion_val[0][xy + wrap    ][0] = s->current_picture.f.motion_val[0][xy][0];

  1945.         s->current_picture.f.motion_val[0][xy + wrap    ][1] = s->current_picture.f.motion_val[0][xy][1];

  1946.         s->current_picture.f.motion_val[0][xy + wrap + 1][0] = s->current_picture.f.motion_val[0][xy][0];

  1947.         s->current_picture.f.motion_val[0][xy + wrap + 1][1] = s->current_picture.f.motion_val[0][xy][1];

  1948.     } else if (mvn == 2) { /* duplicate motion data for 2-Field MV block */

  1949.         s->current_picture.f.motion_val[0][xy + 1][0] = s->current_picture.f.motion_val[0][xy][0];

  1950.         s->current_picture.f.motion_val[0][xy + 1][1] = s->current_picture.f.motion_val[0][xy][1];

  1951.         s->mv[0][n + 1][0] = s->mv[0][n][0];

  1952.         s->mv[0][n + 1][1] = s->mv[0][n][1];

  1953.     }

  1954. }

  1955. 

  1956. /** Motion compensation for direct or interpolated blocks in B-frames

  1957.  */

  1958. static void vc1_interp_mc(VC1Context *v)

  1959. {

  1960.     MpegEncContext *s = &v->s;

  1961.     DSPContext *dsp = &v->s.dsp;

  1962.     uint8_t *srcY, *srcU, *srcV;

  1963.     int dxy, mx, my, uvmx, uvmy, src_x, src_y, uvsrc_x, uvsrc_y;

  1964.     int off, off_uv;

  1965.     int v_edge_pos = s->v_edge_pos >> v->field_mode;

  1966. 

  1967.     if (!v->field_mode && !v->s.next_picture.f.data[0])

  1968.         return;

  1969. 

  1970.     mx   = s->mv[1][0][0];

  1971.     my   = s->mv[1][0][1];

  1972.     uvmx = (mx + ((mx & 3) == 3)) >> 1;

  1973.     uvmy = (my + ((my & 3) == 3)) >> 1;

  1974.     if (v->field_mode) {

  1975.         if (v->cur_field_type != v->ref_field_type[1])

  1976.             my   = my   - 2 + 4 * v->cur_field_type;

  1977.             uvmy = uvmy - 2 + 4 * v->cur_field_type;

  1978.     }

  1979.     if (v->fastuvmc) {

  1980.         uvmx = uvmx + ((uvmx < 0) ? -(uvmx & 1) : (uvmx & 1));

  1981.         uvmy = uvmy + ((uvmy < 0) ? -(uvmy & 1) : (uvmy & 1));

  1982.     }

  1983.     srcY = s->next_picture.f.data[0];

  1984.     srcU = s->next_picture.f.data[1];

  1985.     srcV = s->next_picture.f.data[2];

  1986. 

  1987.     src_x   = s->mb_x * 16 + (mx   >> 2);

  1988.     src_y   = s->mb_y * 16 + (my   >> 2);

  1989.     uvsrc_x = s->mb_x *  8 + (uvmx >> 2);

  1990.     uvsrc_y = s->mb_y *  8 + (uvmy >> 2);

  1991. 

  1992.     if (v->profile != PROFILE_ADVANCED) {

  1993.         src_x   = av_clip(  src_x, -16, s->mb_width  * 16);

  1994.         src_y   = av_clip(  src_y, -16, s->mb_height * 16);

  1995.         uvsrc_x = av_clip(uvsrc_x,  -8, s->mb_width  *  8);

  1996.         uvsrc_y = av_clip(uvsrc_y,  -8, s->mb_height *  8);

  1997.     } else {

  1998.         src_x   = av_clip(  src_x, -17, s->avctx->coded_width);

  1999.         src_y   = av_clip(  src_y, -18, s->avctx->coded_height + 1);

  2000.         uvsrc_x = av_clip(uvsrc_x,  -8, s->avctx->coded_width  >> 1);

  2001.         uvsrc_y = av_clip(uvsrc_y,  -8, s->avctx->coded_height >> 1);

  2002.     }

  2003. 

  2004.     srcY += src_y   * s->linesize   + src_x;

  2005.     srcU += uvsrc_y * s->uvlinesize + uvsrc_x;

  2006.     srcV += uvsrc_y * s->uvlinesize + uvsrc_x;

  2007. 

  2008.     if (v->field_mode && v->ref_field_type[1]) {

  2009.         srcY += s->current_picture_ptr->f.linesize[0];

  2010.         srcU += s->current_picture_ptr->f.linesize[1];

  2011.         srcV += s->current_picture_ptr->f.linesize[2];

  2012.     }

  2013. 

  2014.     /* for grayscale we should not try to read from unknown area */

  2015.     if (s->flags & CODEC_FLAG_GRAY) {

  2016.         srcU = s->edge_emu_buffer + 18 * s->linesize;

  2017.         srcV = s->edge_emu_buffer + 18 * s->linesize;

  2018.     }

  2019. 

  2020.     if (v->rangeredfrm || s->h_edge_pos < 22 || v_edge_pos < 22

  2021.         || (unsigned)(src_x - s->mspel) > s->h_edge_pos - (mx & 3) - 16 - s->mspel * 3

  2022.         || (unsigned)(src_y - s->mspel) > v_edge_pos    - (my & 3) - 16 - s->mspel * 3) {

  2023.         uint8_t *uvbuf = s->edge_emu_buffer + 19 * s->linesize;

  2024. 

  2025.         srcY -= s->mspel * (1 + s->linesize);

  2026.         s->dsp.emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize,

  2027.                                 17 + s->mspel * 2, 17 + s->mspel * 2,

  2028.                                 src_x - s->mspel, src_y - s->mspel,

  2029.                                 s->h_edge_pos, v_edge_pos);

  2030.         srcY = s->edge_emu_buffer;

  2031.         s->dsp.emulated_edge_mc(uvbuf     , srcU, s->uvlinesize, 8 + 1, 8 + 1,

  2032.                                 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1);

  2033.         s->dsp.emulated_edge_mc(uvbuf + 16, srcV, s->uvlinesize, 8 + 1, 8 + 1,

  2034.                                 uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, v_edge_pos >> 1);

  2035.         srcU = uvbuf;

  2036.         srcV = uvbuf + 16;

  2037.         /* if we deal with range reduction we need to scale source blocks */

  2038.         if (v->rangeredfrm) {

  2039.             int i, j;

  2040.             uint8_t *src, *src2;

  2041. 

  2042.             src = srcY;

  2043.             for (j = 0; j < 17 + s->mspel * 2; j++) {

  2044.                 for (i = 0; i < 17 + s->mspel * 2; i++)

  2045.                     src[i] = ((src[i] - 128) >> 1) + 128;

  2046.                 src += s->linesize;

  2047.             }

  2048.             src = srcU;

  2049.             src2 = srcV;

  2050.             for (j = 0; j < 9; j++) {

  2051.                 for (i = 0; i < 9; i++) {

  2052.                     src[i]  = ((src[i]  - 128) >> 1) + 128;

  2053.                     src2[i] = ((src2[i] - 128) >> 1) + 128;

  2054.                 }

  2055.                 src  += s->uvlinesize;

  2056.                 src2 += s->uvlinesize;

  2057.             }

  2058.         }

  2059.         srcY += s->mspel * (1 + s->linesize);

  2060.     }

  2061. 

  2062.     if (v->field_mode && v->second_field) {

  2063.         off    = s->current_picture_ptr->f.linesize[0];

  2064.         off_uv = s->current_picture_ptr->f.linesize[1];

  2065.     } else {

  2066.         off    = 0;

  2067.         off_uv = 0;

  2068.     }

  2069. 

  2070.     if (s->mspel) {

  2071.         dxy = ((my & 3) << 2) | (mx & 3);

  2072.         v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off    , srcY    , s->linesize, v->rnd);

  2073.         v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8, srcY + 8, s->linesize, v->rnd);

  2074.         srcY += s->linesize * 8;

  2075.         v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8 * s->linesize    , srcY    , s->linesize, v->rnd);

  2076.         v->vc1dsp.avg_vc1_mspel_pixels_tab[dxy](s->dest[0] + off + 8 * s->linesize + 8, srcY + 8, s->linesize, v->rnd);

  2077.     } else { // hpel mc

  2078.         dxy = (my & 2) | ((mx & 2) >> 1);

  2079. 

  2080.         if (!v->rnd)

  2081.             dsp->avg_pixels_tab[0][dxy](s->dest[0] + off, srcY, s->linesize, 16);

  2082.         else

  2083.             dsp->avg_no_rnd_pixels_tab[0][dxy](s->dest[0] + off, srcY, s->linesize, 16);

  2084.     }

  2085. 

  2086.     if (s->flags & CODEC_FLAG_GRAY) return;

  2087.     /* Chroma MC always uses qpel blilinear */

  2088.     uvmx = (uvmx & 3) << 1;

  2089.     uvmy = (uvmy & 3) << 1;

  2090.     if (!v->rnd) {

  2091.         dsp->avg_h264_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy);

  2092.         dsp->avg_h264_chroma_pixels_tab[0](s->dest[2] + off_uv, srcV, s->uvlinesize, 8, uvmx, uvmy);

  2093.     } else {

  2094.         v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[0](s->dest[1] + off_uv, srcU, s->uvlinesize, 8, uvmx, uvmy);

  2095.         v->vc1dsp.avg_no_rnd_vc1_chroma_pixels_tab[0](s->dest[2] + off_uv, srcV, s->uvlinesize, 8, uvmx, uvmy);

  2096.     }

  2097. }

  2098. 

  2099. static av_always_inline int scale_mv(int value, int bfrac, int inv, int qs)

  2100. {

  2101.     int n = bfrac;

  2102. 

  2103. #if B_FRACTION_DEN==256

  2104.     if (inv)

  2105.         n -= 256;

  2106.     if (!qs)

  2107.         return 2 * ((value * n + 255) >> 9);

  2108.     return (value * n + 128) >> 8;

  2109. #else

  2110.     if (inv)

  2111.         n -= B_FRACTION_DEN;

  2112.     if (!qs)

  2113.         return 2 * ((value * n + B_FRACTION_DEN - 1) / (2 * B_FRACTION_DEN));

  2114.     return (value * n + B_FRACTION_DEN/2) / B_FRACTION_DEN;

  2115. #endif

  2116. }

  2117. 

  2118. static av_always_inline int scale_mv_intfi(int value, int bfrac, int inv,

  2119.                                            int qs, int qs_last)

  2120. {

  2121.     int n = bfrac;

  2122. 

  2123.     if (inv)

  2124.         n -= 256;

  2125.     n <<= !qs_last;

  2126.     if (!qs)

  2127.         return (value * n + 255) >> 9;

  2128.     else

  2129.         return (value * n + 128) >> 8;

  2130. }

  2131. 

  2132. /** Reconstruct motion vector for B-frame and do motion compensation

  2133.  */

  2134. static inline void vc1_b_mc(VC1Context *v, int dmv_x[2], int dmv_y[2],

  2135.                             int direct, int mode)

  2136. {

  2137.     if (v->use_ic) {

  2138.         v->mv_mode2 = v->mv_mode;

  2139.         v->mv_mode  = MV_PMODE_INTENSITY_COMP;

  2140.     }

  2141.     if (direct) {

  2142.         vc1_mc_1mv(v, 0);

  2143.         vc1_interp_mc(v);

  2144.         if (v->use_ic)

  2145.             v->mv_mode = v->mv_mode2;

  2146.         return;

  2147.     }

  2148.     if (mode == BMV_TYPE_INTERPOLATED) {

  2149.         vc1_mc_1mv(v, 0);

  2150.         vc1_interp_mc(v);

  2151.         if (v->use_ic)

  2152.             v->mv_mode = v->mv_mode2;

  2153.         return;

  2154.     }

  2155. 

  2156.     if (v->use_ic && (mode == BMV_TYPE_BACKWARD))

  2157.         v->mv_mode = v->mv_mode2;

  2158.     vc1_mc_1mv(v, (mode == BMV_TYPE_BACKWARD));

  2159.     if (v->use_ic)

  2160.         v->mv_mode = v->mv_mode2;

  2161. }

  2162. 

  2163. static inline void vc1_pred_b_mv(VC1Context *v, int dmv_x[2], int dmv_y[2],

  2164.                                  int direct, int mvtype)

  2165. {

  2166.     MpegEncContext *s = &v->s;

  2167.     int xy, wrap, off = 0;

  2168.     int16_t *A, *B, *C;

  2169.     int px, py;

  2170.     int sum;

  2171.     int r_x, r_y;

  2172.     const uint8_t *is_intra = v->mb_type[0];

  2173. 

  2174.     r_x = v->range_x;

  2175.     r_y = v->range_y;

  2176.     /* scale MV difference to be quad-pel */

  2177.     dmv_x[0] <<= 1 - s->quarter_sample;

  2178.     dmv_y[0] <<= 1 - s->quarter_sample;

  2179.     dmv_x[1] <<= 1 - s->quarter_sample;

  2180.     dmv_y[1] <<= 1 - s->quarter_sample;

  2181. 

  2182.     wrap = s->b8_stride;

  2183.     xy = s->block_index[0];

  2184. 

  2185.     if (s->mb_intra) {

  2186.         s->current_picture.f.motion_val[0][xy + v->blocks_off][0] =

  2187.         s->current_picture.f.motion_val[0][xy + v->blocks_off][1] =

  2188.         s->current_picture.f.motion_val[1][xy + v->blocks_off][0] =

  2189.         s->current_picture.f.motion_val[1][xy + v->blocks_off][1] = 0;

  2190.         return;

  2191.     }

  2192.     if (!v->field_mode) {

  2193.         s->mv[0][0][0] = scale_mv(s->next_picture.f.motion_val[1][xy][0], v->bfraction, 0, s->quarter_sample);

  2194.         s->mv[0][0][1] = scale_mv(s->next_picture.f.motion_val[1][xy][1], v->bfraction, 0, s->quarter_sample);

  2195.         s->mv[1][0][0] = scale_mv(s->next_picture.f.motion_val[1][xy][0], v->bfraction, 1, s->quarter_sample);

  2196.         s->mv[1][0][1] = scale_mv(s->next_picture.f.motion_val[1][xy][1], v->bfraction, 1, s->quarter_sample);

  2197. 

  2198.         /* Pullback predicted motion vectors as specified in 8.4.5.4 */

  2199.         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));

  2200.         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));

  2201.         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));

  2202.         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));

  2203.     }

  2204.     if (direct) {

  2205.         s->current_picture.f.motion_val[0][xy + v->blocks_off][0] = s->mv[0][0][0];

  2206.         s->current_picture.f.motion_val[0][xy + v->blocks_off][1] = s->mv[0][0][1];

  2207.         s->current_picture.f.motion_val[1][xy + v->blocks_off][0] = s->mv[1][0][0];

  2208.         s->current_picture.f.motion_val[1][xy + v->blocks_off][1] = s->mv[1][0][1];

  2209.         return;

  2210.     }

  2211. 

  2212.     if ((mvtype == BMV_TYPE_FORWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {

  2213.         C   = s->current_picture.f.motion_val[0][xy - 2];

  2214.         A   = s->current_picture.f.motion_val[0][xy - wrap * 2];

  2215.         off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;

  2216.         B   = s->current_picture.f.motion_val[0][xy - wrap * 2 + off];

  2217. 

  2218.         if (!s->mb_x) C[0] = C[1] = 0;

  2219.         if (!s->first_slice_line) { // predictor A is not out of bounds

  2220.             if (s->mb_width == 1) {

  2221.                 px = A[0];

  2222.                 py = A[1];

  2223.             } else {

  2224.                 px = mid_pred(A[0], B[0], C[0]);

  2225.                 py = mid_pred(A[1], B[1], C[1]);

  2226.             }

  2227.         } else if (s->mb_x) { // predictor C is not out of bounds

  2228.             px = C[0];

  2229.             py = C[1];

  2230.         } else {

  2231.             px = py = 0;

  2232.         }

  2233.         /* Pullback MV as specified in 8.3.5.3.4 */

  2234.         {

  2235.             int qx, qy, X, Y;

  2236.             if (v->profile < PROFILE_ADVANCED) {

  2237.                 qx = (s->mb_x << 5);

  2238.                 qy = (s->mb_y << 5);

  2239.                 X  = (s->mb_width  << 5) - 4;

  2240.                 Y  = (s->mb_height << 5) - 4;

  2241.                 if (qx + px < -28) px = -28 - qx;

  2242.                 if (qy + py < -28) py = -28 - qy;

  2243.                 if (qx + px > X) px = X - qx;

  2244.                 if (qy + py > Y) py = Y - qy;

  2245.             } else {

  2246.                 qx = (s->mb_x << 6);

  2247.                 qy = (s->mb_y << 6);

  2248.                 X  = (s->mb_width  << 6) - 4;

  2249.                 Y  = (s->mb_height << 6) - 4;

  2250.                 if (qx + px < -60) px = -60 - qx;

  2251.                 if (qy + py < -60) py = -60 - qy;

  2252.                 if (qx + px > X) px = X - qx;

  2253.                 if (qy + py > Y) py = Y - qy;

  2254.             }

  2255.         }

  2256.         /* Calculate hybrid prediction as specified in 8.3.5.3.5 */

  2257.         if (0 && !s->first_slice_line && s->mb_x) {

  2258.             if (is_intra[xy - wrap])

  2259.                 sum = FFABS(px) + FFABS(py);

  2260.             else

  2261.                 sum = FFABS(px - A[0]) + FFABS(py - A[1]);

  2262.             if (sum > 32) {

  2263.                 if (get_bits1(&s->gb)) {

  2264.                     px = A[0];

  2265.                     py = A[1];

  2266.                 } else {

  2267.                     px = C[0];

  2268.                     py = C[1];

  2269.                 }

  2270.             } else {

  2271.                 if (is_intra[xy - 2])

  2272.                     sum = FFABS(px) + FFABS(py);

  2273.                 else

  2274.                     sum = FFABS(px - C[0]) + FFABS(py - C[1]);

  2275.                 if (sum > 32) {

  2276.                     if (get_bits1(&s->gb)) {

  2277.                         px = A[0];

  2278.                         py = A[1];

  2279.                     } else {

  2280.                         px = C[0];

  2281.                         py = C[1];

  2282.                     }

  2283.                 }

  2284.             }

  2285.         }

  2286.         /* store MV using signed modulus of MV range defined in 4.11 */

  2287.         s->mv[0][0][0] = ((px + dmv_x[0] + r_x) & ((r_x << 1) - 1)) - r_x;

  2288.         s->mv[0][0][1] = ((py + dmv_y[0] + r_y) & ((r_y << 1) - 1)) - r_y;

  2289.     }

  2290.     if ((mvtype == BMV_TYPE_BACKWARD) || (mvtype == BMV_TYPE_INTERPOLATED)) {

  2291.         C   = s->current_picture.f.motion_val[1][xy - 2];

  2292.         A   = s->current_picture.f.motion_val[1][xy - wrap * 2];

  2293.         off = (s->mb_x == (s->mb_width - 1)) ? -2 : 2;

  2294.         B   = s->current_picture.f.motion_val[1][xy - wrap * 2 + off];

  2295. 

  2296.         if (!s->mb_x)

  2297.             C[0] = C[1] = 0;

  2298.         if (!s->first_slice_line) { // predictor A is not out of bounds

  2299.             if (s->mb_width == 1) {

  2300.                 px = A[0];

  2301.                 py = A[1];

  2302.             } else {

  2303.                 px = mid_pred(A[0], B[0], C[0]);

  2304.                 py = mid_pred(A[1], B[1], C[1]);

  2305.             }

  2306.         } else if (s->mb_x) { // predictor C is not out of bounds

  2307.             px = C[0];

  2308.             py = C[1];

  2309.         } else {

  2310.             px = py = 0;

  2311.         }

  2312.         /* Pullback MV as specified in 8.3.5.3.4 */

  2313.         {

  2314.             int qx, qy, X, Y;

  2315.             if (v->profile < PROFILE_ADVANCED) {

  2316.                 qx = (s->mb_x << 5);

  2317.                 qy = (s->mb_y << 5);

  2318.                 X  = (s->mb_width  << 5) - 4;

  2319.                 Y  = (s->mb_height << 5) - 4;

  2320.                 if (qx + px < -28) px = -28 - qx;

  2321.                 if (qy + py < -28) py = -28 - qy;

  2322.                 if (qx + px > X) px = X - qx;

  2323.                 if (qy + py > Y) py = Y - qy;

  2324.             } else {

  2325.                 qx = (s->mb_x << 6);

  2326.                 qy = (s->mb_y << 6);

  2327.                 X  = (s->mb_width  << 6) - 4;

  2328.                 Y  = (s->mb_height << 6) - 4;

  2329.                 if (qx + px < -60) px = -60 - qx;

  2330.                 if (qy + py < -60) py = -60 - qy;

  2331.                 if (qx + px > X) px = X - qx;

  2332.                 if (qy + py > Y) py = Y - qy;

  2333.             }

  2334.         }

  2335.         /* Calculate hybrid prediction as specified in 8.3.5.3.5 */

  2336.         if (0 && !s->first_slice_line && s->mb_x) {

  2337.             if (is_intra[xy - wrap])

  2338.                 sum = FFABS(px) + FFABS(py);

  2339.             else

  2340.                 sum = FFABS(px - A[0]) + FFABS(py - A[1]);

  2341.             if (sum > 32) {

  2342.                 if (get_bits1(&s->gb)) {

  2343.                     px = A[0];

  2344.                     py = A[1];

  2345.                 } else {

  2346.                     px = C[0];

  2347.                     py = C[1];

  2348.                 }

  2349.             } else {

  2350.                 if (is_intra[xy - 2])

  2351.                     sum = FFABS(px) + FFABS(py);

  2352.                 else

  2353.                     sum = FFABS(px - C[0]) + FFABS(py - C[1]);

  2354.                 if (sum > 32) {

  2355.                     if (get_bits1(&s->gb)) {

  2356.                         px = A[0];

  2357.                         py = A[1];

  2358.                     } else {

  2359.                         px = C[0];

  2360.                         py = C[1];

  2361.                     }

  2362.                 }

  2363.             }

  2364.         }

  2365.         /* store MV using signed modulus of MV range defined in 4.11 */

  2366. 

  2367.         s->mv[1][0][0] = ((px + dmv_x[1] + r_x) & ((r_x << 1) - 1)) - r_x;

  2368.         s->mv[1][0][1] = ((py + dmv_y[1] + r_y) & ((r_y << 1) - 1)) - r_y;

  2369.     }

  2370.     s->current_picture.f.motion_val[0][xy][0] = s->mv[0][0][0];

  2371.     s->current_picture.f.motion_val[0][xy][1] = s->mv[0][0][1];

  2372.     s->current_picture.f.motion_val[1][xy][0] = s->mv[1][0][0];

  2373.     s->current_picture.f.motion_val[1][xy][1] = s->mv[1][0][1];

  2374. }

  2375. 

  2376. static inline void vc1_pred_b_mv_intfi(VC1Context *v, int n, int *dmv_x, int *dmv_y, int mv1, int *pred_flag)

  2377. {

  2378.     int dir = (v->bmvtype == BMV_TYPE_BACKWARD) ? 1 : 0;

  2379.     MpegEncContext *s = &v->s;

  2380.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  2381. 

  2382.     if (v->bmvtype == BMV_TYPE_DIRECT) {

  2383.         int total_opp, k, f;

  2384.         if (s->next_picture.f.mb_type[mb_pos + v->mb_off] != MB_TYPE_INTRA) {

  2385.             s->mv[0][0][0] = scale_mv_intfi(s->next_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0],

  2386.                                             v->bfraction, 0, s->quarter_sample, v->qs_last);

  2387.             s->mv[0][0][1] = scale_mv_intfi(s->next_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1],

  2388.                                             v->bfraction, 0, s->quarter_sample, v->qs_last);

  2389.             s->mv[1][0][0] = scale_mv_intfi(s->next_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0],

  2390.                                             v->bfraction, 1, s->quarter_sample, v->qs_last);

  2391.             s->mv[1][0][1] = scale_mv_intfi(s->next_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1],

  2392.                                             v->bfraction, 1, s->quarter_sample, v->qs_last);

  2393. 

