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.
6144 Commits
32 Branches
837MB
Tree: d8cfa5534a
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'd8cfa5534a'
${ noResults }
FFmpeg/libavcodec/vc1.c

2999 lines
98KB
Raw Blame History 

  1. /*

  2.  * VC-1 and WMV3 decoder

  3.  * Copyright (c) 2006 Konstantin Shishkov

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

  5.  *

  6.  * This library is free software; you can redistribute it and/or

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

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

  9.  * version 2 of the License, or (at your option) any later version.

  10.  *

  11.  * This library is distributed in the hope that it will be useful,

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

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

  14.  * Lesser General Public License for more details.

  15.  *

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

  17.  * License along with this library; if not, write to the Free Software

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

  19.  *

  20.  */

  21. 

  22. /**

  23.  * @file vc1.c

  24.  * VC-1 and WMV3 decoder

  25.  *

  26.  */

  27. #include "common.h"

  28. #include "dsputil.h"

  29. #include "avcodec.h"

  30. #include "mpegvideo.h"

  31. #include "vc1data.h"

  32. #include "vc1acdata.h"

  33. 

  34. #undef NDEBUG

  35. #include <assert.h>

  36. 

  37. extern const uint32_t ff_table0_dc_lum[120][2], ff_table1_dc_lum[120][2];

  38. extern const uint32_t ff_table0_dc_chroma[120][2], ff_table1_dc_chroma[120][2];

  39. extern VLC ff_msmp4_dc_luma_vlc[2], ff_msmp4_dc_chroma_vlc[2];

  40. #define MB_INTRA_VLC_BITS 9

  41. extern VLC ff_msmp4_mb_i_vlc;

  42. extern const uint16_t ff_msmp4_mb_i_table[64][2];

  43. #define DC_VLC_BITS 9

  44. #define AC_VLC_BITS 9

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

  46. 

  47. 

  48. /** Available Profiles */

  49. //@{

  50. enum Profile {

  51.     PROFILE_SIMPLE,

  52.     PROFILE_MAIN,

  53.     PROFILE_COMPLEX, ///< TODO: WMV9 specific

  54.     PROFILE_ADVANCED

  55. };

  56. //@}

  57. 

  58. /** Sequence quantizer mode */

  59. //@{

  60. enum QuantMode {

  61.     QUANT_FRAME_IMPLICIT,    ///< Implicitly specified at frame level

  62.     QUANT_FRAME_EXPLICIT,    ///< Explicitly specified at frame level

  63.     QUANT_NON_UNIFORM,       ///< Non-uniform quant used for all frames

  64.     QUANT_UNIFORM            ///< Uniform quant used for all frames

  65. };

  66. //@}

  67. 

  68. /** Where quant can be changed */

  69. //@{

  70. enum DQProfile {

  71.     DQPROFILE_FOUR_EDGES,

  72.     DQPROFILE_DOUBLE_EDGES,

  73.     DQPROFILE_SINGLE_EDGE,

  74.     DQPROFILE_ALL_MBS

  75. };

  76. //@}

  77. 

  78. /** @name Where quant can be changed

  79.  */

  80. //@{

  81. enum DQSingleEdge {

  82.     DQSINGLE_BEDGE_LEFT,

  83.     DQSINGLE_BEDGE_TOP,

  84.     DQSINGLE_BEDGE_RIGHT,

  85.     DQSINGLE_BEDGE_BOTTOM

  86. };

  87. //@}

  88. 

  89. /** Which pair of edges is quantized with ALTPQUANT */

  90. //@{

  91. enum DQDoubleEdge {

  92.     DQDOUBLE_BEDGE_TOPLEFT,

  93.     DQDOUBLE_BEDGE_TOPRIGHT,

  94.     DQDOUBLE_BEDGE_BOTTOMRIGHT,

  95.     DQDOUBLE_BEDGE_BOTTOMLEFT

  96. };

  97. //@}

  98. 