  2394.             total_opp = v->mv_f_next[0][s->block_index[0] + v->blocks_off]

  2395.                       + v->mv_f_next[0][s->block_index[1] + v->blocks_off]

  2396.                       + v->mv_f_next[0][s->block_index[2] + v->blocks_off]

  2397.                       + v->mv_f_next[0][s->block_index[3] + v->blocks_off];

  2398.             f = (total_opp > 2) ? 1 : 0;

  2399.         } else {

  2400.             s->mv[0][0][0] = s->mv[0][0][1] = 0;

  2401.             s->mv[1][0][0] = s->mv[1][0][1] = 0;

  2402.             f = 0;

  2403.         }

  2404.         v->ref_field_type[0] = v->ref_field_type[1] = v->cur_field_type ^ f;

  2405.         for (k = 0; k < 4; k++) {

  2406.             s->current_picture.f.motion_val[0][s->block_index[k] + v->blocks_off][0] = s->mv[0][0][0];

  2407.             s->current_picture.f.motion_val[0][s->block_index[k] + v->blocks_off][1] = s->mv[0][0][1];

  2408.             s->current_picture.f.motion_val[1][s->block_index[k] + v->blocks_off][0] = s->mv[1][0][0];

  2409.             s->current_picture.f.motion_val[1][s->block_index[k] + v->blocks_off][1] = s->mv[1][0][1];

  2410.             v->mv_f[0][s->block_index[k] + v->blocks_off] = f;

  2411.             v->mv_f[1][s->block_index[k] + v->blocks_off] = f;

  2412.         }

  2413.         return;

  2414.     }

  2415.     if (v->bmvtype == BMV_TYPE_INTERPOLATED) {

  2416.         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);

  2417.         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);

  2418.         return;

  2419.     }

  2420.     if (dir) { // backward

  2421.         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);

  2422.         if (n == 3 || mv1) {

  2423.             vc1_pred_mv(v, 0, dmv_x[0], dmv_y[0],   1, v->range_x, v->range_y, v->mb_type[0], 0, 0);

  2424.         }

  2425.     } else { // forward

  2426.         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);

  2427.         if (n == 3 || mv1) {

  2428.             vc1_pred_mv(v, 0, dmv_x[1], dmv_y[1],   1, v->range_x, v->range_y, v->mb_type[0], 0, 1);

  2429.         }

  2430.     }

  2431. }

  2432. 

  2433. /** Get predicted DC value for I-frames only

  2434.  * prediction dir: left=0, top=1

  2435.  * @param s MpegEncContext

  2436.  * @param overlap flag indicating that overlap filtering is used

  2437.  * @param pq integer part of picture quantizer

  2438.  * @param[in] n block index in the current MB

  2439.  * @param dc_val_ptr Pointer to DC predictor

  2440.  * @param dir_ptr Prediction direction for use in AC prediction

  2441.  */

  2442. static inline int vc1_i_pred_dc(MpegEncContext *s, int overlap, int pq, int n,

  2443.                                 int16_t **dc_val_ptr, int *dir_ptr)

  2444. {

  2445.     int a, b, c, wrap, pred, scale;

  2446.     int16_t *dc_val;

  2447.     static const uint16_t dcpred[32] = {

  2448.         -1, 1024,  512,  341,  256,  205,  171,  146,  128,

  2449.              114,  102,   93,   85,   79,   73,   68,   64,

  2450.               60,   57,   54,   51,   49,   47,   45,   43,

  2451.               41,   39,   38,   37,   35,   34,   33

  2452.     };

  2453. 

  2454.     /* find prediction - wmv3_dc_scale always used here in fact */

  2455.     if (n < 4) scale = s->y_dc_scale;

  2456.     else       scale = s->c_dc_scale;

  2457. 

  2458.     wrap   = s->block_wrap[n];

  2459.     dc_val = s->dc_val[0] + s->block_index[n];

  2460. 

  2461.     /* B A

  2462.      * C X

  2463.      */

  2464.     c = dc_val[ - 1];

  2465.     b = dc_val[ - 1 - wrap];

  2466.     a = dc_val[ - wrap];

  2467. 

  2468.     if (pq < 9 || !overlap) {

  2469.         /* Set outer values */

  2470.         if (s->first_slice_line && (n != 2 && n != 3))

  2471.             b = a = dcpred[scale];

  2472.         if (s->mb_x == 0 && (n != 1 && n != 3))

  2473.             b = c = dcpred[scale];

  2474.     } else {

  2475.         /* Set outer values */

  2476.         if (s->first_slice_line && (n != 2 && n != 3))

  2477.             b = a = 0;

  2478.         if (s->mb_x == 0 && (n != 1 && n != 3))

  2479.             b = c = 0;

  2480.     }

  2481. 

  2482.     if (abs(a - b) <= abs(b - c)) {

  2483.         pred     = c;

  2484.         *dir_ptr = 1; // left

  2485.     } else {

  2486.         pred     = a;

  2487.         *dir_ptr = 0; // top

  2488.     }

  2489. 

  2490.     /* update predictor */

  2491.     *dc_val_ptr = &dc_val[0];

  2492.     return pred;

  2493. }

  2494. 

  2495. 

  2496. /** Get predicted DC value

  2497.  * prediction dir: left=0, top=1

  2498.  * @param s MpegEncContext

  2499.  * @param overlap flag indicating that overlap filtering is used

  2500.  * @param pq integer part of picture quantizer

  2501.  * @param[in] n block index in the current MB

  2502.  * @param a_avail flag indicating top block availability

  2503.  * @param c_avail flag indicating left block availability

  2504.  * @param dc_val_ptr Pointer to DC predictor

  2505.  * @param dir_ptr Prediction direction for use in AC prediction

  2506.  */

  2507. static inline int vc1_pred_dc(MpegEncContext *s, int overlap, int pq, int n,

  2508.                               int a_avail, int c_avail,

  2509.                               int16_t **dc_val_ptr, int *dir_ptr)

  2510. {

  2511.     int a, b, c, wrap, pred;

  2512.     int16_t *dc_val;

  2513.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  2514.     int q1, q2 = 0;

  2515.     int dqscale_index;

  2516. 

  2517.     wrap = s->block_wrap[n];

  2518.     dc_val = s->dc_val[0] + s->block_index[n];

  2519. 

  2520.     /* B A

  2521.      * C X

  2522.      */

  2523.     c = dc_val[ - 1];

  2524.     b = dc_val[ - 1 - wrap];

  2525.     a = dc_val[ - wrap];

  2526.     /* scale predictors if needed */

  2527.     q1 = s->current_picture.f.qscale_table[mb_pos];

  2528.     dqscale_index = s->y_dc_scale_table[q1] - 1;

  2529.     if (dqscale_index < 0)

  2530.         return 0;

  2531.     if (c_avail && (n != 1 && n != 3)) {

  2532.         q2 = s->current_picture.f.qscale_table[mb_pos - 1];

  2533.         if (q2 && q2 != q1)

  2534.             c = (c * s->y_dc_scale_table[q2] * ff_vc1_dqscale[dqscale_index] + 0x20000) >> 18;

  2535.     }

  2536.     if (a_avail && (n != 2 && n != 3)) {

  2537.         q2 = s->current_picture.f.qscale_table[mb_pos - s->mb_stride];

  2538.         if (q2 && q2 != q1)

  2539.             a = (a * s->y_dc_scale_table[q2] * ff_vc1_dqscale[dqscale_index] + 0x20000) >> 18;

  2540.     }

  2541.     if (a_avail && c_avail && (n != 3)) {

  2542.         int off = mb_pos;

  2543.         if (n != 1)

  2544.             off--;

  2545.         if (n != 2)

  2546.             off -= s->mb_stride;

  2547.         q2 = s->current_picture.f.qscale_table[off];

  2548.         if (q2 && q2 != q1)

  2549.             b = (b * s->y_dc_scale_table[q2] * ff_vc1_dqscale[dqscale_index] + 0x20000) >> 18;

  2550.     }

  2551. 

  2552.     if (a_avail && c_avail) {

  2553.         if (abs(a - b) <= abs(b - c)) {

  2554.             pred     = c;

  2555.             *dir_ptr = 1; // left

  2556.         } else {

  2557.             pred     = a;

  2558.             *dir_ptr = 0; // top

  2559.         }

  2560.     } else if (a_avail) {

  2561.         pred     = a;

  2562.         *dir_ptr = 0; // top

  2563.     } else if (c_avail) {

  2564.         pred     = c;

  2565.         *dir_ptr = 1; // left

  2566.     } else {

  2567.         pred     = 0;

  2568.         *dir_ptr = 1; // left

  2569.     }

  2570. 

  2571.     /* update predictor */

  2572.     *dc_val_ptr = &dc_val[0];

  2573.     return pred;

  2574. }

  2575. 

  2576. /** @} */ // Block group

  2577. 

  2578. /**

  2579.  * @name VC1 Macroblock-level functions in Simple/Main Profiles

  2580.  * @see 7.1.4, p91 and 8.1.1.7, p(1)04

  2581.  * @{

  2582.  */

  2583. 

  2584. static inline int vc1_coded_block_pred(MpegEncContext * s, int n,

  2585.                                        uint8_t **coded_block_ptr)

  2586. {

  2587.     int xy, wrap, pred, a, b, c;

  2588. 

  2589.     xy   = s->block_index[n];

  2590.     wrap = s->b8_stride;

  2591. 

  2592.     /* B C

  2593.      * A X

  2594.      */

  2595.     a = s->coded_block[xy - 1       ];

  2596.     b = s->coded_block[xy - 1 - wrap];

  2597.     c = s->coded_block[xy     - wrap];

  2598. 

  2599.     if (b == c) {

  2600.         pred = a;

  2601.     } else {

  2602.         pred = c;

  2603.     }

  2604. 

  2605.     /* store value */

  2606.     *coded_block_ptr = &s->coded_block[xy];

  2607. 

  2608.     return pred;

  2609. }

  2610. 

  2611. /**

  2612.  * Decode one AC coefficient

  2613.  * @param v The VC1 context

  2614.  * @param last Last coefficient

  2615.  * @param skip How much zero coefficients to skip

  2616.  * @param value Decoded AC coefficient value

  2617.  * @param codingset set of VLC to decode data

  2618.  * @see 8.1.3.4

  2619.  */

  2620. static void vc1_decode_ac_coeff(VC1Context *v, int *last, int *skip,

  2621.                                 int *value, int codingset)

  2622. {

  2623.     GetBitContext *gb = &v->s.gb;

  2624.     int index, escape, run = 0, level = 0, lst = 0;

  2625. 

  2626.     index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);

  2627.     if (index != vc1_ac_sizes[codingset] - 1) {

  2628.         run   = vc1_index_decode_table[codingset][index][0];

  2629.         level = vc1_index_decode_table[codingset][index][1];

  2630.         lst   = index >= vc1_last_decode_table[codingset] || get_bits_left(gb) < 0;

  2631.         if (get_bits1(gb))

  2632.             level = -level;

  2633.     } else {

  2634.         escape = decode210(gb);

  2635.         if (escape != 2) {

  2636.             index = get_vlc2(gb, ff_vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);

  2637.             run   = vc1_index_decode_table[codingset][index][0];

  2638.             level = vc1_index_decode_table[codingset][index][1];

  2639.             lst   = index >= vc1_last_decode_table[codingset];

  2640.             if (escape == 0) {

  2641.                 if (lst)

  2642.                     level += vc1_last_delta_level_table[codingset][run];

  2643.                 else

  2644.                     level += vc1_delta_level_table[codingset][run];

  2645.             } else {

  2646.                 if (lst)

  2647.                     run += vc1_last_delta_run_table[codingset][level] + 1;

  2648.                 else

  2649.                     run += vc1_delta_run_table[codingset][level] + 1;

  2650.             }

  2651.             if (get_bits1(gb))

  2652.                 level = -level;

  2653.         } else {

  2654.             int sign;

  2655.             lst = get_bits1(gb);

  2656.             if (v->s.esc3_level_length == 0) {

  2657.                 if (v->pq < 8 || v->dquantfrm) { // table 59

  2658.                     v->s.esc3_level_length = get_bits(gb, 3);

  2659.                     if (!v->s.esc3_level_length)

  2660.                         v->s.esc3_level_length = get_bits(gb, 2) + 8;

  2661.                 } else { // table 60

  2662.                     v->s.esc3_level_length = get_unary(gb, 1, 6) + 2;

  2663.                 }

  2664.                 v->s.esc3_run_length = 3 + get_bits(gb, 2);

  2665.             }

  2666.             run   = get_bits(gb, v->s.esc3_run_length);

  2667.             sign  = get_bits1(gb);

  2668.             level = get_bits(gb, v->s.esc3_level_length);

  2669.             if (sign)

  2670.                 level = -level;

  2671.         }

  2672.     }

  2673. 

  2674.     *last  = lst;

  2675.     *skip  = run;

  2676.     *value = level;

  2677. }

  2678. 

  2679. /** Decode intra block in intra frames - should be faster than decode_intra_block

  2680.  * @param v VC1Context

  2681.  * @param block block to decode

  2682.  * @param[in] n subblock index

  2683.  * @param coded are AC coeffs present or not

  2684.  * @param codingset set of VLC to decode data

  2685.  */

  2686. static int vc1_decode_i_block(VC1Context *v, DCTELEM block[64], int n,

  2687.                               int coded, int codingset)

  2688. {

  2689.     GetBitContext *gb = &v->s.gb;

  2690.     MpegEncContext *s = &v->s;

  2691.     int dc_pred_dir = 0; /* Direction of the DC prediction used */

  2692.     int i;

  2693.     int16_t *dc_val;

  2694.     int16_t *ac_val, *ac_val2;

  2695.     int dcdiff;

  2696. 

  2697.     /* Get DC differential */

  2698.     if (n < 4) {

  2699.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2700.     } else {

  2701.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2702.     }

  2703.     if (dcdiff < 0) {

  2704.         av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");

  2705.         return -1;

  2706.     }

  2707.     if (dcdiff) {

  2708.         if (dcdiff == 119 /* ESC index value */) {

  2709.             /* TODO: Optimize */

  2710.             if (v->pq == 1)      dcdiff = get_bits(gb, 10);

  2711.             else if (v->pq == 2) dcdiff = get_bits(gb, 9);

  2712.             else                 dcdiff = get_bits(gb, 8);

  2713.         } else {

  2714.             if (v->pq == 1)

  2715.                 dcdiff = (dcdiff << 2) + get_bits(gb, 2) - 3;

  2716.             else if (v->pq == 2)

  2717.                 dcdiff = (dcdiff << 1) + get_bits1(gb)   - 1;

  2718.         }

  2719.         if (get_bits1(gb))

  2720.             dcdiff = -dcdiff;

  2721.     }

  2722. 

  2723.     /* Prediction */

  2724.     dcdiff += vc1_i_pred_dc(&v->s, v->overlap, v->pq, n, &dc_val, &dc_pred_dir);

  2725.     *dc_val = dcdiff;

  2726. 

  2727.     /* Store the quantized DC coeff, used for prediction */

  2728.     if (n < 4) {

  2729.         block[0] = dcdiff * s->y_dc_scale;

  2730.     } else {

  2731.         block[0] = dcdiff * s->c_dc_scale;

  2732.     }

  2733.     /* Skip ? */

  2734.     if (!coded) {

  2735.         goto not_coded;

  2736.     }

  2737. 

  2738.     // AC Decoding

  2739.     i = 1;

  2740. 

  2741.     {

  2742.         int last = 0, skip, value;

  2743.         const uint8_t *zz_table;

  2744.         int scale;

  2745.         int k;

  2746. 

  2747.         scale = v->pq * 2 + v->halfpq;

  2748. 

  2749.         if (v->s.ac_pred) {

  2750.             if (!dc_pred_dir)

  2751.                 zz_table = v->zz_8x8[2];

  2752.             else

  2753.                 zz_table = v->zz_8x8[3];

  2754.         } else

  2755.             zz_table = v->zz_8x8[1];

  2756. 

  2757.         ac_val  = s->ac_val[0][0] + s->block_index[n] * 16;

  2758.         ac_val2 = ac_val;

  2759.         if (dc_pred_dir) // left

  2760.             ac_val -= 16;

  2761.         else // top

  2762.             ac_val -= 16 * s->block_wrap[n];

  2763. 

  2764.         while (!last) {

  2765.             vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);

  2766.             i += skip;

  2767.             if (i > 63)

  2768.                 break;

  2769.             block[zz_table[i++]] = value;

  2770.         }

  2771. 

  2772.         /* apply AC prediction if needed */

  2773.         if (s->ac_pred) {

  2774.             if (dc_pred_dir) { // left

  2775.                 for (k = 1; k < 8; k++)

  2776.                     block[k << v->left_blk_sh] += ac_val[k];

  2777.             } else { // top

  2778.                 for (k = 1; k < 8; k++)

  2779.                     block[k << v->top_blk_sh] += ac_val[k + 8];

  2780.             }

  2781.         }

  2782.         /* save AC coeffs for further prediction */

  2783.         for (k = 1; k < 8; k++) {

  2784.             ac_val2[k]     = block[k << v->left_blk_sh];

  2785.             ac_val2[k + 8] = block[k << v->top_blk_sh];

  2786.         }

  2787. 

  2788.         /* scale AC coeffs */

  2789.         for (k = 1; k < 64; k++)

  2790.             if (block[k]) {

  2791.                 block[k] *= scale;

  2792.                 if (!v->pquantizer)

  2793.                     block[k] += (block[k] < 0) ? -v->pq : v->pq;

  2794.             }

  2795. 

  2796.         if (s->ac_pred) i = 63;

  2797.     }

  2798. 

  2799. not_coded:

  2800.     if (!coded) {

  2801.         int k, scale;

  2802.         ac_val  = s->ac_val[0][0] + s->block_index[n] * 16;

  2803.         ac_val2 = ac_val;

  2804. 

  2805.         i = 0;

  2806.         scale = v->pq * 2 + v->halfpq;

  2807.         memset(ac_val2, 0, 16 * 2);

  2808.         if (dc_pred_dir) { // left

  2809.             ac_val -= 16;

  2810.             if (s->ac_pred)

  2811.                 memcpy(ac_val2, ac_val, 8 * 2);

  2812.         } else { // top

  2813.             ac_val -= 16 * s->block_wrap[n];

  2814.             if (s->ac_pred)

  2815.                 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);

  2816.         }

  2817. 

  2818.         /* apply AC prediction if needed */

  2819.         if (s->ac_pred) {

  2820.             if (dc_pred_dir) { //left

  2821.                 for (k = 1; k < 8; k++) {

  2822.                     block[k << v->left_blk_sh] = ac_val[k] * scale;

  2823.                     if (!v->pquantizer && block[k << v->left_blk_sh])

  2824.                         block[k << v->left_blk_sh] += (block[k << v->left_blk_sh] < 0) ? -v->pq : v->pq;

  2825.                 }

  2826.             } else { // top

  2827.                 for (k = 1; k < 8; k++) {

  2828.                     block[k << v->top_blk_sh] = ac_val[k + 8] * scale;

  2829.                     if (!v->pquantizer && block[k << v->top_blk_sh])

  2830.                         block[k << v->top_blk_sh] += (block[k << v->top_blk_sh] < 0) ? -v->pq : v->pq;

  2831.                 }

  2832.             }

  2833.             i = 63;

  2834.         }

  2835.     }

  2836.     s->block_last_index[n] = i;

  2837. 

  2838.     return 0;

  2839. }

  2840. 

  2841. /** Decode intra block in intra frames - should be faster than decode_intra_block

  2842.  * @param v VC1Context

  2843.  * @param block block to decode

  2844.  * @param[in] n subblock number

  2845.  * @param coded are AC coeffs present or not

  2846.  * @param codingset set of VLC to decode data

  2847.  * @param mquant quantizer value for this macroblock

  2848.  */

  2849. static int vc1_decode_i_block_adv(VC1Context *v, DCTELEM block[64], int n,

  2850.                                   int coded, int codingset, int mquant)

  2851. {

  2852.     GetBitContext *gb = &v->s.gb;

  2853.     MpegEncContext *s = &v->s;

  2854.     int dc_pred_dir = 0; /* Direction of the DC prediction used */

  2855.     int i;

  2856.     int16_t *dc_val;

  2857.     int16_t *ac_val, *ac_val2;

  2858.     int dcdiff;

  2859.     int a_avail = v->a_avail, c_avail = v->c_avail;

  2860.     int use_pred = s->ac_pred;

  2861.     int scale;

  2862.     int q1, q2 = 0;

  2863.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  2864. 

  2865.     /* Get DC differential */

  2866.     if (n < 4) {

  2867.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2868.     } else {

  2869.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2870.     }

  2871.     if (dcdiff < 0) {

  2872.         av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");

  2873.         return -1;

  2874.     }

  2875.     if (dcdiff) {

  2876.         if (dcdiff == 119 /* ESC index value */) {

  2877.             /* TODO: Optimize */

  2878.             if (mquant == 1)      dcdiff = get_bits(gb, 10);

  2879.             else if (mquant == 2) dcdiff = get_bits(gb, 9);

  2880.             else                  dcdiff = get_bits(gb, 8);

  2881.         } else {

  2882.             if (mquant == 1)

  2883.                 dcdiff = (dcdiff << 2) + get_bits(gb, 2) - 3;

  2884.             else if (mquant == 2)

  2885.                 dcdiff = (dcdiff << 1) + get_bits1(gb)   - 1;

  2886.         }

  2887.         if (get_bits1(gb))

  2888.             dcdiff = -dcdiff;

  2889.     }

  2890. 

  2891.     /* Prediction */

  2892.     dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, v->a_avail, v->c_avail, &dc_val, &dc_pred_dir);

  2893.     *dc_val = dcdiff;

  2894. 

  2895.     /* Store the quantized DC coeff, used for prediction */

  2896.     if (n < 4) {

  2897.         block[0] = dcdiff * s->y_dc_scale;

  2898.     } else {

  2899.         block[0] = dcdiff * s->c_dc_scale;

  2900.     }

  2901. 

  2902.     //AC Decoding

  2903.     i = 1;

  2904. 

  2905.     /* check if AC is needed at all */

  2906.     if (!a_avail && !c_avail)

  2907.         use_pred = 0;

  2908.     ac_val  = s->ac_val[0][0] + s->block_index[n] * 16;

  2909.     ac_val2 = ac_val;

  2910. 

  2911.     scale = mquant * 2 + ((mquant == v->pq) ? v->halfpq : 0);

  2912. 

  2913.     if (dc_pred_dir) // left

  2914.         ac_val -= 16;

  2915.     else // top

  2916.         ac_val -= 16 * s->block_wrap[n];

  2917. 

  2918.     q1 = s->current_picture.f.qscale_table[mb_pos];

  2919.     if ( dc_pred_dir && c_avail && mb_pos)

  2920.         q2 = s->current_picture.f.qscale_table[mb_pos - 1];

  2921.     if (!dc_pred_dir && a_avail && mb_pos >= s->mb_stride)

  2922.         q2 = s->current_picture.f.qscale_table[mb_pos - s->mb_stride];

  2923.     if ( dc_pred_dir && n == 1)

  2924.         q2 = q1;

  2925.     if (!dc_pred_dir && n == 2)

  2926.         q2 = q1;

  2927.     if (n == 3)

  2928.         q2 = q1;

  2929. 

  2930.     if (coded) {

  2931.         int last = 0, skip, value;

  2932.         const uint8_t *zz_table;

  2933.         int k;

  2934. 

  2935.         if (v->s.ac_pred) {

  2936.             if (!use_pred && v->fcm == ILACE_FRAME) {

  2937.                 zz_table = v->zzi_8x8;

  2938.             } else {

  2939.                 if (!dc_pred_dir) // top

  2940.                     zz_table = v->zz_8x8[2];

  2941.                 else // left

  2942.                     zz_table = v->zz_8x8[3];

  2943.             }

  2944.         } else {

  2945.             if (v->fcm != ILACE_FRAME)

  2946.                 zz_table = v->zz_8x8[1];

  2947.             else

  2948.                 zz_table = v->zzi_8x8;

  2949.         }

  2950. 

  2951.         while (!last) {

  2952.             vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);

  2953.             i += skip;

  2954.             if (i > 63)

  2955.                 break;

  2956.             block[zz_table[i++]] = value;

  2957.         }

  2958. 

  2959.         /* apply AC prediction if needed */

  2960.         if (use_pred) {

  2961.             /* scale predictors if needed*/

  2962.             if (q2 && q1 != q2) {

  2963.                 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  2964.                 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  2965. 

  2966.                 if (q1 < 1)

  2967.                     return AVERROR_INVALIDDATA;

  2968.                 if (dc_pred_dir) { // left

  2969.                     for (k = 1; k < 8; k++)

  2970.                         block[k << v->left_blk_sh] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2971.                 } else { // top

  2972.                     for (k = 1; k < 8; k++)

  2973.                         block[k << v->top_blk_sh] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2974.                 }

  2975.             } else {

  2976.                 if (dc_pred_dir) { //left

  2977.                     for (k = 1; k < 8; k++)

  2978.                         block[k << v->left_blk_sh] += ac_val[k];

  2979.                 } else { //top

  2980.                     for (k = 1; k < 8; k++)

  2981.                         block[k << v->top_blk_sh] += ac_val[k + 8];

  2982.                 }

  2983.             }

  2984.         }

  2985.         /* save AC coeffs for further prediction */

  2986.         for (k = 1; k < 8; k++) {

  2987.             ac_val2[k    ] = block[k << v->left_blk_sh];

  2988.             ac_val2[k + 8] = block[k << v->top_blk_sh];

  2989.         }

  2990. 

  2991.         /* scale AC coeffs */

  2992.         for (k = 1; k < 64; k++)

  2993.             if (block[k]) {

  2994.                 block[k] *= scale;

  2995.                 if (!v->pquantizer)

  2996.                     block[k] += (block[k] < 0) ? -mquant : mquant;

  2997.             }

  2998. 

  2999.         if (use_pred) i = 63;

  3000.     } else { // no AC coeffs

  3001.         int k;

  3002. 

  3003.         memset(ac_val2, 0, 16 * 2);

  3004.         if (dc_pred_dir) { // left

  3005.             if (use_pred) {

  3006.                 memcpy(ac_val2, ac_val, 8 * 2);

  3007.                 if (q2 && q1 != q2) {

  3008.                     q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  3009.                     q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  3010.                     if (q1 < 1)

  3011.                         return AVERROR_INVALIDDATA;

  3012.                     for (k = 1; k < 8; k++)

  3013.                         ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  3014.                 }

  3015.             }

  3016.         } else { // top

  3017.             if (use_pred) {

  3018.                 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);

  3019.                 if (q2 && q1 != q2) {

  3020.                     q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  3021.                     q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  3022.                     if (q1 < 1)

  3023.                         return AVERROR_INVALIDDATA;

  3024.                     for (k = 1; k < 8; k++)

  3025.                         ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  3026.                 }

  3027.             }

  3028.         }

  3029. 

  3030.         /* apply AC prediction if needed */

  3031.         if (use_pred) {

  3032.             if (dc_pred_dir) { // left

  3033.                 for (k = 1; k < 8; k++) {

  3034.                     block[k << v->left_blk_sh] = ac_val2[k] * scale;

  3035.                     if (!v->pquantizer && block[k << v->left_blk_sh])

  3036.                         block[k << v->left_blk_sh] += (block[k << v->left_blk_sh] < 0) ? -mquant : mquant;

  3037.                 }

  3038.             } else { // top

  3039.                 for (k = 1; k < 8; k++) {

  3040.                     block[k << v->top_blk_sh] = ac_val2[k + 8] * scale;

  3041.                     if (!v->pquantizer && block[k << v->top_blk_sh])

  3042.                         block[k << v->top_blk_sh] += (block[k << v->top_blk_sh] < 0) ? -mquant : mquant;

  3043.                 }

  3044.             }

  3045.             i = 63;

  3046.         }

  3047.     }

  3048.     s->block_last_index[n] = i;

  3049. 

  3050.     return 0;

  3051. }

  3052. 

  3053. /** Decode intra block in inter frames - more generic version than vc1_decode_i_block

  3054.  * @param v VC1Context

  3055.  * @param block block to decode

  3056.  * @param[in] n subblock index

  3057.  * @param coded are AC coeffs present or not

  3058.  * @param mquant block quantizer

  3059.  * @param codingset set of VLC to decode data

  3060.  */

  3061. static int vc1_decode_intra_block(VC1Context *v, DCTELEM block[64], int n,

  3062.                                   int coded, int mquant, int codingset)

  3063. {

  3064.     GetBitContext *gb = &v->s.gb;

  3065.     MpegEncContext *s = &v->s;

  3066.     int dc_pred_dir = 0; /* Direction of the DC prediction used */

  3067.     int i;

  3068.     int16_t *dc_val;

  3069.     int16_t *ac_val, *ac_val2;

  3070.     int dcdiff;

  3071.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  3072.     int a_avail = v->a_avail, c_avail = v->c_avail;

  3073.     int use_pred = s->ac_pred;

  3074.     int scale;

  3075.     int q1, q2 = 0;

  3076. 

  3077.     s->dsp.clear_block(block);

  3078. 

  3079.     /* XXX: Guard against dumb values of mquant */

  3080.     mquant = (mquant < 1) ? 0 : ((mquant > 31) ? 31 : mquant);

  3081. 

  3082.     /* Set DC scale - y and c use the same */

  3083.     s->y_dc_scale = s->y_dc_scale_table[mquant];

  3084.     s->c_dc_scale = s->c_dc_scale_table[mquant];

  3085. 

  3086.     /* Get DC differential */

  3087.     if (n < 4) {

  3088.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  3089.     } else {

  3090.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  3091.     }

  3092.     if (dcdiff < 0) {

  3093.         av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");

  3094.         return -1;

  3095.     }

  3096.     if (dcdiff) {

  3097.         if (dcdiff == 119 /* ESC index value */) {

  3098.             /* TODO: Optimize */

  3099.             if (mquant == 1)      dcdiff = get_bits(gb, 10);

  3100.             else if (mquant == 2) dcdiff = get_bits(gb, 9);

  3101.             else                  dcdiff = get_bits(gb, 8);

  3102.         } else {

  3103.             if (mquant == 1)

  3104.                 dcdiff = (dcdiff << 2) + get_bits(gb, 2) - 3;

  3105.             else if (mquant == 2)

  3106.                 dcdiff = (dcdiff << 1) + get_bits1(gb)   - 1;

  3107.         }

  3108.         if (get_bits1(gb))

  3109.             dcdiff = -dcdiff;

  3110.     }

  3111. 

  3112.     /* Prediction */

  3113.     dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, a_avail, c_avail, &dc_val, &dc_pred_dir);

  3114.     *dc_val = dcdiff;

  3115. 

  3116.     /* Store the quantized DC coeff, used for prediction */

  3117. 

  3118.     if (n < 4) {

  3119.         block[0] = dcdiff * s->y_dc_scale;

  3120.     } else {

  3121.         block[0] = dcdiff * s->c_dc_scale;

  3122.     }

  3123. 

  3124.     //AC Decoding

  3125.     i = 1;

  3126. 

  3127.     /* check if AC is needed at all and adjust direction if needed */

  3128.     if (!a_avail) dc_pred_dir = 1;

  3129.     if (!c_avail) dc_pred_dir = 0;

  3130.     if (!a_avail && !c_avail) use_pred = 0;

  3131.     ac_val = s->ac_val[0][0] + s->block_index[n] * 16;

  3132.     ac_val2 = ac_val;

  3133. 

  3134.     scale = mquant * 2 + v->halfpq;

  3135. 

  3136.     if (dc_pred_dir) //left

  3137.         ac_val -= 16;

  3138.     else //top

  3139.         ac_val -= 16 * s->block_wrap[n];

  3140. 

  3141.     q1 = s->current_picture.f.qscale_table[mb_pos];

  3142.     if (dc_pred_dir && c_avail && mb_pos)

  3143.         q2 = s->current_picture.f.qscale_table[mb_pos - 1];

  3144.     if (!dc_pred_dir && a_avail && mb_pos >= s->mb_stride)

  3145.         q2 = s->current_picture.f.qscale_table[mb_pos - s->mb_stride];

  3146.     if ( dc_pred_dir && n == 1)

  3147.         q2 = q1;

  3148.     if (!dc_pred_dir && n == 2)

  3149.         q2 = q1;

  3150.     if (n == 3) q2 = q1;

  3151. 

  3152.     if (coded) {

  3153.         int last = 0, skip, value;

  3154.         int k;

  3155. 

  3156.         while (!last) {

  3157.             vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);

  3158.             i += skip;

  3159.             if (i > 63)

  3160.                 break;

  3161.             if (v->fcm == PROGRESSIVE)

  3162.                 block[v->zz_8x8[0][i++]] = value;

  3163.             else {

  3164.                 if (use_pred && (v->fcm == ILACE_FRAME)) {

  3165.                     if (!dc_pred_dir) // top

  3166.                         block[v->zz_8x8[2][i++]] = value;

  3167.                     else // left

  3168.                         block[v->zz_8x8[3][i++]] = value;

  3169.                 } else {

  3170.                     block[v->zzi_8x8[i++]] = value;

  3171.                 }

  3172.             }

  3173.         }

  3174. 

  3175.         /* apply AC prediction if needed */

  3176.         if (use_pred) {

  3177.             /* scale predictors if needed*/

  3178.             if (q2 && q1 != q2) {

  3179.                 q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  3180.                 q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  3181. 

  3182.                 if (q1 < 1)

  3183.                     return AVERROR_INVALIDDATA;

  3184.                 if (dc_pred_dir) { // left

  3185.                     for (k = 1; k < 8; k++)

  3186.                         block[k << v->left_blk_sh] += (ac_val[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  3187.                 } else { //top

  3188.                     for (k = 1; k < 8; k++)

  3189.                         block[k << v->top_blk_sh] += (ac_val[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  3190.                 }

  3191.             } else {

  3192.                 if (dc_pred_dir) { // left

  3193.                     for (k = 1; k < 8; k++)

  3194.                         block[k << v->left_blk_sh] += ac_val[k];

  3195.                 } else { // top

  3196.                     for (k = 1; k < 8; k++)

  3197.                         block[k << v->top_blk_sh] += ac_val[k + 8];

  3198.                 }

  3199.             }

  3200.         }

  3201.         /* save AC coeffs for further prediction */

  3202.         for (k = 1; k < 8; k++) {

  3203.             ac_val2[k    ] = block[k << v->left_blk_sh];

  3204.             ac_val2[k + 8] = block[k << v->top_blk_sh];

  3205.         }

  3206. 

  3207.         /* scale AC coeffs */

  3208.         for (k = 1; k < 64; k++)

  3209.             if (block[k]) {

  3210.                 block[k] *= scale;

  3211.                 if (!v->pquantizer)

  3212.                     block[k] += (block[k] < 0) ? -mquant : mquant;

  3213.             }

  3214. 

  3215.         if (use_pred) i = 63;

  3216.     } else { // no AC coeffs

  3217.         int k;

  3218. 

  3219.         memset(ac_val2, 0, 16 * 2);

  3220.         if (dc_pred_dir) { // left

  3221.             if (use_pred) {

  3222.                 memcpy(ac_val2, ac_val, 8 * 2);

  3223.                 if (q2 && q1 != q2) {

  3224.                     q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  3225.                     q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  3226.                     if (q1 < 1)

  3227.                         return AVERROR_INVALIDDATA;

  3228.                     for (k = 1; k < 8; k++)

  3229.                         ac_val2[k] = (ac_val2[k] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  3230.                 }

  3231.             }

  3232.         } else { // top

  3233.             if (use_pred) {

  3234.                 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);

  3235.                 if (q2 && q1 != q2) {

  3236.                     q1 = q1 * 2 + ((q1 == v->pq) ? v->halfpq : 0) - 1;

  3237.                     q2 = q2 * 2 + ((q2 == v->pq) ? v->halfpq : 0) - 1;

  3238.                     if (q1 < 1)

  3239.                         return AVERROR_INVALIDDATA;

  3240.                     for (k = 1; k < 8; k++)

  3241.                         ac_val2[k + 8] = (ac_val2[k + 8] * q2 * ff_vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  3242.                 }

  3243.             }

  3244.         }

  3245. 

  3246.         /* apply AC prediction if needed */

  3247.         if (use_pred) {

  3248.             if (dc_pred_dir) { // left

  3249.                 for (k = 1; k < 8; k++) {

  3250.                     block[k << v->left_blk_sh] = ac_val2[k] * scale;

  3251.                     if (!v->pquantizer && block[k << v->left_blk_sh])

  3252.                         block[k << v->left_blk_sh] += (block[k << v->left_blk_sh] < 0) ? -mquant : mquant;

  3253.                 }

  3254.             } else { // top

  3255.                 for (k = 1; k < 8; k++) {

  3256.                     block[k << v->top_blk_sh] = ac_val2[k + 8] * scale;

  3257.                     if (!v->pquantizer && block[k << v->top_blk_sh])

  3258.                         block[k << v->top_blk_sh] += (block[k << v->top_blk_sh] < 0) ? -mquant : mquant;

  3259.                 }

  3260.             }

  3261.             i = 63;

  3262.         }

  3263.     }

  3264.     s->block_last_index[n] = i;

  3265. 

  3266.     return 0;

  3267. }

  3268. 

  3269. /** Decode P block

  3270.  */

  3271. static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n,

  3272.                               int mquant, int ttmb, int first_block,

  3273.                               uint8_t *dst, int linesize, int skip_block,

  3274.                               int *ttmb_out)

  3275. {

  3276.     MpegEncContext *s = &v->s;

  3277.     GetBitContext *gb = &s->gb;

  3278.     int i, j;

  3279.     int subblkpat = 0;

  3280.     int scale, off, idx, last, skip, value;

  3281.     int ttblk = ttmb & 7;

  3282.     int pat = 0;

  3283. 

  3284.     s->dsp.clear_block(block);

  3285. 

  3286.     if (ttmb == -1) {

  3287.         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)];

  3288.     }

  3289.     if (ttblk == TT_4X4) {

  3290.         subblkpat = ~(get_vlc2(gb, ff_vc1_subblkpat_vlc[v->tt_index].table, VC1_SUBBLKPAT_VLC_BITS, 1) + 1);

  3291.     }

  3292.     if ((ttblk != TT_8X8 && ttblk != TT_4X4)

  3293.         && ((v->ttmbf || (ttmb != -1 && (ttmb & 8) && !first_block))

  3294.             || (!v->res_rtm_flag && !first_block))) {

  3295.         subblkpat = decode012(gb);

  3296.         if (subblkpat)

  3297.             subblkpat ^= 3; // swap decoded pattern bits

  3298.         if (ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM)

  3299.             ttblk = TT_8X4;

  3300.         if (ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT)

  3301.             ttblk = TT_4X8;

  3302.     }

  3303.     scale = 2 * mquant + ((v->pq == mquant) ? v->halfpq : 0);

  3304. 

  3305.     // convert transforms like 8X4_TOP to generic TT and SUBBLKPAT

  3306.     if (ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) {

  3307.         subblkpat = 2 - (ttblk == TT_8X4_TOP);

  3308.         ttblk     = TT_8X4;

  3309.     }

  3310.     if (ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) {

  3311.         subblkpat = 2 - (ttblk == TT_4X8_LEFT);

  3312.         ttblk     = TT_4X8;

  3313.     }

  3314.     switch (ttblk) {

  3315.     case TT_8X8:

  3316.         pat  = 0xF;

  3317.         i    = 0;

  3318.         last = 0;

  3319.         while (!last) {

  3320.             vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  3321.             i += skip;

  3322.             if (i > 63)

  3323.                 break;

  3324.             if (!v->fcm)

  3325.                 idx = v->zz_8x8[0][i++];

  3326.             else

  3327.                 idx = v->zzi_8x8[i++];

  3328.             block[idx] = value * scale;

  3329.             if (!v->pquantizer)

  3330.                 block[idx] += (block[idx] < 0) ? -mquant : mquant;

  3331.         }

  3332.         if (!skip_block) {

  3333.             if (i == 1)

  3334.                 v->vc1dsp.vc1_inv_trans_8x8_dc(dst, linesize, block);

  3335.             else {

  3336.                 v->vc1dsp.vc1_inv_trans_8x8(block);

  3337.                 s->dsp.add_pixels_clamped(block, dst, linesize);

  3338.             }

  3339.         }

  3340.         break;

  3341.     case TT_4X4:

  3342.         pat = ~subblkpat & 0xF;

  3343.         for (j = 0; j < 4; j++) {

  3344.             last = subblkpat & (1 << (3 - j));

  3345.             i    = 0;

  3346.             off  = (j & 1) * 4 + (j & 2) * 16;

  3347.             while (!last) {

  3348.                 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  3349.                 i += skip;

  3350.                 if (i > 15)

  3351.                     break;

  3352.                 if (!v->fcm)

  3353.                     idx = ff_vc1_simple_progressive_4x4_zz[i++];

  3354.                 else

  3355.                     idx = ff_vc1_adv_interlaced_4x4_zz[i++];

  3356.                 block[idx + off] = value * scale;

  3357.                 if (!v->pquantizer)

  3358.                     block[idx + off] += (block[idx + off] < 0) ? -mquant : mquant;

  3359.             }

  3360.             if (!(subblkpat & (1 << (3 - j))) && !skip_block) {

  3361.                 if (i == 1)

  3362.                     v->vc1dsp.vc1_inv_trans_4x4_dc(dst + (j & 1) * 4 + (j & 2) * 2 * linesize, linesize, block + off);

  3363.                 else

  3364.                     v->vc1dsp.vc1_inv_trans_4x4(dst + (j & 1) * 4 + (j & 2) *  2 * linesize, linesize, block + off);

  3365.             }

  3366.         }

  3367.         break;

  3368.     case TT_8X4:

  3369.         pat = ~((subblkpat & 2) * 6 + (subblkpat & 1) * 3) & 0xF;

  3370.         for (j = 0; j < 2; j++) {

  3371.             last = subblkpat & (1 << (1 - j));

  3372.             i    = 0;

  3373.             off  = j * 32;

  3374.             while (!last) {

  3375.                 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  3376.                 i += skip;

  3377.                 if (i > 31)

  3378.                     break;

  3379.                 if (!v->fcm)

  3380.                     idx = v->zz_8x4[i++] + off;

  3381.                 else

  3382.                     idx = ff_vc1_adv_interlaced_8x4_zz[i++] + off;

  3383.                 block[idx] = value * scale;

  3384.                 if (!v->pquantizer)

  3385.                     block[idx] += (block[idx] < 0) ? -mquant : mquant;

  3386.             }

  3387.             if (!(subblkpat & (1 << (1 - j))) && !skip_block) {

  3388.                 if (i == 1)

  3389.                     v->vc1dsp.vc1_inv_trans_8x4_dc(dst + j * 4 * linesize, linesize, block + off);

  3390.                 else

  3391.                     v->vc1dsp.vc1_inv_trans_8x4(dst + j * 4 * linesize, linesize, block + off);

  3392.             }

  3393.         }

  3394.         break;

  3395.     case TT_4X8:

  3396.         pat = ~(subblkpat * 5) & 0xF;

  3397.         for (j = 0; j < 2; j++) {

  3398.             last = subblkpat & (1 << (1 - j));

  3399.             i    = 0;

  3400.             off  = j * 4;

  3401.             while (!last) {

  3402.                 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  3403.                 i += skip;

  3404.                 if (i > 31)

  3405.                     break;

  3406.                 if (!v->fcm)

  3407.                     idx = v->zz_4x8[i++] + off;

  3408.                 else

  3409.                     idx = ff_vc1_adv_interlaced_4x8_zz[i++] + off;

  3410.                 block[idx] = value * scale;

  3411.                 if (!v->pquantizer)

  3412.                     block[idx] += (block[idx] < 0) ? -mquant : mquant;

  3413.             }

  3414.             if (!(subblkpat & (1 << (1 - j))) && !skip_block) {

  3415.                 if (i == 1)

  3416.                     v->vc1dsp.vc1_inv_trans_4x8_dc(dst + j * 4, linesize, block + off);

  3417.                 else

  3418.                     v->vc1dsp.vc1_inv_trans_4x8(dst + j*4, linesize, block + off);

  3419.             }

  3420.         }

  3421.         break;

  3422.     }

  3423.     if (ttmb_out)

  3424.         *ttmb_out |= ttblk << (n * 4);

  3425.     return pat;

  3426. }

  3427. 

  3428. /** @} */ // Macroblock group

  3429. 

  3430. static const int size_table  [6] = { 0, 2, 3, 4,  5,  8 };

  3431. static const int offset_table[6] = { 0, 1, 3, 7, 15, 31 };

  3432. 

  3433. static av_always_inline void vc1_apply_p_v_loop_filter(VC1Context *v, int block_num)

  3434. {

  3435.     MpegEncContext *s  = &v->s;

  3436.     int mb_cbp         = v->cbp[s->mb_x - s->mb_stride],

  3437.         block_cbp      = mb_cbp      >> (block_num * 4), bottom_cbp,

  3438.         mb_is_intra    = v->is_intra[s->mb_x - s->mb_stride],

  3439.         block_is_intra = mb_is_intra >> (block_num * 4), bottom_is_intra;

  3440.     int idx, linesize  = block_num > 3 ? s->uvlinesize : s->linesize, ttblk;

  3441.     uint8_t *dst;

  3442. 

  3443.     if (block_num > 3) {

  3444.         dst      = s->dest[block_num - 3];

  3445.     } else {

  3446.         dst      = s->dest[0] + (block_num & 1) * 8 + ((block_num & 2) * 4 - 8) * linesize;

  3447.     }

  3448.     if (s->mb_y != s->end_mb_y || block_num < 2) {

  3449.         int16_t (*mv)[2];

  3450.         int mv_stride;

  3451. 

  3452.         if (block_num > 3) {

  3453.             bottom_cbp      = v->cbp[s->mb_x]      >> (block_num * 4);

  3454.             bottom_is_intra = v->is_intra[s->mb_x] >> (block_num * 4);

  3455.             mv              = &v->luma_mv[s->mb_x - s->mb_stride];

  3456.             mv_stride       = s->mb_stride;

  3457.         } else {

  3458.             bottom_cbp      = (block_num < 2) ? (mb_cbp               >> ((block_num + 2) * 4))

  3459.                                               : (v->cbp[s->mb_x]      >> ((block_num - 2) * 4));

  3460.             bottom_is_intra = (block_num < 2) ? (mb_is_intra          >> ((block_num + 2) * 4))

  3461.                                               : (v->is_intra[s->mb_x] >> ((block_num - 2) * 4));

  3462.             mv_stride       = s->b8_stride;

  3463.             mv              = &s->current_picture.f.motion_val[0][s->block_index[block_num] - 2 * mv_stride];

  3464.         }

  3465. 

  3466.         if (bottom_is_intra & 1 || block_is_intra & 1 ||

  3467.             mv[0][0] != mv[mv_stride][0] || mv[0][1] != mv[mv_stride][1]) {

  3468.             v->vc1dsp.vc1_v_loop_filter8(dst, linesize, v->pq);

  3469.         } else {

  3470.             idx = ((bottom_cbp >> 2) | block_cbp) & 3;

  3471.             if (idx == 3) {

  3472.                 v->vc1dsp.vc1_v_loop_filter8(dst, linesize, v->pq);

  3473.             } else if (idx) {

  3474.                 if (idx == 1)

  3475.                     v->vc1dsp.vc1_v_loop_filter4(dst + 4, linesize, v->pq);

  3476.                 else

  3477.                     v->vc1dsp.vc1_v_loop_filter4(dst,     linesize, v->pq);

  3478.             }

  3479.         }

  3480.     }

  3481. 

  3482.     dst -= 4 * linesize;

  3483.     ttblk = (v->ttblk[s->mb_x - s->mb_stride] >> (block_num * 4)) & 0xF;

  3484.     if (ttblk == TT_4X4 || ttblk == TT_8X4) {

  3485.         idx = (block_cbp | (block_cbp >> 2)) & 3;

  3486.         if (idx == 3) {

  3487.             v->vc1dsp.vc1_v_loop_filter8(dst, linesize, v->pq);

  3488.         } else if (idx) {

  3489.             if (idx == 1)

  3490.                 v->vc1dsp.vc1_v_loop_filter4(dst + 4, linesize, v->pq);

  3491.             else

  3492.                 v->vc1dsp.vc1_v_loop_filter4(dst,     linesize, v->pq);

  3493.         }

  3494.     }

  3495. }

  3496. 

  3497. static av_always_inline void vc1_apply_p_h_loop_filter(VC1Context *v, int block_num)

  3498. {

  3499.     MpegEncContext *s  = &v->s;

  3500.     int mb_cbp         = v->cbp[s->mb_x - 1 - s->mb_stride],

  3501.         block_cbp      = mb_cbp      >> (block_num * 4), right_cbp,

  3502.         mb_is_intra    = v->is_intra[s->mb_x - 1 - s->mb_stride],

  3503.         block_is_intra = mb_is_intra >> (block_num * 4), right_is_intra;

  3504.     int idx, linesize  = block_num > 3 ? s->uvlinesize : s->linesize, ttblk;

  3505.     uint8_t *dst;

  3506. 

  3507.     if (block_num > 3) {

  3508.         dst = s->dest[block_num - 3] - 8 * linesize;

  3509.     } else {

  3510.         dst = s->dest[0] + (block_num & 1) * 8 + ((block_num & 2) * 4 - 16) * linesize - 8;

  3511.     }

  3512. 