  99. /** MV modes for P frames */

  100. //@{

  101. enum MVModes {

  102.     MV_PMODE_1MV_HPEL_BILIN,

  103.     MV_PMODE_1MV,

  104.     MV_PMODE_1MV_HPEL,

  105.     MV_PMODE_MIXED_MV,

  106.     MV_PMODE_INTENSITY_COMP

  107. };

  108. //@}

  109. 

  110. /** @name MV types for B frames */

  111. //@{

  112. enum BMVTypes {

  113.     BMV_TYPE_BACKWARD,

  114.     BMV_TYPE_FORWARD,

  115.     BMV_TYPE_INTERPOLATED = 3 //XXX: ??

  116. };

  117. //@}

  118. 

  119. /** @name Block types for P/B frames */

  120. //@{

  121. enum TransformTypes {

  122.     TT_8X8,

  123.     TT_8X4_BOTTOM,

  124.     TT_8X4_TOP,

  125.     TT_8X4, //Both halves

  126.     TT_4X8_RIGHT,

  127.     TT_4X8_LEFT,

  128.     TT_4X8, //Both halves

  129.     TT_4X4

  130. };

  131. //@}

  132. 

  133. /** Table for conversion between TTBLK and TTMB */

  134. static const int ttblk_to_tt[3][8] = {

  135.   { TT_8X4, TT_4X8, TT_8X8, TT_4X4, TT_8X4_TOP, TT_8X4_BOTTOM, TT_4X8_RIGHT, TT_4X8_LEFT },

  136.   { TT_8X8, TT_4X8_RIGHT, TT_4X8_LEFT, TT_4X4, TT_8X4, TT_4X8, TT_8X4_BOTTOM, TT_8X4_TOP },

  137.   { TT_8X8, TT_4X8, TT_4X4, TT_8X4_BOTTOM, TT_4X8_RIGHT, TT_4X8_LEFT, TT_8X4, TT_8X4_TOP }

  138. };

  139. 

  140. static const int ttfrm_to_tt[4] = { TT_8X8, TT_8X4, TT_4X8, TT_4X4 };

  141. 

  142. /** MV P mode - the 5th element is only used for mode 1 */

  143. static const uint8_t mv_pmode_table[2][5] = {

  144.   { MV_PMODE_1MV_HPEL_BILIN, MV_PMODE_1MV, MV_PMODE_1MV_HPEL, MV_PMODE_INTENSITY_COMP, MV_PMODE_MIXED_MV },

  145.   { MV_PMODE_1MV, MV_PMODE_MIXED_MV, MV_PMODE_1MV_HPEL, MV_PMODE_INTENSITY_COMP, MV_PMODE_1MV_HPEL_BILIN }

  146. };

  147. static const uint8_t mv_pmode_table2[2][4] = {

  148.   { MV_PMODE_1MV_HPEL_BILIN, MV_PMODE_1MV, MV_PMODE_1MV_HPEL, MV_PMODE_MIXED_MV },

  149.   { MV_PMODE_1MV, MV_PMODE_MIXED_MV, MV_PMODE_1MV_HPEL, MV_PMODE_1MV_HPEL_BILIN }

  150. };

  151. 

  152. /** One more frame type */

  153. #define BI_TYPE 7

  154. 

  155. static const int fps_nr[5] = { 24, 25, 30, 50, 60 },

  156.   fps_dr[2] = { 1000, 1001 };

  157. static const uint8_t pquant_table[3][32] = {

  158.   {  /* Implicit quantizer */

  159.      0,  1,  2,  3,  4,  5,  6,  7,  8,  6,  7,  8,  9, 10, 11, 12,

  160.     13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 27, 29, 31

  161.   },

  162.   {  /* Explicit quantizer, pquantizer uniform */

  163.      0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15,

  164.     16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31

  165.   },

  166.   {  /* Explicit quantizer, pquantizer non-uniform */

  167.      0,  1,  1,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13,

  168.     14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 29, 31

  169.   }

  170. };

  171. 