  3513.     if (s->mb_x != s->mb_width || !(block_num & 5)) {

  3514.         int16_t (*mv)[2];

  3515. 

  3516.         if (block_num > 3) {

  3517.             right_cbp      = v->cbp[s->mb_x - s->mb_stride] >> (block_num * 4);

  3518.             right_is_intra = v->is_intra[s->mb_x - s->mb_stride] >> (block_num * 4);

  3519.             mv             = &v->luma_mv[s->mb_x - s->mb_stride - 1];

  3520.         } else {

  3521.             right_cbp      = (block_num & 1) ? (v->cbp[s->mb_x - s->mb_stride]      >> ((block_num - 1) * 4))

  3522.                                              : (mb_cbp                              >> ((block_num + 1) * 4));

  3523.             right_is_intra = (block_num & 1) ? (v->is_intra[s->mb_x - s->mb_stride] >> ((block_num - 1) * 4))

  3524.                                              : (mb_is_intra                         >> ((block_num + 1) * 4));

  3525.             mv             = &s->current_picture.f.motion_val[0][s->block_index[block_num] - s->b8_stride * 2 - 2];

  3526.         }

  3527.         if (block_is_intra & 1 || right_is_intra & 1 || mv[0][0] != mv[1][0] || mv[0][1] != mv[1][1]) {

  3528.             v->vc1dsp.vc1_h_loop_filter8(dst, linesize, v->pq);

  3529.         } else {

  3530.             idx = ((right_cbp >> 1) | block_cbp) & 5; // FIXME check

  3531.             if (idx == 5) {

  3532.                 v->vc1dsp.vc1_h_loop_filter8(dst, linesize, v->pq);

  3533.             } else if (idx) {

  3534.                 if (idx == 1)

  3535.                     v->vc1dsp.vc1_h_loop_filter4(dst + 4 * linesize, linesize, v->pq);

  3536.                 else

  3537.                     v->vc1dsp.vc1_h_loop_filter4(dst,                linesize, v->pq);

  3538.             }

  3539.         }

  3540.     }

  3541. 

  3542.     dst -= 4;

  3543.     ttblk = (v->ttblk[s->mb_x - s->mb_stride - 1] >> (block_num * 4)) & 0xf;

  3544.     if (ttblk == TT_4X4 || ttblk == TT_4X8) {

  3545.         idx = (block_cbp | (block_cbp >> 1)) & 5;

  3546.         if (idx == 5) {

  3547.             v->vc1dsp.vc1_h_loop_filter8(dst, linesize, v->pq);

  3548.         } else if (idx) {

  3549.             if (idx == 1)

  3550.                 v->vc1dsp.vc1_h_loop_filter4(dst + linesize * 4, linesize, v->pq);

  3551.             else

  3552.                 v->vc1dsp.vc1_h_loop_filter4(dst,                linesize, v->pq);

  3553.         }

  3554.     }

  3555. }

  3556. 

  3557. static void vc1_apply_p_loop_filter(VC1Context *v)

  3558. {

  3559.     MpegEncContext *s = &v->s;

  3560.     int i;

  3561. 

  3562.     for (i = 0; i < 6; i++) {

  3563.         vc1_apply_p_v_loop_filter(v, i);

  3564.     }

  3565. 

  3566.     /* V always precedes H, therefore we run H one MB before V;

  3567.      * at the end of a row, we catch up to complete the row */

  3568.     if (s->mb_x) {

  3569.         for (i = 0; i < 6; i++) {

  3570.             vc1_apply_p_h_loop_filter(v, i);

  3571.         }

  3572.         if (s->mb_x == s->mb_width - 1) {

  3573.             s->mb_x++;

  3574.             ff_update_block_index(s);

  3575.             for (i = 0; i < 6; i++) {

  3576.                 vc1_apply_p_h_loop_filter(v, i);

  3577.             }

  3578.         }

  3579.     }

  3580. }

  3581. 

  3582. /** Decode one P-frame MB

  3583.  */

  3584. static int vc1_decode_p_mb(VC1Context *v)

  3585. {

  3586.     MpegEncContext *s = &v->s;

  3587.     GetBitContext *gb = &s->gb;

  3588.     int i, j;

  3589.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  3590.     int cbp; /* cbp decoding stuff */

  3591.     int mqdiff, mquant; /* MB quantization */

  3592.     int ttmb = v->ttfrm; /* MB Transform type */

  3593. 

  3594.     int mb_has_coeffs = 1; /* last_flag */

  3595.     int dmv_x, dmv_y; /* Differential MV components */

  3596.     int index, index1; /* LUT indexes */

  3597.     int val, sign; /* temp values */

  3598.     int first_block = 1;

  3599.     int dst_idx, off;

  3600.     int skipped, fourmv;

  3601.     int block_cbp = 0, pat, block_tt = 0, block_intra = 0;

  3602. 

  3603.     mquant = v->pq; /* lossy initialization */

  3604. 

  3605.     if (v->mv_type_is_raw)

  3606.         fourmv = get_bits1(gb);

  3607.     else

  3608.         fourmv = v->mv_type_mb_plane[mb_pos];

  3609.     if (v->skip_is_raw)

  3610.         skipped = get_bits1(gb);

  3611.     else

  3612.         skipped = v->s.mbskip_table[mb_pos];

  3613. 

  3614.     if (!fourmv) { /* 1MV mode */

  3615.         if (!skipped) {

  3616.             GET_MVDATA(dmv_x, dmv_y);

  3617. 

  3618.             if (s->mb_intra) {

  3619.                 s->current_picture.f.motion_val[1][s->block_index[0]][0] = 0;

  3620.                 s->current_picture.f.motion_val[1][s->block_index[0]][1] = 0;

  3621.             }

  3622.             s->current_picture.f.mb_type[mb_pos] = s->mb_intra ? MB_TYPE_INTRA : MB_TYPE_16x16;

  3623.             vc1_pred_mv(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], 0, 0);

  3624. 

  3625.             /* FIXME Set DC val for inter block ? */

  3626.             if (s->mb_intra && !mb_has_coeffs) {

  3627.                 GET_MQUANT();

  3628.                 s->ac_pred = get_bits1(gb);

  3629.                 cbp        = 0;

  3630.             } else if (mb_has_coeffs) {

  3631.                 if (s->mb_intra)

  3632.                     s->ac_pred = get_bits1(gb);

  3633.                 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  3634.                 GET_MQUANT();

  3635.             } else {

  3636.                 mquant = v->pq;

  3637.                 cbp    = 0;

  3638.             }

  3639.             s->current_picture.f.qscale_table[mb_pos] = mquant;

  3640. 

  3641.             if (!v->ttmbf && !s->mb_intra && mb_has_coeffs)

  3642.                 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table,

  3643.                                 VC1_TTMB_VLC_BITS, 2);

  3644.             if (!s->mb_intra) vc1_mc_1mv(v, 0);

  3645.             dst_idx = 0;

  3646.             for (i = 0; i < 6; i++) {

  3647.                 s->dc_val[0][s->block_index[i]] = 0;

  3648.                 dst_idx += i >> 2;

  3649.                 val = ((cbp >> (5 - i)) & 1);

  3650.                 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  3651.                 v->mb_type[0][s->block_index[i]] = s->mb_intra;

  3652.                 if (s->mb_intra) {

  3653.                     /* check if prediction blocks A and C are available */

  3654.                     v->a_avail = v->c_avail = 0;

  3655.                     if (i == 2 || i == 3 || !s->first_slice_line)

  3656.                         v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  3657.                     if (i == 1 || i == 3 || s->mb_x)

  3658.                         v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  3659. 

  3660.                     vc1_decode_intra_block(v, s->block[i], i, val, mquant,

  3661.                                            (i & 4) ? v->codingset2 : v->codingset);

  3662.                     if ((i>3) && (s->flags & CODEC_FLAG_GRAY))

  3663.                         continue;

  3664.                     v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);

  3665.                     if (v->rangeredfrm)

  3666.                         for (j = 0; j < 64; j++)

  3667.                             s->block[i][j] <<= 1;

  3668.                     s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);

  3669.                     if (v->pq >= 9 && v->overlap) {

  3670.                         if (v->c_avail)

  3671.                             v->vc1dsp.vc1_h_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);

  3672.                         if (v->a_avail)

  3673.                             v->vc1dsp.vc1_v_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);

  3674.                     }

  3675.                     block_cbp   |= 0xF << (i << 2);

  3676.                     block_intra |= 1 << i;

  3677.                 } else if (val) {

  3678.                     pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block,

  3679.                                              s->dest[dst_idx] + off, (i & 4) ? s->uvlinesize : s->linesize,

  3680.                                              (i & 4) && (s->flags & CODEC_FLAG_GRAY), &block_tt);

  3681.                     block_cbp |= pat << (i << 2);

  3682.                     if (!v->ttmbf && ttmb < 8)

  3683.                         ttmb = -1;

  3684.                     first_block = 0;

  3685.                 }

  3686.             }

  3687.         } else { // skipped

  3688.             s->mb_intra = 0;

  3689.             for (i = 0; i < 6; i++) {

  3690.                 v->mb_type[0][s->block_index[i]] = 0;

  3691.                 s->dc_val[0][s->block_index[i]]  = 0;

  3692.             }

  3693.             s->current_picture.f.mb_type[mb_pos]      = MB_TYPE_SKIP;

  3694.             s->current_picture.f.qscale_table[mb_pos] = 0;

  3695.             vc1_pred_mv(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0], 0, 0);

  3696.             vc1_mc_1mv(v, 0);

  3697.         }

  3698.     } else { // 4MV mode

  3699.         if (!skipped /* unskipped MB */) {

  3700.             int intra_count = 0, coded_inter = 0;

  3701.             int is_intra[6], is_coded[6];

  3702.             /* Get CBPCY */

  3703.             cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  3704.             for (i = 0; i < 6; i++) {

  3705.                 val = ((cbp >> (5 - i)) & 1);

  3706.                 s->dc_val[0][s->block_index[i]] = 0;

  3707.                 s->mb_intra                     = 0;

  3708.                 if (i < 4) {

  3709.                     dmv_x = dmv_y = 0;

  3710.                     s->mb_intra   = 0;

  3711.                     mb_has_coeffs = 0;

  3712.                     if (val) {

  3713.                         GET_MVDATA(dmv_x, dmv_y);

  3714.                     }

  3715.                     vc1_pred_mv(v, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0], 0, 0);

  3716.                     if (!s->mb_intra)

  3717.                         vc1_mc_4mv_luma(v, i, 0);

  3718.                     intra_count += s->mb_intra;

  3719.                     is_intra[i]  = s->mb_intra;

  3720.                     is_coded[i]  = mb_has_coeffs;

  3721.                 }

  3722.                 if (i & 4) {

  3723.                     is_intra[i] = (intra_count >= 3);

  3724.                     is_coded[i] = val;

  3725.                 }

  3726.                 if (i == 4)

  3727.                     vc1_mc_4mv_chroma(v, 0);

  3728.                 v->mb_type[0][s->block_index[i]] = is_intra[i];

  3729.                 if (!coded_inter)

  3730.                     coded_inter = !is_intra[i] & is_coded[i];

  3731.             }

  3732.             // if there are no coded blocks then don't do anything more

  3733.             dst_idx = 0;

  3734.             if (!intra_count && !coded_inter)

  3735.                 goto end;

  3736.             GET_MQUANT();

  3737.             s->current_picture.f.qscale_table[mb_pos] = mquant;

  3738.             /* test if block is intra and has pred */

  3739.             {

  3740.                 int intrapred = 0;

  3741.                 for (i = 0; i < 6; i++)

  3742.                     if (is_intra[i]) {

  3743.                         if (((!s->first_slice_line || (i == 2 || i == 3)) && v->mb_type[0][s->block_index[i] - s->block_wrap[i]])

  3744.                             || ((s->mb_x || (i == 1 || i == 3)) && v->mb_type[0][s->block_index[i] - 1])) {

  3745.                             intrapred = 1;

  3746.                             break;

  3747.                         }

  3748.                     }

  3749.                 if (intrapred)

  3750.                     s->ac_pred = get_bits1(gb);

  3751.                 else

  3752.                     s->ac_pred = 0;

  3753.             }

  3754.             if (!v->ttmbf && coded_inter)

  3755.                 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  3756.             for (i = 0; i < 6; i++) {

  3757.                 dst_idx    += i >> 2;

  3758.                 off         = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  3759.                 s->mb_intra = is_intra[i];

  3760.                 if (is_intra[i]) {

  3761.                     /* check if prediction blocks A and C are available */

  3762.                     v->a_avail = v->c_avail = 0;

  3763.                     if (i == 2 || i == 3 || !s->first_slice_line)

  3764.                         v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  3765.                     if (i == 1 || i == 3 || s->mb_x)

  3766.                         v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  3767. 

  3768.                     vc1_decode_intra_block(v, s->block[i], i, is_coded[i], mquant,

  3769.                                            (i & 4) ? v->codingset2 : v->codingset);

  3770.                     if ((i>3) && (s->flags & CODEC_FLAG_GRAY))

  3771.                         continue;

  3772.                     v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);

  3773.                     if (v->rangeredfrm)

  3774.                         for (j = 0; j < 64; j++)

  3775.                             s->block[i][j] <<= 1;

  3776.                     s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off,

  3777.                                                      (i & 4) ? s->uvlinesize : s->linesize);

  3778.                     if (v->pq >= 9 && v->overlap) {

  3779.                         if (v->c_avail)

  3780.                             v->vc1dsp.vc1_h_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);

  3781.                         if (v->a_avail)

  3782.                             v->vc1dsp.vc1_v_overlap(s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);

  3783.                     }

  3784.                     block_cbp   |= 0xF << (i << 2);

  3785.                     block_intra |= 1 << i;

  3786.                 } else if (is_coded[i]) {

  3787.                     pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,

  3788.                                              first_block, s->dest[dst_idx] + off,

  3789.                                              (i & 4) ? s->uvlinesize : s->linesize,

  3790.                                              (i & 4) && (s->flags & CODEC_FLAG_GRAY),

  3791.                                              &block_tt);

  3792.                     block_cbp |= pat << (i << 2);

  3793.                     if (!v->ttmbf && ttmb < 8)

  3794.                         ttmb = -1;

  3795.                     first_block = 0;

  3796.                 }

  3797.             }

  3798.         } else { // skipped MB

  3799.             s->mb_intra                               = 0;

  3800.             s->current_picture.f.qscale_table[mb_pos] = 0;

  3801.             for (i = 0; i < 6; i++) {

  3802.                 v->mb_type[0][s->block_index[i]] = 0;

  3803.                 s->dc_val[0][s->block_index[i]]  = 0;

  3804.             }

  3805.             for (i = 0; i < 4; i++) {

  3806.                 vc1_pred_mv(v, i, 0, 0, 0, v->range_x, v->range_y, v->mb_type[0], 0, 0);

  3807.                 vc1_mc_4mv_luma(v, i, 0);

  3808.             }

  3809.             vc1_mc_4mv_chroma(v, 0);

  3810.             s->current_picture.f.qscale_table[mb_pos] = 0;

  3811.         }

  3812.     }

  3813. end:

  3814.     v->cbp[s->mb_x]      = block_cbp;

  3815.     v->ttblk[s->mb_x]    = block_tt;

  3816.     v->is_intra[s->mb_x] = block_intra;

  3817. 

  3818.     return 0;

  3819. }

  3820. 

  3821. /* Decode one macroblock in an interlaced frame p picture */

  3822. 

  3823. static int vc1_decode_p_mb_intfr(VC1Context *v)

  3824. {

  3825.     MpegEncContext *s = &v->s;

  3826.     GetBitContext *gb = &s->gb;

  3827.     int i;

  3828.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  3829.     int cbp = 0; /* cbp decoding stuff */

  3830.     int mqdiff, mquant; /* MB quantization */

  3831.     int ttmb = v->ttfrm; /* MB Transform type */

  3832. 

  3833.     int mb_has_coeffs = 1; /* last_flag */

  3834.     int dmv_x, dmv_y; /* Differential MV components */

  3835.     int val; /* temp value */

  3836.     int first_block = 1;

  3837.     int dst_idx, off;

  3838.     int skipped, fourmv = 0, twomv = 0;

  3839.     int block_cbp = 0, pat, block_tt = 0;

  3840.     int idx_mbmode = 0, mvbp;

  3841.     int stride_y, fieldtx;

  3842. 

  3843.     mquant = v->pq; /* Loosy initialization */

  3844. 

  3845.     if (v->skip_is_raw)

  3846.         skipped = get_bits1(gb);

  3847.     else

  3848.         skipped = v->s.mbskip_table[mb_pos];

  3849.     if (!skipped) {

  3850.         if (v->fourmvswitch)

  3851.             idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_INTFR_4MV_MBMODE_VLC_BITS, 2); // try getting this done

  3852.         else

  3853.             idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_INTFR_NON4MV_MBMODE_VLC_BITS, 2); // in a single line

  3854.         switch (ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0]) {

  3855.         /* store the motion vector type in a flag (useful later) */

  3856.         case MV_PMODE_INTFR_4MV:

  3857.             fourmv = 1;

  3858.             v->blk_mv_type[s->block_index[0]] = 0;

  3859.             v->blk_mv_type[s->block_index[1]] = 0;

  3860.             v->blk_mv_type[s->block_index[2]] = 0;

  3861.             v->blk_mv_type[s->block_index[3]] = 0;

  3862.             break;

  3863.         case MV_PMODE_INTFR_4MV_FIELD:

  3864.             fourmv = 1;

  3865.             v->blk_mv_type[s->block_index[0]] = 1;

  3866.             v->blk_mv_type[s->block_index[1]] = 1;

  3867.             v->blk_mv_type[s->block_index[2]] = 1;

  3868.             v->blk_mv_type[s->block_index[3]] = 1;

  3869.             break;

  3870.         case MV_PMODE_INTFR_2MV_FIELD:

  3871.             twomv = 1;

  3872.             v->blk_mv_type[s->block_index[0]] = 1;

  3873.             v->blk_mv_type[s->block_index[1]] = 1;

  3874.             v->blk_mv_type[s->block_index[2]] = 1;

  3875.             v->blk_mv_type[s->block_index[3]] = 1;

  3876.             break;

  3877.         case MV_PMODE_INTFR_1MV:

  3878.             v->blk_mv_type[s->block_index[0]] = 0;

  3879.             v->blk_mv_type[s->block_index[1]] = 0;

  3880.             v->blk_mv_type[s->block_index[2]] = 0;

  3881.             v->blk_mv_type[s->block_index[3]] = 0;

  3882.             break;

  3883.         }

  3884.         if (ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0] == MV_PMODE_INTFR_INTRA) { // intra MB

  3885.             s->current_picture.f.motion_val[1][s->block_index[0]][0] = 0;

  3886.             s->current_picture.f.motion_val[1][s->block_index[0]][1] = 0;

  3887.             s->current_picture.f.mb_type[mb_pos]                     = MB_TYPE_INTRA;

  3888.             s->mb_intra = v->is_intra[s->mb_x] = 1;

  3889.             for (i = 0; i < 6; i++)

  3890.                 v->mb_type[0][s->block_index[i]] = 1;

  3891.             fieldtx = v->fieldtx_plane[mb_pos] = get_bits1(gb);

  3892.             mb_has_coeffs = get_bits1(gb);

  3893.             if (mb_has_coeffs)

  3894.                 cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  3895.             v->s.ac_pred = v->acpred_plane[mb_pos] = get_bits1(gb);

  3896.             GET_MQUANT();

  3897.             s->current_picture.f.qscale_table[mb_pos] = mquant;

  3898.             /* Set DC scale - y and c use the same (not sure if necessary here) */

  3899.             s->y_dc_scale = s->y_dc_scale_table[mquant];

  3900.             s->c_dc_scale = s->c_dc_scale_table[mquant];

  3901.             dst_idx = 0;

  3902.             for (i = 0; i < 6; i++) {

  3903.                 s->dc_val[0][s->block_index[i]] = 0;

  3904.                 dst_idx += i >> 2;

  3905.                 val = ((cbp >> (5 - i)) & 1);

  3906.                 v->mb_type[0][s->block_index[i]] = s->mb_intra;

  3907.                 v->a_avail = v->c_avail = 0;

  3908.                 if (i == 2 || i == 3 || !s->first_slice_line)

  3909.                     v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  3910.                 if (i == 1 || i == 3 || s->mb_x)

  3911.                     v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  3912. 

  3913.                 vc1_decode_intra_block(v, s->block[i], i, val, mquant,

  3914.                                        (i & 4) ? v->codingset2 : v->codingset);

  3915.                 if ((i>3) && (s->flags & CODEC_FLAG_GRAY)) continue;

  3916.                 v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);

  3917.                 if (i < 4) {

  3918.                     stride_y = s->linesize << fieldtx;

  3919.                     off = (fieldtx) ? ((i & 1) * 8) + ((i & 2) >> 1) * s->linesize : (i & 1) * 8 + 4 * (i & 2) * s->linesize;

  3920.                 } else {

  3921.                     stride_y = s->uvlinesize;

  3922.                     off = 0;

  3923.                 }

  3924.                 s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, stride_y);

  3925.                 //TODO: loop filter

  3926.             }

  3927. 

  3928.         } else { // inter MB

  3929.             mb_has_coeffs = ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][3];

  3930.             if (mb_has_coeffs)

  3931.                 cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  3932.             if (ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0] == MV_PMODE_INTFR_2MV_FIELD) {

  3933.                 v->twomvbp = get_vlc2(gb, v->twomvbp_vlc->table, VC1_2MV_BLOCK_PATTERN_VLC_BITS, 1);

  3934.             } else {

  3935.                 if ((ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0] == MV_PMODE_INTFR_4MV)

  3936.                     || (ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][0] == MV_PMODE_INTFR_4MV_FIELD)) {

  3937.                     v->fourmvbp = get_vlc2(gb, v->fourmvbp_vlc->table, VC1_4MV_BLOCK_PATTERN_VLC_BITS, 1);

  3938.                 }

  3939.             }

  3940.             s->mb_intra = v->is_intra[s->mb_x] = 0;

  3941.             for (i = 0; i < 6; i++)

  3942.                 v->mb_type[0][s->block_index[i]] = 0;

  3943.             fieldtx = v->fieldtx_plane[mb_pos] = ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][1];

  3944.             /* for all motion vector read MVDATA and motion compensate each block */

  3945.             dst_idx = 0;

  3946.             if (fourmv) {

  3947.                 mvbp = v->fourmvbp;

  3948.                 for (i = 0; i < 6; i++) {

  3949.                     if (i < 4) {

  3950.                         dmv_x = dmv_y = 0;

  3951.                         val   = ((mvbp >> (3 - i)) & 1);

  3952.                         if (val) {

  3953.                             get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);

  3954.                         }

  3955.                         vc1_pred_mv_intfr(v, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0]);

  3956.                         vc1_mc_4mv_luma(v, i, 0);

  3957.                     } else if (i == 4) {

  3958.                         vc1_mc_4mv_chroma4(v);

  3959.                     }

  3960.                 }

  3961.             } else if (twomv) {

  3962.                 mvbp  = v->twomvbp;

  3963.                 dmv_x = dmv_y = 0;

  3964.                 if (mvbp & 2) {

  3965.                     get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);

  3966.                 }

  3967.                 vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0]);

  3968.                 vc1_mc_4mv_luma(v, 0, 0);

  3969.                 vc1_mc_4mv_luma(v, 1, 0);

  3970.                 dmv_x = dmv_y = 0;

  3971.                 if (mvbp & 1) {

  3972.                     get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);

  3973.                 }

  3974.                 vc1_pred_mv_intfr(v, 2, dmv_x, dmv_y, 2, v->range_x, v->range_y, v->mb_type[0]);

  3975.                 vc1_mc_4mv_luma(v, 2, 0);

  3976.                 vc1_mc_4mv_luma(v, 3, 0);

  3977.                 vc1_mc_4mv_chroma4(v);

  3978.             } else {

  3979.                 mvbp = ff_vc1_mbmode_intfrp[v->fourmvswitch][idx_mbmode][2];

  3980.                 if (mvbp) {

  3981.                     get_mvdata_interlaced(v, &dmv_x, &dmv_y, 0);

  3982.                 }

  3983.                 vc1_pred_mv_intfr(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0]);

  3984.                 vc1_mc_1mv(v, 0);

  3985.             }

  3986.             if (cbp)

  3987.                 GET_MQUANT();  // p. 227

  3988.             s->current_picture.f.qscale_table[mb_pos] = mquant;

  3989.             if (!v->ttmbf && cbp)

  3990.                 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  3991.             for (i = 0; i < 6; i++) {

  3992.                 s->dc_val[0][s->block_index[i]] = 0;

  3993.                 dst_idx += i >> 2;

  3994.                 val = ((cbp >> (5 - i)) & 1);

  3995.                 if (!fieldtx)

  3996.                     off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  3997.                 else

  3998.                     off = (i & 4) ? 0 : ((i & 1) * 8 + ((i > 1) * s->linesize));

  3999.                 if (val) {

  4000.                     pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,

  4001.                                              first_block, s->dest[dst_idx] + off,

  4002.                                              (i & 4) ? s->uvlinesize : (s->linesize << fieldtx),

  4003.                                              (i & 4) && (s->flags & CODEC_FLAG_GRAY), &block_tt);

  4004.                     block_cbp |= pat << (i << 2);

  4005.                     if (!v->ttmbf && ttmb < 8)

  4006.                         ttmb = -1;

  4007.                     first_block = 0;

  4008.                 }

  4009.             }

  4010.         }

  4011.     } else { // skipped

  4012.         s->mb_intra = v->is_intra[s->mb_x] = 0;

  4013.         for (i = 0; i < 6; i++) {

  4014.             v->mb_type[0][s->block_index[i]] = 0;

  4015.             s->dc_val[0][s->block_index[i]] = 0;

  4016.         }

  4017.         s->current_picture.f.mb_type[mb_pos]      = MB_TYPE_SKIP;

  4018.         s->current_picture.f.qscale_table[mb_pos] = 0;

  4019.         v->blk_mv_type[s->block_index[0]] = 0;

  4020.         v->blk_mv_type[s->block_index[1]] = 0;

  4021.         v->blk_mv_type[s->block_index[2]] = 0;

  4022.         v->blk_mv_type[s->block_index[3]] = 0;

  4023.         vc1_pred_mv_intfr(v, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0]);

  4024.         vc1_mc_1mv(v, 0);

  4025.     }

  4026.     if (s->mb_x == s->mb_width - 1)

  4027.         memmove(v->is_intra_base, v->is_intra, sizeof(v->is_intra_base[0])*s->mb_stride);

  4028.     return 0;

  4029. }

  4030. 

  4031. static int vc1_decode_p_mb_intfi(VC1Context *v)

  4032. {

  4033.     MpegEncContext *s = &v->s;

  4034.     GetBitContext *gb = &s->gb;

  4035.     int i;

  4036.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  4037.     int cbp = 0; /* cbp decoding stuff */

  4038.     int mqdiff, mquant; /* MB quantization */

  4039.     int ttmb = v->ttfrm; /* MB Transform type */

  4040. 

  4041.     int mb_has_coeffs = 1; /* last_flag */

  4042.     int dmv_x, dmv_y; /* Differential MV components */

  4043.     int val; /* temp values */

  4044.     int first_block = 1;

  4045.     int dst_idx, off;

  4046.     int pred_flag;

  4047.     int block_cbp = 0, pat, block_tt = 0;

  4048.     int idx_mbmode = 0;

  4049. 

  4050.     mquant = v->pq; /* Loosy initialization */

  4051. 

  4052.     idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_IF_MBMODE_VLC_BITS, 2);

  4053.     if (idx_mbmode <= 1) { // intra MB

  4054.         s->mb_intra = v->is_intra[s->mb_x] = 1;

  4055.         s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0] = 0;

  4056.         s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1] = 0;

  4057.         s->current_picture.f.mb_type[mb_pos + v->mb_off] = MB_TYPE_INTRA;

  4058.         GET_MQUANT();

  4059.         s->current_picture.f.qscale_table[mb_pos] = mquant;

  4060.         /* Set DC scale - y and c use the same (not sure if necessary here) */

  4061.         s->y_dc_scale = s->y_dc_scale_table[mquant];

  4062.         s->c_dc_scale = s->c_dc_scale_table[mquant];

  4063.         v->s.ac_pred  = v->acpred_plane[mb_pos] = get_bits1(gb);

  4064.         mb_has_coeffs = idx_mbmode & 1;

  4065.         if (mb_has_coeffs)

  4066.             cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_ICBPCY_VLC_BITS, 2);

  4067.         dst_idx = 0;

  4068.         for (i = 0; i < 6; i++) {

  4069.             s->dc_val[0][s->block_index[i]]  = 0;

  4070.             v->mb_type[0][s->block_index[i]] = 1;

  4071.             dst_idx += i >> 2;

  4072.             val = ((cbp >> (5 - i)) & 1);

  4073.             v->a_avail = v->c_avail = 0;

  4074.             if (i == 2 || i == 3 || !s->first_slice_line)

  4075.                 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  4076.             if (i == 1 || i == 3 || s->mb_x)

  4077.                 v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  4078. 

  4079.             vc1_decode_intra_block(v, s->block[i], i, val, mquant,

  4080.                                    (i & 4) ? v->codingset2 : v->codingset);

  4081.             if ((i>3) && (s->flags & CODEC_FLAG_GRAY))

  4082.                 continue;

  4083.             v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);

  4084.             off  = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  4085.             off += v->second_field ? ((i & 4) ? s->current_picture_ptr->f.linesize[1] : s->current_picture_ptr->f.linesize[0]) : 0;

  4086.             s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i & 4) ? s->uvlinesize : s->linesize);

  4087.             // TODO: loop filter

  4088.         }

  4089.     } else {

  4090.         s->mb_intra = v->is_intra[s->mb_x] = 0;

  4091.         s->current_picture.f.mb_type[mb_pos + v->mb_off] = MB_TYPE_16x16;

  4092.         for (i = 0; i < 6; i++) v->mb_type[0][s->block_index[i]] = 0;

  4093.         if (idx_mbmode <= 5) { // 1-MV

  4094.             dmv_x = dmv_y = 0;

  4095.             if (idx_mbmode & 1) {

  4096.                 get_mvdata_interlaced(v, &dmv_x, &dmv_y, &pred_flag);

  4097.             }

  4098.             vc1_pred_mv(v, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0], pred_flag, 0);

  4099.             vc1_mc_1mv(v, 0);

  4100.             mb_has_coeffs = !(idx_mbmode & 2);

  4101.         } else { // 4-MV

  4102.             v->fourmvbp = get_vlc2(gb, v->fourmvbp_vlc->table, VC1_4MV_BLOCK_PATTERN_VLC_BITS, 1);

  4103.             for (i = 0; i < 6; i++) {

  4104.                 if (i < 4) {

  4105.                     dmv_x = dmv_y = pred_flag = 0;

  4106.                     val   = ((v->fourmvbp >> (3 - i)) & 1);

  4107.                     if (val) {

  4108.                         get_mvdata_interlaced(v, &dmv_x, &dmv_y, &pred_flag);

  4109.                     }

  4110.                     vc1_pred_mv(v, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0], pred_flag, 0);

  4111.                     vc1_mc_4mv_luma(v, i, 0);

  4112.                 } else if (i == 4)

  4113.                     vc1_mc_4mv_chroma(v, 0);

  4114.             }

  4115.             mb_has_coeffs = idx_mbmode & 1;

  4116.         }

  4117.         if (mb_has_coeffs)

  4118.             cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  4119.         if (cbp) {

  4120.             GET_MQUANT();

  4121.         }

  4122.         s->current_picture.f.qscale_table[mb_pos] = mquant;

  4123.         if (!v->ttmbf && cbp) {

  4124.             ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  4125.         }

  4126.         dst_idx = 0;

  4127.         for (i = 0; i < 6; i++) {

  4128.             s->dc_val[0][s->block_index[i]] = 0;

  4129.             dst_idx += i >> 2;

  4130.             val = ((cbp >> (5 - i)) & 1);

  4131.             off = (i & 4) ? 0 : (i & 1) * 8 + (i & 2) * 4 * s->linesize;

  4132.             if (v->second_field)

  4133.                 off += (i & 4) ? s->current_picture_ptr->f.linesize[1] : s->current_picture_ptr->f.linesize[0];

  4134.             if (val) {

  4135.                 pat = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,

  4136.                                          first_block, s->dest[dst_idx] + off,

  4137.                                          (i & 4) ? s->uvlinesize : s->linesize,

  4138.                                          (i & 4) && (s->flags & CODEC_FLAG_GRAY),

  4139.                                          &block_tt);

  4140.                 block_cbp |= pat << (i << 2);

  4141.                 if (!v->ttmbf && ttmb < 8) ttmb = -1;

  4142.                 first_block = 0;

  4143.             }

  4144.         }

  4145.     }

  4146.     if (s->mb_x == s->mb_width - 1)

  4147.         memmove(v->is_intra_base, v->is_intra, sizeof(v->is_intra_base[0]) * s->mb_stride);

  4148.     return 0;

  4149. }

  4150. 

  4151. /** Decode one B-frame MB (in Main profile)

  4152.  */

  4153. static void vc1_decode_b_mb(VC1Context *v)

  4154. {

  4155.     MpegEncContext *s = &v->s;

  4156.     GetBitContext *gb = &s->gb;

  4157.     int i, j;

  4158.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  4159.     int cbp = 0; /* cbp decoding stuff */

  4160.     int mqdiff, mquant; /* MB quantization */

  4161.     int ttmb = v->ttfrm; /* MB Transform type */

  4162.     int mb_has_coeffs = 0; /* last_flag */

  4163.     int index, index1; /* LUT indexes */

  4164.     int val, sign; /* temp values */

  4165.     int first_block = 1;

  4166.     int dst_idx, off;

  4167.     int skipped, direct;

  4168.     int dmv_x[2], dmv_y[2];

  4169.     int bmvtype = BMV_TYPE_BACKWARD;

  4170. 

  4171.     mquant      = v->pq; /* lossy initialization */

  4172.     s->mb_intra = 0;

  4173. 

  4174.     if (v->dmb_is_raw)

  4175.         direct = get_bits1(gb);

  4176.     else

  4177.         direct = v->direct_mb_plane[mb_pos];

  4178.     if (v->skip_is_raw)

  4179.         skipped = get_bits1(gb);

  4180.     else

  4181.         skipped = v->s.mbskip_table[mb_pos];

  4182. 

  4183.     dmv_x[0] = dmv_x[1] = dmv_y[0] = dmv_y[1] = 0;

  4184.     for (i = 0; i < 6; i++) {

  4185.         v->mb_type[0][s->block_index[i]] = 0;

  4186.         s->dc_val[0][s->block_index[i]]  = 0;

  4187.     }

  4188.     s->current_picture.f.qscale_table[mb_pos] = 0;

  4189. 

  4190.     if (!direct) {

  4191.         if (!skipped) {

  4192.             GET_MVDATA(dmv_x[0], dmv_y[0]);

  4193.             dmv_x[1] = dmv_x[0];

  4194.             dmv_y[1] = dmv_y[0];

  4195.         }

  4196.         if (skipped || !s->mb_intra) {

  4197.             bmvtype = decode012(gb);

  4198.             switch (bmvtype) {

  4199.             case 0:

  4200.                 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_BACKWARD : BMV_TYPE_FORWARD;

  4201.                 break;

  4202.             case 1:

  4203.                 bmvtype = (v->bfraction >= (B_FRACTION_DEN/2)) ? BMV_TYPE_FORWARD : BMV_TYPE_BACKWARD;

  4204.                 break;

  4205.             case 2:

  4206.                 bmvtype  = BMV_TYPE_INTERPOLATED;

  4207.                 dmv_x[0] = dmv_y[0] = 0;

  4208.             }

  4209.         }

  4210.     }

  4211.     for (i = 0; i < 6; i++)

  4212.         v->mb_type[0][s->block_index[i]] = s->mb_intra;

  4213. 

  4214.     if (skipped) {

  4215.         if (direct)

  4216.             bmvtype = BMV_TYPE_INTERPOLATED;

  4217.         vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  4218.         vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  4219.         return;

  4220.     }

  4221.     if (direct) {

  4222.         cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  4223.         GET_MQUANT();

  4224.         s->mb_intra = 0;

  4225.         s->current_picture.f.qscale_table[mb_pos] = mquant;

  4226.         if (!v->ttmbf)

  4227.             ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  4228.         dmv_x[0] = dmv_y[0] = dmv_x[1] = dmv_y[1] = 0;

  4229.         vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  4230.         vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  4231.     } else {

  4232.         if (!mb_has_coeffs && !s->mb_intra) {

  4233.             /* no coded blocks - effectively skipped */

  4234.             vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  4235.             vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  4236.             return;

  4237.         }

  4238.         if (s->mb_intra && !mb_has_coeffs) {

  4239.             GET_MQUANT();

  4240.             s->current_picture.f.qscale_table[mb_pos] = mquant;

  4241.             s->ac_pred = get_bits1(gb);

  4242.             cbp = 0;

  4243.             vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  4244.         } else {

  4245.             if (bmvtype == BMV_TYPE_INTERPOLATED) {

  4246.                 GET_MVDATA(dmv_x[0], dmv_y[0]);

  4247.                 if (!mb_has_coeffs) {

  4248.                     /* interpolated skipped block */

  4249.                     vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  4250.                     vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  4251.                     return;

  4252.                 }

  4253.             }

  4254.             vc1_pred_b_mv(v, dmv_x, dmv_y, direct, bmvtype);

  4255.             if (!s->mb_intra) {

  4256.                 vc1_b_mc(v, dmv_x, dmv_y, direct, bmvtype);

  4257.             }

  4258.             if (s->mb_intra)

  4259.                 s->ac_pred = get_bits1(gb);

  4260.             cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  4261.             GET_MQUANT();

  4262.             s->current_picture.f.qscale_table[mb_pos] = mquant;

  4263.             if (!v->ttmbf && !s->mb_intra && mb_has_coeffs)

  4264.                 ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  4265.         }

  4266.     }

  4267.     dst_idx = 0;

  4268.     for (i = 0; i < 6; i++) {

  4269.         s->dc_val[0][s->block_index[i]] = 0;

  4270.         dst_idx += i >> 2;

  4271.         val = ((cbp >> (5 - i)) & 1);

  4272.         off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  4273.         v->mb_type[0][s->block_index[i]] = s->mb_intra;

  4274.         if (s->mb_intra) {

  4275.             /* check if prediction blocks A and C are available */

  4276.             v->a_avail = v->c_avail = 0;

  4277.             if (i == 2 || i == 3 || !s->first_slice_line)

  4278.                 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  4279.             if (i == 1 || i == 3 || s->mb_x)

  4280.                 v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  4281. 

  4282.             vc1_decode_intra_block(v, s->block[i], i, val, mquant,

  4283.                                    (i & 4) ? v->codingset2 : v->codingset);

  4284.             if ((i>3) && (s->flags & CODEC_FLAG_GRAY))

  4285.                 continue;

  4286.             v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);

  4287.             if (v->rangeredfrm)

  4288.                 for (j = 0; j < 64; j++)

  4289.                     s->block[i][j] <<= 1;

  4290.             s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, i & 4 ? s->uvlinesize : s->linesize);

  4291.         } else if (val) {

  4292.             vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,

  4293.                                first_block, s->dest[dst_idx] + off,

  4294.                                (i & 4) ? s->uvlinesize : s->linesize,

  4295.                                (i & 4) && (s->flags & CODEC_FLAG_GRAY), NULL);

  4296.             if (!v->ttmbf && ttmb < 8)

  4297.                 ttmb = -1;

  4298.             first_block = 0;

  4299.         }

  4300.     }

  4301. }

  4302. 

  4303. /** Decode one B-frame MB (in interlaced field B picture)

  4304.  */

  4305. static void vc1_decode_b_mb_intfi(VC1Context *v)

  4306. {

  4307.     MpegEncContext *s = &v->s;

  4308.     GetBitContext *gb = &s->gb;

  4309.     int i, j;

  4310.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  4311.     int cbp = 0; /* cbp decoding stuff */

  4312.     int mqdiff, mquant; /* MB quantization */

  4313.     int ttmb = v->ttfrm; /* MB Transform type */

  4314.     int mb_has_coeffs = 0; /* last_flag */

  4315.     int val; /* temp value */

  4316.     int first_block = 1;

  4317.     int dst_idx, off;

  4318.     int fwd;

  4319.     int dmv_x[2], dmv_y[2], pred_flag[2];

  4320.     int bmvtype = BMV_TYPE_BACKWARD;

  4321.     int idx_mbmode, interpmvp;

  4322. 

  4323.     mquant      = v->pq; /* Loosy initialization */

  4324.     s->mb_intra = 0;

  4325. 

  4326.     idx_mbmode = get_vlc2(gb, v->mbmode_vlc->table, VC1_IF_MBMODE_VLC_BITS, 2);

  4327.     if (idx_mbmode <= 1) { // intra MB

  4328.         s->mb_intra = v->is_intra[s->mb_x] = 1;

  4329.         s->current_picture.f.motion_val[1][s->block_index[0]][0] = 0;

  4330.         s->current_picture.f.motion_val[1][s->block_index[0]][1] = 0;

  4331.         s->current_picture.f.mb_type[mb_pos + v->mb_off]         = MB_TYPE_INTRA;

  4332.         GET_MQUANT();

  4333.         s->current_picture.f.qscale_table[mb_pos] = mquant;

  4334.         /* Set DC scale - y and c use the same (not sure if necessary here) */

  4335.         s->y_dc_scale = s->y_dc_scale_table[mquant];

  4336.         s->c_dc_scale = s->c_dc_scale_table[mquant];

  4337.         v->s.ac_pred  = v->acpred_plane[mb_pos] = get_bits1(gb);

  4338.         mb_has_coeffs = idx_mbmode & 1;

  4339.         if (mb_has_coeffs)

  4340.             cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_ICBPCY_VLC_BITS, 2);

  4341.         dst_idx = 0;

  4342.         for (i = 0; i < 6; i++) {

  4343.             s->dc_val[0][s->block_index[i]] = 0;

  4344.             dst_idx += i >> 2;

  4345.             val = ((cbp >> (5 - i)) & 1);

  4346.             v->mb_type[0][s->block_index[i]] = s->mb_intra;

  4347.             v->a_avail                       = v->c_avail = 0;

  4348.             if (i == 2 || i == 3 || !s->first_slice_line)

  4349.                 v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  4350.             if (i == 1 || i == 3 || s->mb_x)

  4351.                 v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  4352. 

  4353.             vc1_decode_intra_block(v, s->block[i], i, val, mquant,

  4354.                                    (i & 4) ? v->codingset2 : v->codingset);

  4355.             if ((i>3) && (s->flags & CODEC_FLAG_GRAY))

  4356.                 continue;

  4357.             v->vc1dsp.vc1_inv_trans_8x8(s->block[i]);

  4358.             if (v->rangeredfrm)

  4359.                 for (j = 0; j < 64; j++)

  4360.                     s->block[i][j] <<= 1;

  4361.             off  = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  4362.             off += v->second_field ? ((i & 4) ? s->current_picture_ptr->f.linesize[1] : s->current_picture_ptr->f.linesize[0]) : 0;

  4363.             s->dsp.put_signed_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i & 4) ? s->uvlinesize : s->linesize);

  4364.             // TODO: yet to perform loop filter

  4365.         }

  4366.     } else {

  4367.         s->mb_intra = v->is_intra[s->mb_x] = 0;

  4368.         s->current_picture.f.mb_type[mb_pos + v->mb_off] = MB_TYPE_16x16;

  4369.         for (i = 0; i < 6; i++) v->mb_type[0][s->block_index[i]] = 0;

  4370.         if (v->fmb_is_raw)

  4371.             fwd = v->forward_mb_plane[mb_pos] = get_bits1(gb);

  4372.         else

  4373.             fwd = v->forward_mb_plane[mb_pos];

  4374.         if (idx_mbmode <= 5) { // 1-MV

  4375.             dmv_x[0]     = dmv_x[1] = dmv_y[0] = dmv_y[1] = 0;

  4376.             pred_flag[0] = pred_flag[1] = 0;

  4377.             if (fwd)

  4378.                 bmvtype = BMV_TYPE_FORWARD;

  4379.             else {

  4380.                 bmvtype = decode012(gb);

  4381.                 switch (bmvtype) {

  4382.                 case 0:

  4383.                     bmvtype = BMV_TYPE_BACKWARD;

  4384.                     break;

  4385.                 case 1:

  4386.                     bmvtype = BMV_TYPE_DIRECT;

  4387.                     break;

  4388.                 case 2:

  4389.                     bmvtype   = BMV_TYPE_INTERPOLATED;

  4390.                     interpmvp = get_bits1(gb);

  4391.                 }

  4392.             }

  4393.             v->bmvtype = bmvtype;

  4394.             if (bmvtype != BMV_TYPE_DIRECT && idx_mbmode & 1) {

  4395.                 get_mvdata_interlaced(v, &dmv_x[bmvtype == BMV_TYPE_BACKWARD], &dmv_y[bmvtype == BMV_TYPE_BACKWARD], &pred_flag[bmvtype == BMV_TYPE_BACKWARD]);

  4396.             }

  4397.             if (bmvtype == BMV_TYPE_INTERPOLATED && interpmvp) {

  4398.                 get_mvdata_interlaced(v, &dmv_x[1], &dmv_y[1], &pred_flag[1]);

  4399.             }

  4400.             if (bmvtype == BMV_TYPE_DIRECT) {

  4401.                 dmv_x[0] = dmv_y[0] = pred_flag[0] = 0;

  4402.                 dmv_x[1] = dmv_y[1] = pred_flag[0] = 0;

  4403.             }

  4404.             vc1_pred_b_mv_intfi(v, 0, dmv_x, dmv_y, 1, pred_flag);

  4405.             vc1_b_mc(v, dmv_x, dmv_y, (bmvtype == BMV_TYPE_DIRECT), bmvtype);

  4406.             mb_has_coeffs = !(idx_mbmode & 2);

  4407.         } else { // 4-MV

  4408.             if (fwd)

  4409.                 bmvtype = BMV_TYPE_FORWARD;

  4410.             v->bmvtype  = bmvtype;

  4411.             v->fourmvbp = get_vlc2(gb, v->fourmvbp_vlc->table, VC1_4MV_BLOCK_PATTERN_VLC_BITS, 1);

  4412.             for (i = 0; i < 6; i++) {

  4413.                 if (i < 4) {

  4414.                     dmv_x[0] = dmv_y[0] = pred_flag[0] = 0;

  4415.                     dmv_x[1] = dmv_y[1] = pred_flag[1] = 0;

  4416.                     val = ((v->fourmvbp >> (3 - i)) & 1);

  4417.                     if (val) {

  4418.                         get_mvdata_interlaced(v, &dmv_x[bmvtype == BMV_TYPE_BACKWARD],

  4419.                                                  &dmv_y[bmvtype == BMV_TYPE_BACKWARD],

  4420.                                              &pred_flag[bmvtype == BMV_TYPE_BACKWARD]);

  4421.                     }

  4422.                     vc1_pred_b_mv_intfi(v, i, dmv_x, dmv_y, 0, pred_flag);

  4423.                     vc1_mc_4mv_luma(v, i, bmvtype == BMV_TYPE_BACKWARD);

  4424.                 } else if (i == 4)

  4425.                     vc1_mc_4mv_chroma(v, bmvtype == BMV_TYPE_BACKWARD);

  4426.             }

  4427.             mb_has_coeffs = idx_mbmode & 1;

  4428.         }

  4429.         if (mb_has_coeffs)

  4430.             cbp = 1 + get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  4431.         if (cbp) {

  4432.             GET_MQUANT();

  4433.         }

  4434.         s->current_picture.f.qscale_table[mb_pos] = mquant;

  4435.         if (!v->ttmbf && cbp) {

  4436.             ttmb = get_vlc2(gb, ff_vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 2);

  4437.         }

  4438.         dst_idx = 0;

  4439.         for (i = 0; i < 6; i++) {

  4440.             s->dc_val[0][s->block_index[i]] = 0;

  4441.             dst_idx += i >> 2;

  4442.             val = ((cbp >> (5 - i)) & 1);

  4443.             off = (i & 4) ? 0 : (i & 1) * 8 + (i & 2) * 4 * s->linesize;

  4444.             if (v->second_field)

  4445.                 off += (i & 4) ? s->current_picture_ptr->f.linesize[1] : s->current_picture_ptr->f.linesize[0];

  4446.             if (val) {

  4447.                 vc1_decode_p_block(v, s->block[i], i, mquant, ttmb,

  4448.                                    first_block, s->dest[dst_idx] + off,

  4449.                                    (i & 4) ? s->uvlinesize : s->linesize,

  4450.                                    (i & 4) && (s->flags & CODEC_FLAG_GRAY), NULL);

  4451.                 if (!v->ttmbf && ttmb < 8)

  4452.                     ttmb = -1;

  4453.                 first_block = 0;

  4454.             }

  4455.         }

  4456.     }

  4457. }

  4458. 