  172. /** @name VC-1 VLC tables and defines

  173.  *  @todo TODO move this into the context

  174.  */

  175. //@{

  176. #define VC1_BFRACTION_VLC_BITS 7

  177. static VLC vc1_bfraction_vlc;

  178. #define VC1_IMODE_VLC_BITS 4

  179. static VLC vc1_imode_vlc;

  180. #define VC1_NORM2_VLC_BITS 3

  181. static VLC vc1_norm2_vlc;

  182. #define VC1_NORM6_VLC_BITS 9

  183. static VLC vc1_norm6_vlc;

  184. /* Could be optimized, one table only needs 8 bits */

  185. #define VC1_TTMB_VLC_BITS 9 //12

  186. static VLC vc1_ttmb_vlc[3];

  187. #define VC1_MV_DIFF_VLC_BITS 9 //15

  188. static VLC vc1_mv_diff_vlc[4];

  189. #define VC1_CBPCY_P_VLC_BITS 9 //14

  190. static VLC vc1_cbpcy_p_vlc[4];

  191. #define VC1_4MV_BLOCK_PATTERN_VLC_BITS 6

  192. static VLC vc1_4mv_block_pattern_vlc[4];

  193. #define VC1_TTBLK_VLC_BITS 5

  194. static VLC vc1_ttblk_vlc[3];

  195. #define VC1_SUBBLKPAT_VLC_BITS 6

  196. static VLC vc1_subblkpat_vlc[3];

  197. 

  198. static VLC vc1_ac_coeff_table[8];

  199. //@}

  200. 

  201. enum CodingSet {

  202.     CS_HIGH_MOT_INTRA = 0,

  203.     CS_HIGH_MOT_INTER,

  204.     CS_LOW_MOT_INTRA,

  205.     CS_LOW_MOT_INTER,

  206.     CS_MID_RATE_INTRA,

  207.     CS_MID_RATE_INTER,

  208.     CS_HIGH_RATE_INTRA,

  209.     CS_HIGH_RATE_INTER

  210. };

  211. 

  212. /** The VC1 Context

  213.  * @fixme Change size wherever another size is more efficient

  214.  * Many members are only used for Advanced Profile

  215.  */

  216. typedef struct VC1Context{

  217.     MpegEncContext s;

  218. 

  219.     int bits;

  220. 

  221.     /** Simple/Main Profile sequence header */

  222.     //@{

  223.     int res_sm;           ///< reserved, 2b

  224.     int res_x8;           ///< reserved

  225.     int multires;         ///< frame-level RESPIC syntax element present

  226.     int res_fasttx;       ///< reserved, always 1

  227.     int res_transtab;     ///< reserved, always 0

  228.     int rangered;         ///< RANGEREDFRM (range reduction) syntax element present

  229.                           ///< at frame level

  230.     int res_rtm_flag;     ///< reserved, set to 1

  231.     int reserved;         ///< reserved

  232.     //@}

  233. 

  234.     /** Advanced Profile */

  235.     //@{

  236.     int level;            ///< 3bits, for Advanced/Simple Profile, provided by TS layer

  237.     int chromaformat;     ///< 2bits, 2=4:2:0, only defined

  238.     int postprocflag;     ///< Per-frame processing suggestion flag present

  239.     int broadcast;        ///< TFF/RFF present

  240.     int interlace;        ///< Progressive/interlaced (RPTFTM syntax element)

  241.     int tfcntrflag;       ///< TFCNTR present

  242.     int panscanflag;      ///< NUMPANSCANWIN, TOPLEFT{X,Y}, BOTRIGHT{X,Y} present

  243.     int extended_dmv;     ///< Additional extended dmv range at P/B frame-level

  244.     int color_prim;       ///< 8bits, chroma coordinates of the color primaries

  245.     int transfer_char;    ///< 8bits, Opto-electronic transfer characteristics

  246.     int matrix_coef;      ///< 8bits, Color primaries->YCbCr transform matrix

  247.     int hrd_param_flag;   ///< Presence of Hypothetical Reference

  248.                           ///< Decoder parameters

  249.     //@}

  250. 