  4459. /** Decode blocks of I-frame

  4460.  */

  4461. static void vc1_decode_i_blocks(VC1Context *v)

  4462. {

  4463.     int k, j;

  4464.     MpegEncContext *s = &v->s;

  4465.     int cbp, val;

  4466.     uint8_t *coded_val;

  4467.     int mb_pos;

  4468. 

  4469.     /* select codingmode used for VLC tables selection */

  4470.     switch (v->y_ac_table_index) {

  4471.     case 0:

  4472.         v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;

  4473.         break;

  4474.     case 1:

  4475.         v->codingset = CS_HIGH_MOT_INTRA;

  4476.         break;

  4477.     case 2:

  4478.         v->codingset = CS_MID_RATE_INTRA;

  4479.         break;

  4480.     }

  4481. 

  4482.     switch (v->c_ac_table_index) {

  4483.     case 0:

  4484.         v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;

  4485.         break;

  4486.     case 1:

  4487.         v->codingset2 = CS_HIGH_MOT_INTER;

  4488.         break;

  4489.     case 2:

  4490.         v->codingset2 = CS_MID_RATE_INTER;

  4491.         break;

  4492.     }

  4493. 

  4494.     /* Set DC scale - y and c use the same */

  4495.     s->y_dc_scale = s->y_dc_scale_table[v->pq];

  4496.     s->c_dc_scale = s->c_dc_scale_table[v->pq];

  4497. 

  4498.     //do frame decode

  4499.     s->mb_x = s->mb_y = 0;

  4500.     s->mb_intra         = 1;

  4501.     s->first_slice_line = 1;

  4502.     for (s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {

  4503.         s->mb_x = 0;

  4504.         ff_init_block_index(s);

  4505.         for (; s->mb_x < s->mb_width; s->mb_x++) {

  4506.             uint8_t *dst[6];

  4507.             ff_update_block_index(s);

  4508.             dst[0] = s->dest[0];

  4509.             dst[1] = dst[0] + 8;

  4510.             dst[2] = s->dest[0] + s->linesize * 8;

  4511.             dst[3] = dst[2] + 8;

  4512.             dst[4] = s->dest[1];

  4513.             dst[5] = s->dest[2];

  4514.             s->dsp.clear_blocks(s->block[0]);

  4515.             mb_pos = s->mb_x + s->mb_y * s->mb_width;

  4516.             s->current_picture.f.mb_type[mb_pos]                     = MB_TYPE_INTRA;

  4517.             s->current_picture.f.qscale_table[mb_pos]                = v->pq;

  4518.             s->current_picture.f.motion_val[1][s->block_index[0]][0] = 0;

  4519.             s->current_picture.f.motion_val[1][s->block_index[0]][1] = 0;

  4520. 

  4521.             // do actual MB decoding and displaying

  4522.             cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);

  4523.             v->s.ac_pred = get_bits1(&v->s.gb);

  4524. 

  4525.             for (k = 0; k < 6; k++) {

  4526.                 val = ((cbp >> (5 - k)) & 1);

  4527. 

  4528.                 if (k < 4) {

  4529.                     int pred   = vc1_coded_block_pred(&v->s, k, &coded_val);

  4530.                     val        = val ^ pred;

  4531.                     *coded_val = val;

  4532.                 }

  4533.                 cbp |= val << (5 - k);

  4534. 

  4535.                 vc1_decode_i_block(v, s->block[k], k, val, (k < 4) ? v->codingset : v->codingset2);

  4536. 

  4537.                 if (k > 3 && (s->flags & CODEC_FLAG_GRAY))

  4538.                     continue;

  4539.                 v->vc1dsp.vc1_inv_trans_8x8(s->block[k]);

  4540.                 if (v->pq >= 9 && v->overlap) {

  4541.                     if (v->rangeredfrm)

  4542.                         for (j = 0; j < 64; j++)

  4543.                             s->block[k][j] <<= 1;

  4544.                     s->dsp.put_signed_pixels_clamped(s->block[k], dst[k], k & 4 ? s->uvlinesize : s->linesize);

  4545.                 } else {

  4546.                     if (v->rangeredfrm)

  4547.                         for (j = 0; j < 64; j++)

  4548.                             s->block[k][j] = (s->block[k][j] - 64) << 1;

  4549.                     s->dsp.put_pixels_clamped(s->block[k], dst[k], k & 4 ? s->uvlinesize : s->linesize);

  4550.                 }

  4551.             }

  4552. 

  4553.             if (v->pq >= 9 && v->overlap) {

  4554.                 if (s->mb_x) {

  4555.                     v->vc1dsp.vc1_h_overlap(s->dest[0], s->linesize);

  4556.                     v->vc1dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);

  4557.                     if (!(s->flags & CODEC_FLAG_GRAY)) {

  4558.                         v->vc1dsp.vc1_h_overlap(s->dest[1], s->uvlinesize);

  4559.                         v->vc1dsp.vc1_h_overlap(s->dest[2], s->uvlinesize);

  4560.                     }

  4561.                 }

  4562.                 v->vc1dsp.vc1_h_overlap(s->dest[0] + 8, s->linesize);

  4563.                 v->vc1dsp.vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);

  4564.                 if (!s->first_slice_line) {

  4565.                     v->vc1dsp.vc1_v_overlap(s->dest[0], s->linesize);

  4566.                     v->vc1dsp.vc1_v_overlap(s->dest[0] + 8, s->linesize);

  4567.                     if (!(s->flags & CODEC_FLAG_GRAY)) {

  4568.                         v->vc1dsp.vc1_v_overlap(s->dest[1], s->uvlinesize);

  4569.                         v->vc1dsp.vc1_v_overlap(s->dest[2], s->uvlinesize);

  4570.                     }

  4571.                 }

  4572.                 v->vc1dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);

  4573.                 v->vc1dsp.vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);

  4574.             }

  4575.             if (v->s.loop_filter) vc1_loop_filter_iblk(v, v->pq);

  4576. 

  4577.             if (get_bits_count(&s->gb) > v->bits) {

  4578.                 ff_er_add_slice(s, 0, 0, s->mb_x, s->mb_y, ER_MB_ERROR);

  4579.                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n",

  4580.                        get_bits_count(&s->gb), v->bits);

  4581.                 return;

  4582.             }

  4583.         }

  4584.         if (!v->s.loop_filter)

  4585.             ff_draw_horiz_band(s, s->mb_y * 16, 16);

  4586.         else if (s->mb_y)

  4587.             ff_draw_horiz_band(s, (s->mb_y - 1) * 16, 16);

  4588. 

  4589.         s->first_slice_line = 0;

  4590.     }

  4591.     if (v->s.loop_filter)

  4592.         ff_draw_horiz_band(s, (s->mb_height - 1) * 16, 16);

  4593.     ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, ER_MB_END);

  4594. }

  4595. 

  4596. /** Decode blocks of I-frame for advanced profile

  4597.  */

  4598. static void vc1_decode_i_blocks_adv(VC1Context *v)

  4599. {

  4600.     int k;

  4601.     MpegEncContext *s = &v->s;

  4602.     int cbp, val;

  4603.     uint8_t *coded_val;

  4604.     int mb_pos;

  4605.     int mquant = v->pq;

  4606.     int mqdiff;

  4607.     GetBitContext *gb = &s->gb;

  4608. 

  4609.     /* select codingmode used for VLC tables selection */

  4610.     switch (v->y_ac_table_index) {

  4611.     case 0:

  4612.         v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;

  4613.         break;

  4614.     case 1:

  4615.         v->codingset = CS_HIGH_MOT_INTRA;

  4616.         break;

  4617.     case 2:

  4618.         v->codingset = CS_MID_RATE_INTRA;

  4619.         break;

  4620.     }

  4621. 

  4622.     switch (v->c_ac_table_index) {

  4623.     case 0:

  4624.         v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;

  4625.         break;

  4626.     case 1:

  4627.         v->codingset2 = CS_HIGH_MOT_INTER;

  4628.         break;

  4629.     case 2:

  4630.         v->codingset2 = CS_MID_RATE_INTER;

  4631.         break;

  4632.     }

  4633. 

  4634.     // do frame decode

  4635.     s->mb_x             = s->mb_y = 0;

  4636.     s->mb_intra         = 1;

  4637.     s->first_slice_line = 1;

  4638.     s->mb_y             = s->start_mb_y;

  4639.     if (s->start_mb_y) {

  4640.         s->mb_x = 0;

  4641.         ff_init_block_index(s);

  4642.         memset(&s->coded_block[s->block_index[0] - s->b8_stride], 0,

  4643.                (1 + s->b8_stride) * sizeof(*s->coded_block));

  4644.     }

  4645.     for (; s->mb_y < s->end_mb_y; s->mb_y++) {

  4646.         s->mb_x = 0;

  4647.         ff_init_block_index(s);

  4648.         for (;s->mb_x < s->mb_width; s->mb_x++) {

  4649.             DCTELEM (*block)[64] = v->block[v->cur_blk_idx];

  4650.             ff_update_block_index(s);

  4651.             s->dsp.clear_blocks(block[0]);

  4652.             mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  4653.             s->current_picture.f.mb_type[mb_pos + v->mb_off]                         = MB_TYPE_INTRA;

  4654.             s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][0] = 0;

  4655.             s->current_picture.f.motion_val[1][s->block_index[0] + v->blocks_off][1] = 0;

  4656. 

  4657.             // do actual MB decoding and displaying

  4658.             if (v->fieldtx_is_raw)

  4659.                 v->fieldtx_plane[mb_pos] = get_bits1(&v->s.gb);

  4660.             cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);

  4661.             if ( v->acpred_is_raw)

  4662.                 v->s.ac_pred = get_bits1(&v->s.gb);

  4663.             else

  4664.                 v->s.ac_pred = v->acpred_plane[mb_pos];

  4665. 

  4666.             if (v->condover == CONDOVER_SELECT && v->overflg_is_raw)

  4667.                 v->over_flags_plane[mb_pos] = get_bits1(&v->s.gb);

  4668. 

  4669.             GET_MQUANT();

  4670. 

  4671.             s->current_picture.f.qscale_table[mb_pos] = mquant;

  4672.             /* Set DC scale - y and c use the same */

  4673.             s->y_dc_scale = s->y_dc_scale_table[mquant];

  4674.             s->c_dc_scale = s->c_dc_scale_table[mquant];

  4675. 

  4676.             for (k = 0; k < 6; k++) {

  4677.                 val = ((cbp >> (5 - k)) & 1);

  4678. 

  4679.                 if (k < 4) {

  4680.                     int pred   = vc1_coded_block_pred(&v->s, k, &coded_val);

  4681.                     val        = val ^ pred;

  4682.                     *coded_val = val;

  4683.                 }

  4684.                 cbp |= val << (5 - k);

  4685. 

  4686.                 v->a_avail = !s->first_slice_line || (k == 2 || k == 3);

  4687.                 v->c_avail = !!s->mb_x || (k == 1 || k == 3);

  4688. 

  4689.                 vc1_decode_i_block_adv(v, block[k], k, val,

  4690.                                        (k < 4) ? v->codingset : v->codingset2, mquant);

  4691. 

  4692.                 if (k > 3 && (s->flags & CODEC_FLAG_GRAY))

  4693.                     continue;

  4694.                 v->vc1dsp.vc1_inv_trans_8x8(block[k]);

  4695.             }

  4696. 

  4697.             vc1_smooth_overlap_filter_iblk(v);

  4698.             vc1_put_signed_blocks_clamped(v);

  4699.             if (v->s.loop_filter) vc1_loop_filter_iblk_delayed(v, v->pq);

  4700. 

  4701.             if (get_bits_count(&s->gb) > v->bits) {

  4702.                 // TODO: may need modification to handle slice coding

  4703.                 ff_er_add_slice(s, 0, s->start_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR);

  4704.                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n",

  4705.                        get_bits_count(&s->gb), v->bits);

  4706.                 return;

  4707.             }

  4708.         }

  4709.         if (!v->s.loop_filter)

  4710.             ff_draw_horiz_band(s, s->mb_y * 16, 16);

  4711.         else if (s->mb_y)

  4712.             ff_draw_horiz_band(s, (s->mb_y-1) * 16, 16);

  4713.         s->first_slice_line = 0;

  4714.     }

  4715. 

  4716.     /* raw bottom MB row */

  4717.     s->mb_x = 0;

  4718.     ff_init_block_index(s);

  4719.     for (;s->mb_x < s->mb_width; s->mb_x++) {

  4720.         ff_update_block_index(s);

  4721.         vc1_put_signed_blocks_clamped(v);

  4722.         if (v->s.loop_filter)

  4723.             vc1_loop_filter_iblk_delayed(v, v->pq);

  4724.     }

  4725.     if (v->s.loop_filter)

  4726.         ff_draw_horiz_band(s, (s->end_mb_y-1)*16, 16);

  4727.     ff_er_add_slice(s, 0, s->start_mb_y << v->field_mode, s->mb_width - 1,

  4728.                     (s->end_mb_y << v->field_mode) - 1, ER_MB_END);

  4729. }

  4730. 

  4731. static void vc1_decode_p_blocks(VC1Context *v)

  4732. {

  4733.     MpegEncContext *s = &v->s;

  4734.     int apply_loop_filter;

  4735. 

  4736.     /* select codingmode used for VLC tables selection */

  4737.     switch (v->c_ac_table_index) {

  4738.     case 0:

  4739.         v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;

  4740.         break;

  4741.     case 1:

  4742.         v->codingset = CS_HIGH_MOT_INTRA;

  4743.         break;

  4744.     case 2:

  4745.         v->codingset = CS_MID_RATE_INTRA;

  4746.         break;

  4747.     }

  4748. 

  4749.     switch (v->c_ac_table_index) {

  4750.     case 0:

  4751.         v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;

  4752.         break;

  4753.     case 1:

  4754.         v->codingset2 = CS_HIGH_MOT_INTER;

  4755.         break;

  4756.     case 2:

  4757.         v->codingset2 = CS_MID_RATE_INTER;

  4758.         break;

  4759.     }

  4760. 

  4761.     apply_loop_filter   = s->loop_filter && !(s->avctx->skip_loop_filter >= AVDISCARD_NONKEY);

  4762.     s->first_slice_line = 1;

  4763.     memset(v->cbp_base, 0, sizeof(v->cbp_base[0])*2*s->mb_stride);

  4764.     for (s->mb_y = s->start_mb_y; s->mb_y < s->end_mb_y; s->mb_y++) {

  4765.         s->mb_x = 0;

  4766.         ff_init_block_index(s);

  4767.         for (; s->mb_x < s->mb_width; s->mb_x++) {

  4768.             ff_update_block_index(s);

  4769. 

  4770.             if (v->fcm == ILACE_FIELD)

  4771.                 vc1_decode_p_mb_intfi(v);

  4772.             else if (v->fcm == ILACE_FRAME)

  4773.                 vc1_decode_p_mb_intfr(v);

  4774.             else vc1_decode_p_mb(v);

  4775.             if (s->mb_y != s->start_mb_y && apply_loop_filter && v->fcm == PROGRESSIVE)

  4776.                 vc1_apply_p_loop_filter(v);

  4777.             if (get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {

  4778.                 // TODO: may need modification to handle slice coding

  4779.                 ff_er_add_slice(s, 0, s->start_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR);

  4780.                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i at %ix%i\n",

  4781.                        get_bits_count(&s->gb), v->bits, s->mb_x, s->mb_y);

  4782.                 return;

  4783.             }

  4784.         }

  4785.         memmove(v->cbp_base,      v->cbp,      sizeof(v->cbp_base[0])      * s->mb_stride);

  4786.         memmove(v->ttblk_base,    v->ttblk,    sizeof(v->ttblk_base[0])    * s->mb_stride);

  4787.         memmove(v->is_intra_base, v->is_intra, sizeof(v->is_intra_base[0]) * s->mb_stride);

  4788.         memmove(v->luma_mv_base,  v->luma_mv,  sizeof(v->luma_mv_base[0])  * s->mb_stride);

  4789.         if (s->mb_y != s->start_mb_y) ff_draw_horiz_band(s, (s->mb_y - 1) * 16, 16);

  4790.         s->first_slice_line = 0;

  4791.     }

  4792.     if (apply_loop_filter) {

  4793.         s->mb_x = 0;

  4794.         ff_init_block_index(s);

  4795.         for (; s->mb_x < s->mb_width; s->mb_x++) {

  4796.             ff_update_block_index(s);

  4797.             vc1_apply_p_loop_filter(v);

  4798.         }

  4799.     }

  4800.     if (s->end_mb_y >= s->start_mb_y)

  4801.         ff_draw_horiz_band(s, (s->end_mb_y - 1) * 16, 16);

  4802.     ff_er_add_slice(s, 0, s->start_mb_y << v->field_mode, s->mb_width - 1,

  4803.                     (s->end_mb_y << v->field_mode) - 1, ER_MB_END);

  4804. }

  4805. 

  4806. static void vc1_decode_b_blocks(VC1Context *v)

  4807. {

  4808.     MpegEncContext *s = &v->s;

  4809. 

  4810.     /* select codingmode used for VLC tables selection */

  4811.     switch (v->c_ac_table_index) {

  4812.     case 0:

  4813.         v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;

  4814.         break;

  4815.     case 1:

  4816.         v->codingset = CS_HIGH_MOT_INTRA;

  4817.         break;

  4818.     case 2:

  4819.         v->codingset = CS_MID_RATE_INTRA;

  4820.         break;

  4821.     }

  4822. 

  4823.     switch (v->c_ac_table_index) {

  4824.     case 0:

  4825.         v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;

  4826.         break;

  4827.     case 1:

  4828.         v->codingset2 = CS_HIGH_MOT_INTER;

  4829.         break;

  4830.     case 2:

  4831.         v->codingset2 = CS_MID_RATE_INTER;

  4832.         break;

  4833.     }

  4834. 

  4835.     s->first_slice_line = 1;

  4836.     for (s->mb_y = s->start_mb_y; s->mb_y < s->end_mb_y; s->mb_y++) {

  4837.         s->mb_x = 0;

  4838.         ff_init_block_index(s);

  4839.         for (; s->mb_x < s->mb_width; s->mb_x++) {

  4840.             ff_update_block_index(s);

  4841. 

  4842.             if (v->fcm == ILACE_FIELD)

  4843.                 vc1_decode_b_mb_intfi(v);

  4844.             else

  4845.                 vc1_decode_b_mb(v);

  4846.             if (get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {

  4847.                 // TODO: may need modification to handle slice coding

  4848.                 ff_er_add_slice(s, 0, s->start_mb_y, s->mb_x, s->mb_y, ER_MB_ERROR);

  4849.                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i at %ix%i\n",

  4850.                        get_bits_count(&s->gb), v->bits, s->mb_x, s->mb_y);

  4851.                 return;

  4852.             }

  4853.             if (v->s.loop_filter) vc1_loop_filter_iblk(v, v->pq);

  4854.         }

  4855.         if (!v->s.loop_filter)

  4856.             ff_draw_horiz_band(s, s->mb_y * 16, 16);

  4857.         else if (s->mb_y)

  4858.             ff_draw_horiz_band(s, (s->mb_y - 1) * 16, 16);

  4859.         s->first_slice_line = 0;

  4860.     }

  4861.     if (v->s.loop_filter)

  4862.         ff_draw_horiz_band(s, (s->end_mb_y - 1) * 16, 16);

  4863.     ff_er_add_slice(s, 0, s->start_mb_y << v->field_mode, s->mb_width - 1,

  4864.                     (s->end_mb_y << v->field_mode) - 1, ER_MB_END);

  4865. }

  4866. 

  4867. static void vc1_decode_skip_blocks(VC1Context *v)

  4868. {

  4869.     MpegEncContext *s = &v->s;

  4870. 

  4871.     ff_er_add_slice(s, 0, s->start_mb_y, s->mb_width - 1, s->end_mb_y - 1, ER_MB_END);

  4872.     s->first_slice_line = 1;

  4873.     for (s->mb_y = s->start_mb_y; s->mb_y < s->end_mb_y; s->mb_y++) {

  4874.         s->mb_x = 0;

  4875.         ff_init_block_index(s);

  4876.         ff_update_block_index(s);

  4877.         memcpy(s->dest[0], s->last_picture.f.data[0] + s->mb_y * 16 * s->linesize,   s->linesize   * 16);

  4878.         memcpy(s->dest[1], s->last_picture.f.data[1] + s->mb_y *  8 * s->uvlinesize, s->uvlinesize *  8);

  4879.         memcpy(s->dest[2], s->last_picture.f.data[2] + s->mb_y *  8 * s->uvlinesize, s->uvlinesize *  8);

  4880.         ff_draw_horiz_band(s, s->mb_y * 16, 16);

  4881.         s->first_slice_line = 0;

  4882.     }

  4883.     s->pict_type = AV_PICTURE_TYPE_P;

  4884. }

  4885. 

  4886. static void vc1_decode_blocks(VC1Context *v)

  4887. {

  4888. 

  4889.     v->s.esc3_level_length = 0;

  4890.     if (v->x8_type) {

  4891.         ff_intrax8_decode_picture(&v->x8, 2*v->pq + v->halfpq, v->pq * !v->pquantizer);

  4892.     } else {

  4893.         v->cur_blk_idx     =  0;

  4894.         v->left_blk_idx    = -1;

  4895.         v->topleft_blk_idx =  1;

  4896.         v->top_blk_idx     =  2;

  4897.         switch (v->s.pict_type) {

  4898.         case AV_PICTURE_TYPE_I:

  4899.             if (v->profile == PROFILE_ADVANCED)

  4900.                 vc1_decode_i_blocks_adv(v);

  4901.             else

  4902.                 vc1_decode_i_blocks(v);

  4903.             break;

  4904.         case AV_PICTURE_TYPE_P:

  4905.             if (v->p_frame_skipped)

  4906.                 vc1_decode_skip_blocks(v);

  4907.             else

  4908.                 vc1_decode_p_blocks(v);

  4909.             break;

  4910.         case AV_PICTURE_TYPE_B:

  4911.             if (v->bi_type) {

  4912.                 if (v->profile == PROFILE_ADVANCED)

  4913.                     vc1_decode_i_blocks_adv(v);

  4914.                 else

  4915.                     vc1_decode_i_blocks(v);

  4916.             } else

  4917.                 vc1_decode_b_blocks(v);

  4918.             break;

  4919.         }

  4920.     }

  4921. }

  4922. 

  4923. #if CONFIG_WMV3IMAGE_DECODER || CONFIG_VC1IMAGE_DECODER

  4924. 

  4925. typedef struct {

  4926.     /**

  4927.      * Transform coefficients for both sprites in 16.16 fixed point format,

  4928.      * in the order they appear in the bitstream:

  4929.      *  x scale

  4930.      *  rotation 1 (unused)

  4931.      *  x offset

  4932.      *  rotation 2 (unused)

  4933.      *  y scale

  4934.      *  y offset

  4935.      *  alpha

  4936.      */

  4937.     int coefs[2][7];

  4938. 

  4939.     int effect_type, effect_flag;

  4940.     int effect_pcount1, effect_pcount2;   ///< amount of effect parameters stored in effect_params

  4941.     int effect_params1[15], effect_params2[10]; ///< effect parameters in 16.16 fixed point format

  4942. } SpriteData;

  4943. 

  4944. static inline int get_fp_val(GetBitContext* gb)

  4945. {

  4946.     return (get_bits_long(gb, 30) - (1 << 29)) << 1;

  4947. }

  4948. 

  4949. static void vc1_sprite_parse_transform(GetBitContext* gb, int c[7])

  4950. {

  4951.     c[1] = c[3] = 0;

  4952. 

  4953.     switch (get_bits(gb, 2)) {

  4954.     case 0:

  4955.         c[0] = 1 << 16;

  4956.         c[2] = get_fp_val(gb);

  4957.         c[4] = 1 << 16;

  4958.         break;

  4959.     case 1:

  4960.         c[0] = c[4] = get_fp_val(gb);

  4961.         c[2] = get_fp_val(gb);

  4962.         break;

  4963.     case 2:

  4964.         c[0] = get_fp_val(gb);

  4965.         c[2] = get_fp_val(gb);

  4966.         c[4] = get_fp_val(gb);

  4967.         break;

  4968.     case 3:

  4969.         c[0] = get_fp_val(gb);

  4970.         c[1] = get_fp_val(gb);

  4971.         c[2] = get_fp_val(gb);

  4972.         c[3] = get_fp_val(gb);

  4973.         c[4] = get_fp_val(gb);

  4974.         break;

  4975.     }

  4976.     c[5] = get_fp_val(gb);

  4977.     if (get_bits1(gb))

  4978.         c[6] = get_fp_val(gb);

  4979.     else

  4980.         c[6] = 1 << 16;

  4981. }

  4982. 

  4983. static void vc1_parse_sprites(VC1Context *v, GetBitContext* gb, SpriteData* sd)

  4984. {

  4985.     AVCodecContext *avctx = v->s.avctx;

  4986.     int sprite, i;

  4987. 

  4988.     for (sprite = 0; sprite <= v->two_sprites; sprite++) {

  4989.         vc1_sprite_parse_transform(gb, sd->coefs[sprite]);

  4990.         if (sd->coefs[sprite][1] || sd->coefs[sprite][3])

  4991.             av_log_ask_for_sample(avctx, "Rotation coefficients are not zero");

  4992.         av_log(avctx, AV_LOG_DEBUG, sprite ? "S2:" : "S1:");

  4993.         for (i = 0; i < 7; i++)

  4994.             av_log(avctx, AV_LOG_DEBUG, " %d.%.3d",

  4995.                    sd->coefs[sprite][i] / (1<<16),

  4996.                    (abs(sd->coefs[sprite][i]) & 0xFFFF) * 1000 / (1 << 16));

  4997.         av_log(avctx, AV_LOG_DEBUG, "\n");

  4998.     }

  4999. 

  5000.     skip_bits(gb, 2);

  5001.     if (sd->effect_type = get_bits_long(gb, 30)) {

  5002.         switch (sd->effect_pcount1 = get_bits(gb, 4)) {

  5003.         case 7:

  5004.             vc1_sprite_parse_transform(gb, sd->effect_params1);

  5005.             break;

  5006.         case 14:

  5007.             vc1_sprite_parse_transform(gb, sd->effect_params1);

  5008.             vc1_sprite_parse_transform(gb, sd->effect_params1 + 7);

  5009.             break;

  5010.         default:

  5011.             for (i = 0; i < sd->effect_pcount1; i++)

  5012.                 sd->effect_params1[i] = get_fp_val(gb);

  5013.         }

  5014.         if (sd->effect_type != 13 || sd->effect_params1[0] != sd->coefs[0][6]) {

  5015.             // effect 13 is simple alpha blending and matches the opacity above

  5016.             av_log(avctx, AV_LOG_DEBUG, "Effect: %d; params: ", sd->effect_type);

  5017.             for (i = 0; i < sd->effect_pcount1; i++)

  5018.                 av_log(avctx, AV_LOG_DEBUG, " %d.%.2d",

  5019.                        sd->effect_params1[i] / (1 << 16),

  5020.                        (abs(sd->effect_params1[i]) & 0xFFFF) * 1000 / (1 << 16));

  5021.             av_log(avctx, AV_LOG_DEBUG, "\n");

  5022.         }

  5023. 

  5024.         sd->effect_pcount2 = get_bits(gb, 16);

  5025.         if (sd->effect_pcount2 > 10) {

  5026.             av_log(avctx, AV_LOG_ERROR, "Too many effect parameters\n");

  5027.             return;

  5028.         } else if (sd->effect_pcount2) {

  5029.             i = -1;

  5030.             av_log(avctx, AV_LOG_DEBUG, "Effect params 2: ");

  5031.             while (++i < sd->effect_pcount2) {

  5032.                 sd->effect_params2[i] = get_fp_val(gb);

  5033.                 av_log(avctx, AV_LOG_DEBUG, " %d.%.2d",

  5034.                        sd->effect_params2[i] / (1 << 16),

  5035.                        (abs(sd->effect_params2[i]) & 0xFFFF) * 1000 / (1 << 16));

  5036.             }

  5037.             av_log(avctx, AV_LOG_DEBUG, "\n");

  5038.         }

  5039.     }

  5040.     if (sd->effect_flag = get_bits1(gb))

  5041.         av_log(avctx, AV_LOG_DEBUG, "Effect flag set\n");

  5042. 

  5043.     if (get_bits_count(gb) >= gb->size_in_bits +

  5044.        (avctx->codec_id == CODEC_ID_WMV3IMAGE ? 64 : 0))

  5045.         av_log(avctx, AV_LOG_ERROR, "Buffer overrun\n");

  5046.     if (get_bits_count(gb) < gb->size_in_bits - 8)

  5047.         av_log(avctx, AV_LOG_WARNING, "Buffer not fully read\n");

  5048. }

  5049. 

  5050. static void vc1_draw_sprites(VC1Context *v, SpriteData* sd)

  5051. {

  5052.     int i, plane, row, sprite;

  5053.     int sr_cache[2][2] = { { -1, -1 }, { -1, -1 } };

  5054.     uint8_t* src_h[2][2];

  5055.     int xoff[2], xadv[2], yoff[2], yadv[2], alpha;

  5056.     int ysub[2];

  5057.     MpegEncContext *s = &v->s;

  5058. 

  5059.     for (i = 0; i < 2; i++) {

  5060.         xoff[i] = av_clip(sd->coefs[i][2], 0, v->sprite_width-1 << 16);

  5061.         xadv[i] = sd->coefs[i][0];

  5062.         if (xadv[i] != 1<<16 || (v->sprite_width << 16) - (v->output_width << 16) - xoff[i])

  5063.             xadv[i] = av_clip(xadv[i], 0, ((v->sprite_width<<16) - xoff[i] - 1) / v->output_width);

  5064. 

  5065.         yoff[i] = av_clip(sd->coefs[i][5], 0, v->sprite_height-1 << 16);

  5066.         yadv[i] = av_clip(sd->coefs[i][4], 0, ((v->sprite_height << 16) - yoff[i]) / v->output_height);

  5067.     }

  5068.     alpha = av_clip(sd->coefs[1][6], 0, (1<<16) - 1);

  5069. 

  5070.     for (plane = 0; plane < (s->flags&CODEC_FLAG_GRAY ? 1 : 3); plane++) {

  5071.         int width = v->output_width>>!!plane;

  5072. 

  5073.         for (row = 0; row < v->output_height>>!!plane; row++) {

  5074.             uint8_t *dst = v->sprite_output_frame.data[plane] +

  5075.                            v->sprite_output_frame.linesize[plane] * row;

  5076. 

  5077.             for (sprite = 0; sprite <= v->two_sprites; sprite++) {

  5078.                 uint8_t *iplane = s->current_picture.f.data[plane];

  5079.                 int      iline  = s->current_picture.f.linesize[plane];

  5080.                 int      ycoord = yoff[sprite] + yadv[sprite] * row;

  5081.                 int      yline  = ycoord >> 16;

  5082.                 ysub[sprite] = ycoord & 0xFFFF;

  5083.                 if (sprite) {

  5084.                     iplane = s->last_picture.f.data[plane];

  5085.                     iline  = s->last_picture.f.linesize[plane];

  5086.                 }

  5087.                 if (!(xoff[sprite] & 0xFFFF) && xadv[sprite] == 1 << 16) {

  5088.                         src_h[sprite][0] = iplane + (xoff[sprite] >> 16) +  yline      * iline;

  5089.                     if (ysub[sprite])

  5090.                         src_h[sprite][1] = iplane + (xoff[sprite] >> 16) + (yline + 1) * iline;

  5091.                 } else {

  5092.                     if (sr_cache[sprite][0] != yline) {

  5093.                         if (sr_cache[sprite][1] == yline) {

  5094.                             FFSWAP(uint8_t*, v->sr_rows[sprite][0], v->sr_rows[sprite][1]);

  5095.                             FFSWAP(int,        sr_cache[sprite][0],   sr_cache[sprite][1]);

  5096.                         } else {

  5097.                             v->vc1dsp.sprite_h(v->sr_rows[sprite][0], iplane + yline * iline, xoff[sprite], xadv[sprite], width);

  5098.                             sr_cache[sprite][0] = yline;

  5099.                         }

  5100.                     }

  5101.                     if (ysub[sprite] && sr_cache[sprite][1] != yline + 1) {

  5102.                         v->vc1dsp.sprite_h(v->sr_rows[sprite][1], iplane + (yline + 1) * iline, xoff[sprite], xadv[sprite], width);

  5103.                         sr_cache[sprite][1] = yline + 1;

  5104.                     }

  5105.                     src_h[sprite][0] = v->sr_rows[sprite][0];

  5106.                     src_h[sprite][1] = v->sr_rows[sprite][1];

  5107.                 }

  5108.             }

  5109. 

  5110.             if (!v->two_sprites) {

  5111.                 if (ysub[0]) {

  5112.                     v->vc1dsp.sprite_v_single(dst, src_h[0][0], src_h[0][1], ysub[0], width);

  5113.                 } else {

  5114.                     memcpy(dst, src_h[0][0], width);

  5115.                 }

  5116.             } else {

  5117.                 if (ysub[0] && ysub[1]) {

  5118.                     v->vc1dsp.sprite_v_double_twoscale(dst, src_h[0][0], src_h[0][1], ysub[0],

  5119.                                                        src_h[1][0], src_h[1][1], ysub[1], alpha, width);

  5120.                 } else if (ysub[0]) {

  5121.                     v->vc1dsp.sprite_v_double_onescale(dst, src_h[0][0], src_h[0][1], ysub[0],

  5122.                                                        src_h[1][0], alpha, width);

  5123.                 } else if (ysub[1]) {

  5124.                     v->vc1dsp.sprite_v_double_onescale(dst, src_h[1][0], src_h[1][1], ysub[1],

  5125.                                                        src_h[0][0], (1<<16)-1-alpha, width);

  5126.                 } else {

  5127.                     v->vc1dsp.sprite_v_double_noscale(dst, src_h[0][0], src_h[1][0], alpha, width);

  5128.                 }

  5129.             }

  5130.         }

  5131. 

  5132.         if (!plane) {

  5133.             for (i = 0; i < 2; i++) {

  5134.                 xoff[i] >>= 1;

  5135.                 yoff[i] >>= 1;

  5136.             }

  5137.         }

  5138. 

  5139.     }

  5140. }

  5141. 

  5142. 

  5143. static int vc1_decode_sprites(VC1Context *v, GetBitContext* gb)

  5144. {

  5145.     MpegEncContext *s     = &v->s;

  5146.     AVCodecContext *avctx = s->avctx;

  5147.     SpriteData sd;

  5148. 

  5149.     vc1_parse_sprites(v, gb, &sd);

  5150. 

  5151.     if (!s->current_picture.f.data[0]) {

  5152.         av_log(avctx, AV_LOG_ERROR, "Got no sprites\n");

  5153.         return -1;

  5154.     }

  5155. 

  5156.     if (v->two_sprites && (!s->last_picture_ptr || !s->last_picture.f.data[0])) {

  5157.         av_log(avctx, AV_LOG_WARNING, "Need two sprites, only got one\n");

  5158.         v->two_sprites = 0;

  5159.     }

  5160. 

  5161.     if (v->sprite_output_frame.data[0])

  5162.         avctx->release_buffer(avctx, &v->sprite_output_frame);

  5163. 

  5164.     v->sprite_output_frame.buffer_hints = FF_BUFFER_HINTS_VALID;

  5165.     v->sprite_output_frame.reference = 0;

  5166.     if (avctx->get_buffer(avctx, &v->sprite_output_frame) < 0) {

  5167.         av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");

  5168.         return -1;

  5169.     }

  5170. 

  5171.     vc1_draw_sprites(v, &sd);

  5172. 

  5173.     return 0;

  5174. }

  5175. 

  5176. static void vc1_sprite_flush(AVCodecContext *avctx)

  5177. {

  5178.     VC1Context *v     = avctx->priv_data;

  5179.     MpegEncContext *s = &v->s;

  5180.     AVFrame *f = &s->current_picture.f;

  5181.     int plane, i;

  5182. 

  5183.     /* Windows Media Image codecs have a convergence interval of two keyframes.

  5184.        Since we can't enforce it, clear to black the missing sprite. This is

  5185.        wrong but it looks better than doing nothing. */

  5186. 

  5187.     if (f->data[0])

  5188.         for (plane = 0; plane < (s->flags&CODEC_FLAG_GRAY ? 1 : 3); plane++)

  5189.             for (i = 0; i < v->sprite_height>>!!plane; i++)

  5190.                 memset(f->data[plane] + i * f->linesize[plane],

  5191.                        plane ? 128 : 0, f->linesize[plane]);

  5192. }

  5193. 

  5194. #endif

  5195. 

  5196. static av_cold int vc1_decode_init_alloc_tables(VC1Context *v)

  5197. {

  5198.     MpegEncContext *s = &v->s;

  5199.     int i;

  5200. 

  5201.     /* Allocate mb bitplanes */

  5202.     v->mv_type_mb_plane = av_malloc (s->mb_stride * s->mb_height);

  5203.     v->direct_mb_plane  = av_malloc (s->mb_stride * s->mb_height);

  5204.     v->forward_mb_plane = av_malloc (s->mb_stride * s->mb_height);

  5205.     v->fieldtx_plane    = av_mallocz(s->mb_stride * s->mb_height);

  5206.     v->acpred_plane     = av_malloc (s->mb_stride * s->mb_height);

  5207.     v->over_flags_plane = av_malloc (s->mb_stride * s->mb_height);

  5208. 

  5209.     v->n_allocated_blks = s->mb_width + 2;

  5210.     v->block            = av_malloc(sizeof(*v->block) * v->n_allocated_blks);

  5211.     v->cbp_base         = av_malloc(sizeof(v->cbp_base[0]) * 2 * s->mb_stride);

  5212.     v->cbp              = v->cbp_base + s->mb_stride;

  5213.     v->ttblk_base       = av_malloc(sizeof(v->ttblk_base[0]) * 2 * s->mb_stride);

  5214.     v->ttblk            = v->ttblk_base + s->mb_stride;

  5215.     v->is_intra_base    = av_mallocz(sizeof(v->is_intra_base[0]) * 2 * s->mb_stride);

  5216.     v->is_intra         = v->is_intra_base + s->mb_stride;

  5217.     v->luma_mv_base     = av_malloc(sizeof(v->luma_mv_base[0]) * 2 * s->mb_stride);

  5218.     v->luma_mv          = v->luma_mv_base + s->mb_stride;

  5219. 

  5220.     /* allocate block type info in that way so it could be used with s->block_index[] */

  5221.     v->mb_type_base = av_malloc(s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);

  5222.     v->mb_type[0]   = v->mb_type_base + s->b8_stride + 1;

  5223.     v->mb_type[1]   = v->mb_type_base + s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride + 1;

  5224.     v->mb_type[2]   = v->mb_type[1] + s->mb_stride * (s->mb_height + 1);

  5225. 

  5226.     /* allocate memory to store block level MV info */

  5227.     v->blk_mv_type_base = av_mallocz(     s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);

  5228.     v->blk_mv_type      = v->blk_mv_type_base + s->b8_stride + 1;

  5229.     v->mv_f_base        = av_mallocz(2 * (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2));

  5230.     v->mv_f[0]          = v->mv_f_base + s->b8_stride + 1;

  5231.     v->mv_f[1]          = v->mv_f[0] + (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);

  5232.     v->mv_f_last_base   = av_mallocz(2 * (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2));

  5233.     v->mv_f_last[0]     = v->mv_f_last_base + s->b8_stride + 1;

  5234.     v->mv_f_last[1]     = v->mv_f_last[0] + (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);

  5235.     v->mv_f_next_base   = av_mallocz(2 * (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2));

  5236.     v->mv_f_next[0]     = v->mv_f_next_base + s->b8_stride + 1;

  5237.     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);

  5238. 

  5239.     /* Init coded blocks info */

  5240.     if (v->profile == PROFILE_ADVANCED) {

  5241. //        if (alloc_bitplane(&v->over_flags_plane, s->mb_width, s->mb_height) < 0)

  5242. //            return -1;

  5243. //        if (alloc_bitplane(&v->ac_pred_plane, s->mb_width, s->mb_height) < 0)

  5244. //            return -1;

  5245.     }

  5246. 

  5247.     ff_intrax8_common_init(&v->x8,s);

  5248. 

  5249.     if (s->avctx->codec_id == CODEC_ID_WMV3IMAGE || s->avctx->codec_id == CODEC_ID_VC1IMAGE) {

  5250.         for (i = 0; i < 4; i++)

  5251.             if (!(v->sr_rows[i >> 1][i & 1] = av_malloc(v->output_width))) return -1;

  5252.     }

  5253. 

  5254.     if (!v->mv_type_mb_plane || !v->direct_mb_plane || !v->acpred_plane || !v->over_flags_plane ||

  5255.         !v->block || !v->cbp_base || !v->ttblk_base || !v->is_intra_base || !v->luma_mv_base ||

  5256.         !v->mb_type_base)

  5257.             return -1;

  5258. 

  5259.     return 0;

  5260. }

  5261. 

  5262. /** Initialize a VC1/WMV3 decoder

  5263.  * @todo TODO: Handle VC-1 IDUs (Transport level?)

  5264.  * @todo TODO: Decypher remaining bits in extra_data

  5265.  */

  5266. static av_cold int vc1_decode_init(AVCodecContext *avctx)

  5267. {

  5268.     VC1Context *v = avctx->priv_data;

  5269.     MpegEncContext *s = &v->s;

  5270.     GetBitContext gb;

  5271.     int i;

  5272. 

  5273.     /* save the container output size for WMImage */

  5274.     v->output_width  = avctx->width;

  5275.     v->output_height = avctx->height;

  5276. 

  5277.     if (!avctx->extradata_size || !avctx->extradata)

  5278.         return -1;

  5279.     if (!(avctx->flags & CODEC_FLAG_GRAY))

  5280.         avctx->pix_fmt = avctx->get_format(avctx, avctx->codec->pix_fmts);

  5281.     else

  5282.         avctx->pix_fmt = PIX_FMT_GRAY8;

  5283.     avctx->hwaccel = ff_find_hwaccel(avctx->codec->id, avctx->pix_fmt);

  5284.     v->s.avctx = avctx;

  5285.     avctx->flags |= CODEC_FLAG_EMU_EDGE;

  5286.     v->s.flags   |= CODEC_FLAG_EMU_EDGE;

  5287. 

  5288.     if (avctx->idct_algo == FF_IDCT_AUTO) {

  5289.         avctx->idct_algo = FF_IDCT_WMV2;

  5290.     }

  5291. 

  5292.     if (ff_vc1_init_common(v) < 0)

  5293.         return -1;

  5294.     ff_vc1dsp_init(&v->vc1dsp);

  5295. 

  5296.     if (avctx->codec_id == CODEC_ID_WMV3 || avctx->codec_id == CODEC_ID_WMV3IMAGE) {

  5297.         int count = 0;

  5298. 

  5299.         // looks like WMV3 has a sequence header stored in the extradata

  5300.         // advanced sequence header may be before the first frame

  5301.         // the last byte of the extradata is a version number, 1 for the

  5302.         // samples we can decode

  5303. 

  5304.         init_get_bits(&gb, avctx->extradata, avctx->extradata_size*8);

  5305. 

  5306.         if (vc1_decode_sequence_header(avctx, v, &gb) < 0)

  5307.           return -1;

  5308. 

  5309.         count = avctx->extradata_size*8 - get_bits_count(&gb);

  5310.         if (count > 0) {

  5311.             av_log(avctx, AV_LOG_INFO, "Extra data: %i bits left, value: %X\n",

  5312.                    count, get_bits(&gb, count));

  5313.         } else if (count < 0) {

  5314.             av_log(avctx, AV_LOG_INFO, "Read %i bits in overflow\n", -count);

  5315.         }

  5316.     } else { // VC1/WVC1/WVP2

  5317.         const uint8_t *start = avctx->extradata;

  5318.         uint8_t *end = avctx->extradata + avctx->extradata_size;

  5319.         const uint8_t *next;

  5320.         int size, buf2_size;

  5321.         uint8_t *buf2 = NULL;

  5322.         int seq_initialized = 0, ep_initialized = 0;

  5323. 

  5324.         if (avctx->extradata_size < 16) {

  5325.             av_log(avctx, AV_LOG_ERROR, "Extradata size too small: %i\n", avctx->extradata_size);

  5326.             return -1;

  5327.         }

  5328. 

  5329.         buf2  = av_mallocz(avctx->extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);

  5330.         start = find_next_marker(start, end); // in WVC1 extradata first byte is its size, but can be 0 in mkv

  5331.         next  = start;

  5332.         for (; next < end; start = next) {

  5333.             next = find_next_marker(start + 4, end);

  5334.             size = next - start - 4;

  5335.             if (size <= 0)

  5336.                 continue;

  5337.             buf2_size = vc1_unescape_buffer(start + 4, size, buf2);

  5338.             init_get_bits(&gb, buf2, buf2_size * 8);

  5339.             switch (AV_RB32(start)) {

  5340.             case VC1_CODE_SEQHDR:

  5341.                 if (vc1_decode_sequence_header(avctx, v, &gb) < 0) {

  5342.                     av_free(buf2);

  5343.                     return -1;

  5344.                 }

  5345.                 seq_initialized = 1;

  5346.                 break;

  5347.             case VC1_CODE_ENTRYPOINT:

  5348.                 if (vc1_decode_entry_point(avctx, v, &gb) < 0) {

  5349.                     av_free(buf2);

  5350.                     return -1;

  5351.                 }

  5352.                 ep_initialized = 1;

  5353.                 break;

  5354.             }

  5355.         }

  5356.         av_free(buf2);

  5357.         if (!seq_initialized || !ep_initialized) {

  5358.             av_log(avctx, AV_LOG_ERROR, "Incomplete extradata\n");

  5359.             return -1;

  5360.         }

  5361.         v->res_sprite = (avctx->codec_tag == MKTAG('W','V','P','2'));

  5362.     }

  5363. 

  5364.     avctx->profile = v->profile;

  5365.     if (v->profile == PROFILE_ADVANCED)

  5366.         avctx->level = v->level;

  5367. 

  5368.     avctx->has_b_frames = !!avctx->max_b_frames;

  5369. 

  5370.     s->mb_width  = (avctx->coded_width  + 15) >> 4;

  5371.     s->mb_height = (avctx->coded_height + 15) >> 4;

  5372. 

  5373.     if (v->profile == PROFILE_ADVANCED || v->res_fasttx) {

  5374.         for (i = 0; i < 64; i++) {

  5375. #define transpose(x) ((x >> 3) | ((x & 7) << 3))

  5376.             v->zz_8x8[0][i] = transpose(wmv1_scantable[0][i]);

  5377.             v->zz_8x8[1][i] = transpose(wmv1_scantable[1][i]);

  5378.             v->zz_8x8[2][i] = transpose(wmv1_scantable[2][i]);

  5379.             v->zz_8x8[3][i] = transpose(wmv1_scantable[3][i]);

  5380.             v->zzi_8x8[i] = transpose(ff_vc1_adv_interlaced_8x8_zz[i]);

  5381.         }

  5382.         v->left_blk_sh = 0;

  5383.         v->top_blk_sh  = 3;

  5384.     } else {

  5385.         memcpy(v->zz_8x8, wmv1_scantable, 4*64);

  5386.         v->left_blk_sh = 3;

  5387.         v->top_blk_sh  = 0;

  5388.     }

  5389. 

  5390.     if (avctx->codec_id == CODEC_ID_WMV3IMAGE || avctx->codec_id == CODEC_ID_VC1IMAGE) {

  5391.         v->sprite_width  = avctx->coded_width;

  5392.         v->sprite_height = avctx->coded_height;

  5393. 

  5394.         avctx->coded_width  = avctx->width  = v->output_width;

  5395.         avctx->coded_height = avctx->height = v->output_height;

  5396. 

  5397.         // prevent 16.16 overflows

  5398.         if (v->sprite_width  > 1 << 14 ||

  5399.             v->sprite_height > 1 << 14 ||

  5400.             v->output_width  > 1 << 14 ||

  5401.             v->output_height > 1 << 14) return -1;

  5402.     }

  5403.     return 0;

  5404. }

  5405. 

  5406. /** Close a VC1/WMV3 decoder

  5407.  * @warning Initial try at using MpegEncContext stuff

  5408.  */

  5409. static av_cold int vc1_decode_end(AVCodecContext *avctx)

  5410. {

  5411.     VC1Context *v = avctx->priv_data;

  5412.     int i;

  5413. 

  5414.     if ((avctx->codec_id == CODEC_ID_WMV3IMAGE || avctx->codec_id == CODEC_ID_VC1IMAGE)

  5415.         && v->sprite_output_frame.data[0])

  5416.         avctx->release_buffer(avctx, &v->sprite_output_frame);

  5417.     for (i = 0; i < 4; i++)

  5418.         av_freep(&v->sr_rows[i >> 1][i & 1]);

  5419.     av_freep(&v->hrd_rate);

  5420.     av_freep(&v->hrd_buffer);

  5421.     MPV_common_end(&v->s);

  5422.     av_freep(&v->mv_type_mb_plane);

  5423.     av_freep(&v->direct_mb_plane);

  5424.     av_freep(&v->forward_mb_plane);

  5425.     av_freep(&v->fieldtx_plane);

  5426.     av_freep(&v->acpred_plane);

  5427.     av_freep(&v->over_flags_plane);

  5428.     av_freep(&v->mb_type_base);

  5429.     av_freep(&v->blk_mv_type_base);

  5430.     av_freep(&v->mv_f_base);

  5431.     av_freep(&v->mv_f_last_base);

  5432.     av_freep(&v->mv_f_next_base);

  5433.     av_freep(&v->block);

  5434.     av_freep(&v->cbp_base);

  5435.     av_freep(&v->ttblk_base);

  5436.     av_freep(&v->is_intra_base); // FIXME use v->mb_type[]

  5437.     av_freep(&v->luma_mv_base);

  5438.     ff_intrax8_common_end(&v->x8);

  5439.     return 0;

  5440. }

  5441. 

  5442. 

  5443. /** Decode a VC1/WMV3 frame

  5444.  * @todo TODO: Handle VC-1 IDUs (Transport level?)

  5445.  */

  5446. static int vc1_decode_frame(AVCodecContext *avctx, void *data,

  5447.                             int *data_size, AVPacket *avpkt)

  5448. {

  5449.     const uint8_t *buf = avpkt->data;

  5450.     int buf_size = avpkt->size, n_slices = 0, i;

  5451.     VC1Context *v = avctx->priv_data;

  5452.     MpegEncContext *s = &v->s;

  5453.     AVFrame *pict = data;

  5454.     uint8_t *buf2 = NULL;

  5455.     const uint8_t *buf_start = buf;

  5456.     int mb_height, n_slices1=-1;

  5457.     struct {

  5458.         uint8_t *buf;

  5459.         GetBitContext gb;

  5460.         int mby_start;

  5461.     } *slices = NULL, *tmp;

  5462. 

  5463.     if(s->flags & CODEC_FLAG_LOW_DELAY)

  5464.         s->low_delay = 1;

  5465. 

  5466.     /* no supplementary picture */

  5467.     if (buf_size == 0 || (buf_size == 4 && AV_RB32(buf) == VC1_CODE_ENDOFSEQ)) {

  5468.         /* special case for last picture */

  5469.         if (s->low_delay == 0 && s->next_picture_ptr) {

  5470.             *pict = *(AVFrame*)s->next_picture_ptr;

  5471.             s->next_picture_ptr = NULL;

  5472. 

  5473.             *data_size = sizeof(AVFrame);

  5474.         }

  5475. 

  5476.         return 0;

  5477.     }

  5478. 

  5479.     if (s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU) {

  5480.         if (v->profile < PROFILE_ADVANCED)

  5481.             avctx->pix_fmt = PIX_FMT_VDPAU_WMV3;

  5482.         else

  5483.             avctx->pix_fmt = PIX_FMT_VDPAU_VC1;

  5484.     }

  5485. 

  5486.     //for advanced profile we may need to parse and unescape data

  5487.     if (avctx->codec_id == CODEC_ID_VC1 || avctx->codec_id == CODEC_ID_VC1IMAGE) {

  5488.         int buf_size2 = 0;

  5489.         buf2 = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);

  5490. 

  5491.         if (IS_MARKER(AV_RB32(buf))) { /* frame starts with marker and needs to be parsed */

  5492.             const uint8_t *start, *end, *next;

  5493.             int size;

  5494. 

  5495.             next = buf;

  5496.             for (start = buf, end = buf + buf_size; next < end; start = next) {

  5497.                 next = find_next_marker(start + 4, end);

  5498.                 size = next - start - 4;

  5499.                 if (size <= 0) continue;

  5500.                 switch (AV_RB32(start)) {

  5501.                 case VC1_CODE_FRAME:

  5502.                     if (avctx->hwaccel ||

  5503.                         s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU)

  5504.                         buf_start = start;

  5505.                     buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);

  5506.                     break;

  5507.                 case VC1_CODE_FIELD: {

  5508.                     int buf_size3;

  5509.                     slices = av_realloc(slices, sizeof(*slices) * (n_slices+1));

  5510.                     if (!slices)

  5511.                         goto err;

  5512.                     slices[n_slices].buf = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);

  5513.                     if (!slices[n_slices].buf)

  5514.                         goto err;

  5515.                     buf_size3 = vc1_unescape_buffer(start + 4, size,

  5516.                                                     slices[n_slices].buf);

  5517.                     init_get_bits(&slices[n_slices].gb, slices[n_slices].buf,

  5518.                                   buf_size3 << 3);

  5519.                     /* assuming that the field marker is at the exact middle,

  5520.                        hope it's correct */

  5521.                     slices[n_slices].mby_start = s->mb_height >> 1;

  5522.                     n_slices1 = n_slices - 1; // index of the last slice of the first field

  5523.                     n_slices++;

  5524.                     break;

  5525.                 }

  5526.                 case VC1_CODE_ENTRYPOINT: /* it should be before frame data */

  5527.                     buf_size2 = vc1_unescape_buffer(start + 4, size, buf2);

  5528.                     init_get_bits(&s->gb, buf2, buf_size2 * 8);

  5529.                     vc1_decode_entry_point(avctx, v, &s->gb);

  5530.                     break;

  5531.                 case VC1_CODE_SLICE: {

  5532.                     int buf_size3;

  5533.                     slices = av_realloc(slices, sizeof(*slices) * (n_slices+1));

  5534.                     if (!slices)

  5535.                         goto err;

  5536.                     slices[n_slices].buf = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);

  5537.                     if (!slices[n_slices].buf)

  5538.                         goto err;

  5539.                     buf_size3 = vc1_unescape_buffer(start + 4, size,

  5540.                                                     slices[n_slices].buf);

  5541.                     init_get_bits(&slices[n_slices].gb, slices[n_slices].buf,

  5542.                                   buf_size3 << 3);

  5543.                     slices[n_slices].mby_start = get_bits(&slices[n_slices].gb, 9);

  5544.                     n_slices++;

  5545.                     break;

  5546.                 }

  5547.                 }

  5548.             }

  5549.         } else if (v->interlace && ((buf[0] & 0xC0) == 0xC0)) { /* WVC1 interlaced stores both fields divided by marker */

  5550.             const uint8_t *divider;

  5551.             int buf_size3;

  5552. 

  5553.             divider = find_next_marker(buf, buf + buf_size);

  5554.             if ((divider == (buf + buf_size)) || AV_RB32(divider) != VC1_CODE_FIELD) {

  5555.                 av_log(avctx, AV_LOG_ERROR, "Error in WVC1 interlaced frame\n");

  5556.                 goto err;

  5557.             } else { // found field marker, unescape second field

  5558.                 tmp = av_realloc(slices, sizeof(*slices) * (n_slices+1));

  5559.                 if (!tmp)

  5560.                     goto err;

  5561.                 slices = tmp;

  5562.                 slices[n_slices].buf = av_mallocz(buf_size + FF_INPUT_BUFFER_PADDING_SIZE);

  5563.                 if (!slices[n_slices].buf)

  5564.                     goto err;

  5565.                 buf_size3 = vc1_unescape_buffer(divider + 4, buf + buf_size - divider - 4, slices[n_slices].buf);

  5566.                 init_get_bits(&slices[n_slices].gb, slices[n_slices].buf,

  5567.                               buf_size3 << 3);

  5568.                 slices[n_slices].mby_start = s->mb_height >> 1;

  5569.                 n_slices1 = n_slices - 1;

  5570.                 n_slices++;

  5571.             }

  5572.             buf_size2 = vc1_unescape_buffer(buf, divider - buf, buf2);

  5573.         } else {

  5574.             buf_size2 = vc1_unescape_buffer(buf, buf_size, buf2);

  5575.         }

  5576.         init_get_bits(&s->gb, buf2, buf_size2*8);

  5577.     } else

  5578.         init_get_bits(&s->gb, buf, buf_size*8);

  5579. 

  5580.     if (v->res_sprite) {

  5581.         v->new_sprite  = !get_bits1(&s->gb);

  5582.         v->two_sprites =  get_bits1(&s->gb);

  5583.         /* res_sprite means a Windows Media Image stream, CODEC_ID_*IMAGE means

  5584.            we're using the sprite compositor. These are intentionally kept separate

  5585.            so you can get the raw sprites by using the wmv3 decoder for WMVP or

  5586.            the vc1 one for WVP2 */

  5587.         if (avctx->codec_id == CODEC_ID_WMV3IMAGE || avctx->codec_id == CODEC_ID_VC1IMAGE) {

  5588.             if (v->new_sprite) {

  5589.                 // switch AVCodecContext parameters to those of the sprites

  5590.                 avctx->width  = avctx->coded_width  = v->sprite_width;

  5591.                 avctx->height = avctx->coded_height = v->sprite_height;

  5592.             } else {

  5593.                 goto image;

  5594.             }

  5595.         }

  5596.     }

  5597. 

  5598.     if (s->context_initialized &&

  5599.         (s->width  != avctx->coded_width ||

  5600.          s->height != avctx->coded_height)) {

  5601.         vc1_decode_end(avctx);

  5602.     }

  5603. 

  5604.     if (!s->context_initialized) {

  5605.         if (ff_msmpeg4_decode_init(avctx) < 0 || vc1_decode_init_alloc_tables(v) < 0)

  5606.             return -1;

  5607. 

  5608.         s->low_delay = !avctx->has_b_frames || v->res_sprite;

  5609. 

  5610.         if (v->profile == PROFILE_ADVANCED) {

  5611.             s->h_edge_pos = avctx->coded_width;

  5612.             s->v_edge_pos = avctx->coded_height;

  5613.         }

  5614.     }

  5615. 

  5616.     /* We need to set current_picture_ptr before reading the header,

  5617.      * otherwise we cannot store anything in there. */

  5618.     if (s->current_picture_ptr == NULL || s->current_picture_ptr->f.data[0]) {

  5619.         int i = ff_find_unused_picture(s, 0);

  5620.         if (i < 0)

  5621.             goto err;

  5622.         s->current_picture_ptr = &s->picture[i];

  5623.     }

  5624. 

  5625.     // do parse frame header

  5626.     v->pic_header_flag = 0;

  5627.     if (v->profile < PROFILE_ADVANCED) {

  5628.         if (vc1_parse_frame_header(v, &s->gb) == -1) {

  5629.             goto err;

  5630.         }

  5631.     } else {

  5632.         if (vc1_parse_frame_header_adv(v, &s->gb) == -1) {

  5633.             goto err;

  5634.         }

  5635.     }

  5636. 

  5637.     if ((avctx->codec_id == CODEC_ID_WMV3IMAGE || avctx->codec_id == CODEC_ID_VC1IMAGE)

  5638.         && s->pict_type != AV_PICTURE_TYPE_I) {

  5639.         av_log(v->s.avctx, AV_LOG_ERROR, "Sprite decoder: expected I-frame\n");

  5640.         goto err;

  5641.     }

  5642. 

  5643.     // process pulldown flags

  5644.     s->current_picture_ptr->f.repeat_pict = 0;

  5645.     // Pulldown flags are only valid when 'broadcast' has been set.

  5646.     // So ticks_per_frame will be 2

  5647.     if (v->rff) {

  5648.         // repeat field

  5649.         s->current_picture_ptr->f.repeat_pict = 1;

  5650.     } else if (v->rptfrm) {

  5651.         // repeat frames

  5652.         s->current_picture_ptr->f.repeat_pict = v->rptfrm * 2;

  5653.     }

  5654. 

  5655.     // for skipping the frame

  5656.     s->current_picture.f.pict_type = s->pict_type;

  5657.     s->current_picture.f.key_frame = s->pict_type == AV_PICTURE_TYPE_I;

  5658. 

  5659.     /* skip B-frames if we don't have reference frames */

  5660.     if (s->last_picture_ptr == NULL && (s->pict_type == AV_PICTURE_TYPE_B || s->dropable)) {

  5661.         goto err;

  5662.     }

  5663.     if ((avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type == AV_PICTURE_TYPE_B) ||

  5664.         (avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type != AV_PICTURE_TYPE_I) ||

  5665.          avctx->skip_frame >= AVDISCARD_ALL) {

  5666.         goto end;

  5667.     }

  5668. 

  5669.     if (s->next_p_frame_damaged) {

  5670.         if (s->pict_type == AV_PICTURE_TYPE_B)

  5671.             goto end;

  5672.         else

  5673.             s->next_p_frame_damaged = 0;

  5674.     }

  5675. 

  5676.     if (MPV_frame_start(s, avctx) < 0) {

  5677.         goto err;

  5678.     }

  5679. 

  5680.     s->me.qpel_put = s->dsp.put_qpel_pixels_tab;

  5681.     s->me.qpel_avg = s->dsp.avg_qpel_pixels_tab;

  5682. 

  5683.     if ((CONFIG_VC1_VDPAU_DECODER)

  5684.         &&s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU)

  5685.         ff_vdpau_vc1_decode_picture(s, buf_start, (buf + buf_size) - buf_start);

  5686.     else if (avctx->hwaccel) {

  5687.         if (avctx->hwaccel->start_frame(avctx, buf, buf_size) < 0)

  5688.             goto err;

  5689.         if (avctx->hwaccel->decode_slice(avctx, buf_start, (buf + buf_size) - buf_start) < 0)

  5690.             goto err;

  5691.         if (avctx->hwaccel->end_frame(avctx) < 0)

  5692.             goto err;

  5693.     } else {

  5694.         ff_er_frame_start(s);

  5695. 

  5696.         v->bits = buf_size * 8;

  5697.         if (v->field_mode) {

  5698.             uint8_t *tmp[2];

  5699.             s->current_picture.f.linesize[0] <<= 1;

  5700.             s->current_picture.f.linesize[1] <<= 1;

  5701.             s->current_picture.f.linesize[2] <<= 1;

  5702.             s->linesize                      <<= 1;

  5703.             s->uvlinesize                    <<= 1;

  5704.             tmp[0]          = v->mv_f_last[0];

  5705.             tmp[1]          = v->mv_f_last[1];

  5706.             v->mv_f_last[0] = v->mv_f_next[0];

  5707.             v->mv_f_last[1] = v->mv_f_next[1];

  5708.             v->mv_f_next[0] = v->mv_f[0];

  5709.             v->mv_f_next[1] = v->mv_f[1];

  5710.             v->mv_f[0] = tmp[0];

  5711.             v->mv_f[1] = tmp[1];

  5712.         }

  5713.         mb_height = s->mb_height >> v->field_mode;

  5714.         for (i = 0; i <= n_slices; i++) {

  5715.             if (i > 0 &&  slices[i - 1].mby_start >= mb_height) {

  5716.                 v->second_field = 1;

  5717.                 v->blocks_off   = s->mb_width  * s->mb_height << 1;

  5718.                 v->mb_off       = s->mb_stride * s->mb_height >> 1;

  5719.             } else {

  5720.                 v->second_field = 0;

  5721.                 v->blocks_off   = 0;

  5722.                 v->mb_off       = 0;

  5723.             }

  5724.             if (i) {

  5725.                 v->pic_header_flag = 0;

  5726.                 if (v->field_mode && i == n_slices1 + 2)

  5727.                     vc1_parse_frame_header_adv(v, &s->gb);

  5728.                 else if (get_bits1(&s->gb)) {

  5729.                     v->pic_header_flag = 1;

  5730.                     vc1_parse_frame_header_adv(v, &s->gb);

  5731.                 }

  5732.             }

  5733.             s->start_mb_y = (i == 0) ? 0 : FFMAX(0, slices[i-1].mby_start % mb_height);

  5734.             if (!v->field_mode || v->second_field)

  5735.                 s->end_mb_y = (i == n_slices     ) ? mb_height : FFMIN(mb_height, slices[i].mby_start % mb_height);

  5736.             else

  5737.                 s->end_mb_y = (i <= n_slices1 + 1) ? mb_height : FFMIN(mb_height, slices[i].mby_start % mb_height);

  5738.             vc1_decode_blocks(v);

  5739.             if (i != n_slices)

  5740.                 s->gb = slices[i].gb;

  5741.         }

  5742.         if (v->field_mode) {

  5743.             v->second_field = 0;

  5744.             if (s->pict_type == AV_PICTURE_TYPE_B) {

  5745.                 memcpy(v->mv_f_base, v->mv_f_next_base,

  5746.                        2 * (s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2));

  5747.             }

  5748.             s->current_picture.f.linesize[0] >>= 1;

  5749.             s->current_picture.f.linesize[1] >>= 1;

  5750.             s->current_picture.f.linesize[2] >>= 1;

  5751.             s->linesize                      >>= 1;

  5752.             s->uvlinesize                    >>= 1;

  5753.         }

  5754. //av_log(s->avctx, AV_LOG_INFO, "Consumed %i/%i bits\n", get_bits_count(&s->gb), s->gb.size_in_bits);

  5755. //  if (get_bits_count(&s->gb) > buf_size * 8)

  5756. //      return -1;

  5757.         if(s->error_occurred && s->pict_type == AV_PICTURE_TYPE_B)

  5758.             goto err;

  5759.         ff_er_frame_end(s);

  5760.     }

  5761. 

  5762.     MPV_frame_end(s);

  5763. 

  5764.     if (avctx->codec_id == CODEC_ID_WMV3IMAGE || avctx->codec_id == CODEC_ID_VC1IMAGE) {

  5765. image:

  5766.         avctx->width  = avctx->coded_width  = v->output_width;

  5767.         avctx->height = avctx->coded_height = v->output_height;

  5768.         if (avctx->skip_frame >= AVDISCARD_NONREF)

  5769.             goto end;

  5770. #if CONFIG_WMV3IMAGE_DECODER || CONFIG_VC1IMAGE_DECODER

  5771.         if (vc1_decode_sprites(v, &s->gb))

  5772.             goto err;

  5773. #endif

  5774.         *pict      = v->sprite_output_frame;

  5775.         *data_size = sizeof(AVFrame);

  5776.     } else {

  5777.         if (s->pict_type == AV_PICTURE_TYPE_B || s->low_delay) {

  5778.             *pict = *(AVFrame*)s->current_picture_ptr;

  5779.         } else if (s->last_picture_ptr != NULL) {

  5780.             *pict = *(AVFrame*)s->last_picture_ptr;

  5781.         }

  5782.         if (s->last_picture_ptr || s->low_delay) {

  5783.             *data_size = sizeof(AVFrame);

  5784.             ff_print_debug_info(s, pict);

  5785.         }

  5786.     }

  5787. 

  5788. end:

  5789.     av_free(buf2);

  5790.     for (i = 0; i < n_slices; i++)

  5791.         av_free(slices[i].buf);

  5792.     av_free(slices);

  5793.     return buf_size;

  5794. 

  5795. err:

  5796.     av_free(buf2);

  5797.     for (i = 0; i < n_slices; i++)

  5798.         av_free(slices[i].buf);

  5799.     av_free(slices);

  5800.     return -1;

  5801. }

  5802. 

  5803. 

  5804. static const AVProfile profiles[] = {

  5805.     { FF_PROFILE_VC1_SIMPLE,   "Simple"   },

  5806.     { FF_PROFILE_VC1_MAIN,     "Main"     },

  5807.     { FF_PROFILE_VC1_COMPLEX,  "Complex"  },

  5808.     { FF_PROFILE_VC1_ADVANCED, "Advanced" },

  5809.     { FF_PROFILE_UNKNOWN },

  5810. };

  5811. 

  5812. AVCodec ff_vc1_decoder = {

  5813.     .name           = "vc1",

  5814.     .type           = AVMEDIA_TYPE_VIDEO,

  5815.     .id             = CODEC_ID_VC1,

  5816.     .priv_data_size = sizeof(VC1Context),

  5817.     .init           = vc1_decode_init,

  5818.     .close          = vc1_decode_end,

  5819.     .decode         = vc1_decode_frame,

  5820.     .capabilities   = CODEC_CAP_DR1 | CODEC_CAP_DELAY,

  5821.     .long_name      = NULL_IF_CONFIG_SMALL("SMPTE VC-1"),

  5822.     .pix_fmts       = ff_hwaccel_pixfmt_list_420,

  5823.     .profiles       = NULL_IF_CONFIG_SMALL(profiles)

  5824. };

  5825. 

  5826. #if CONFIG_WMV3_DECODER

  5827. AVCodec ff_wmv3_decoder = {

  5828.     .name           = "wmv3",

  5829.     .type           = AVMEDIA_TYPE_VIDEO,

  5830.     .id             = CODEC_ID_WMV3,

  5831.     .priv_data_size = sizeof(VC1Context),

  5832.     .init           = vc1_decode_init,

  5833.     .close          = vc1_decode_end,

  5834.     .decode         = vc1_decode_frame,

  5835.     .capabilities   = CODEC_CAP_DR1 | CODEC_CAP_DELAY,

  5836.     .long_name      = NULL_IF_CONFIG_SMALL("Windows Media Video 9"),

  5837.     .pix_fmts       = ff_hwaccel_pixfmt_list_420,

  5838.     .profiles       = NULL_IF_CONFIG_SMALL(profiles)

  5839. };

  5840. #endif

  5841. 

  5842. #if CONFIG_WMV3_VDPAU_DECODER

  5843. AVCodec ff_wmv3_vdpau_decoder = {

  5844.     .name           = "wmv3_vdpau",

  5845.     .type           = AVMEDIA_TYPE_VIDEO,

  5846.     .id             = CODEC_ID_WMV3,

  5847.     .priv_data_size = sizeof(VC1Context),

  5848.     .init           = vc1_decode_init,

  5849.     .close          = vc1_decode_end,

  5850.     .decode         = vc1_decode_frame,

  5851.     .capabilities   = CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_HWACCEL_VDPAU,

  5852.     .long_name      = NULL_IF_CONFIG_SMALL("Windows Media Video 9 VDPAU"),

  5853.     .pix_fmts       = (const enum PixelFormat[]){PIX_FMT_VDPAU_WMV3, PIX_FMT_NONE},

  5854.     .profiles       = NULL_IF_CONFIG_SMALL(profiles)

  5855. };

  5856. #endif

  5857. 

  5858. #if CONFIG_VC1_VDPAU_DECODER

  5859. AVCodec ff_vc1_vdpau_decoder = {

  5860.     .name           = "vc1_vdpau",

  5861.     .type           = AVMEDIA_TYPE_VIDEO,

  5862.     .id             = CODEC_ID_VC1,

  5863.     .priv_data_size = sizeof(VC1Context),

  5864.     .init           = vc1_decode_init,

  5865.     .close          = vc1_decode_end,

  5866.     .decode         = vc1_decode_frame,

  5867.     .capabilities   = CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_HWACCEL_VDPAU,

  5868.     .long_name      = NULL_IF_CONFIG_SMALL("SMPTE VC-1 VDPAU"),

  5869.     .pix_fmts       = (const enum PixelFormat[]){PIX_FMT_VDPAU_VC1, PIX_FMT_NONE},

  5870.     .profiles       = NULL_IF_CONFIG_SMALL(profiles)

  5871. };

  5872. #endif

  5873. 

  5874. #if CONFIG_WMV3IMAGE_DECODER

  5875. AVCodec ff_wmv3image_decoder = {

  5876.     .name           = "wmv3image",

  5877.     .type           = AVMEDIA_TYPE_VIDEO,

  5878.     .id             = CODEC_ID_WMV3IMAGE,

  5879.     .priv_data_size = sizeof(VC1Context),

  5880.     .init           = vc1_decode_init,

  5881.     .close          = vc1_decode_end,

  5882.     .decode         = vc1_decode_frame,

  5883.     .capabilities   = CODEC_CAP_DR1,

  5884.     .flush          = vc1_sprite_flush,

  5885.     .long_name      = NULL_IF_CONFIG_SMALL("Windows Media Video 9 Image"),

  5886.     .pix_fmts       = ff_pixfmt_list_420

  5887. };

  5888. #endif

  5889. 

  5890. #if CONFIG_VC1IMAGE_DECODER

  5891. AVCodec ff_vc1image_decoder = {

  5892.     .name           = "vc1image",

  5893.     .type           = AVMEDIA_TYPE_VIDEO,

  5894.     .id             = CODEC_ID_VC1IMAGE,

  5895.     .priv_data_size = sizeof(VC1Context),

  5896.     .init           = vc1_decode_init,

  5897.     .close          = vc1_decode_end,

  5898.     .decode         = vc1_decode_frame,

  5899.     .capabilities   = CODEC_CAP_DR1,

  5900.     .flush          = vc1_sprite_flush,

  5901.     .long_name      = NULL_IF_CONFIG_SMALL("Windows Media Video 9 Image v2"),

  5902.     .pix_fmts       = ff_pixfmt_list_420

  5903. };

  5904. #endif