  251.     /** Sequence header data for all Profiles

  252.      * TODO: choose between ints, uint8_ts and monobit flags

  253.      */

  254.     //@{

  255.     int profile;          ///< 2bits, Profile

  256.     int frmrtq_postproc;  ///< 3bits,

  257.     int bitrtq_postproc;  ///< 5bits, quantized framerate-based postprocessing strength

  258.     int fastuvmc;         ///< Rounding of qpel vector to hpel ? (not in Simple)

  259.     int extended_mv;      ///< Ext MV in P/B (not in Simple)

  260.     int dquant;           ///< How qscale varies with MBs, 2bits (not in Simple)

  261.     int vstransform;      ///< variable-size [48]x[48] transform type + info

  262.     int overlap;          ///< overlapped transforms in use

  263.     int quantizer_mode;   ///< 2bits, quantizer mode used for sequence, see QUANT_*

  264.     int finterpflag;      ///< INTERPFRM present

  265.     //@}

  266. 

  267.     /** Frame decoding info for all profiles */

  268.     //@{

  269.     uint8_t mv_mode;      ///< MV coding monde

  270.     uint8_t mv_mode2;     ///< Secondary MV coding mode (B frames)

  271.     int k_x;              ///< Number of bits for MVs (depends on MV range)

  272.     int k_y;              ///< Number of bits for MVs (depends on MV range)

  273.     int range_x, range_y; ///< MV range

  274.     uint8_t pq, altpq;    ///< Current/alternate frame quantizer scale

  275.     /** pquant parameters */

  276.     //@{

  277.     uint8_t dquantfrm;

  278.     uint8_t dqprofile;

  279.     uint8_t dqsbedge;

  280.     uint8_t dqbilevel;

  281.     //@}

  282.     /** AC coding set indexes

  283.      * @see 8.1.1.10, p(1)10

  284.      */

  285.     //@{

  286.     int c_ac_table_index; ///< Chroma index from ACFRM element

  287.     int y_ac_table_index; ///< Luma index from AC2FRM element

  288.     //@}

  289.     int ttfrm;            ///< Transform type info present at frame level

  290.     uint8_t ttmbf;        ///< Transform type flag

  291.     uint8_t ttblk4x4;     ///< Value of ttblk which indicates a 4x4 transform

  292.     int codingset;        ///< index of current table set from 11.8 to use for luma block decoding

  293.     int codingset2;       ///< index of current table set from 11.8 to use for chroma block decoding

  294.     int pqindex;          ///< raw pqindex used in coding set selection

  295.     int a_avail, c_avail;

  296.     uint8_t *mb_type_base, *mb_type[3];

  297. 

  298. 

  299.     /** Luma compensation parameters */

  300.     //@{

  301.     uint8_t lumscale;

  302.     uint8_t lumshift;

  303.     //@}

  304.     int16_t bfraction;    ///< Relative position % anchors=> how to scale MVs

  305.     uint8_t halfpq;       ///< Uniform quant over image and qp+.5

  306.     uint8_t respic;       ///< Frame-level flag for resized images

  307.     int buffer_fullness;  ///< HRD info

  308.     /** Ranges:

  309.      * -# 0 -> [-64n 63.f] x [-32, 31.f]

  310.      * -# 1 -> [-128, 127.f] x [-64, 63.f]

  311.      * -# 2 -> [-512, 511.f] x [-128, 127.f]

  312.      * -# 3 -> [-1024, 1023.f] x [-256, 255.f]

  313.      */

  314.     uint8_t mvrange;

  315.     uint8_t pquantizer;           ///< Uniform (over sequence) quantizer in use

  316.     VLC *cbpcy_vlc;               ///< CBPCY VLC table

  317.     int tt_index;                 ///< Index for Transform Type tables

  318.     uint8_t* mv_type_mb_plane;    ///< bitplane for mv_type == (4MV)

  319. //    BitPlane direct_mb_plane;     ///< bitplane for "direct" MBs

  320.     int mv_type_is_raw;           ///< mv type mb plane is not coded

  321.     int skip_is_raw;              ///< skip mb plane is not coded

  322.     uint8_t luty[256], lutuv[256]; // lookup tables used for intensity compensation

  323.     int rnd;                      ///< rounding control

  324. 

  325.     /** Frame decoding info for S/M profiles only */

  326.     //@{

  327.     uint8_t rangeredfrm; ///< out_sample = CLIP((in_sample-128)*2+128)

  328.     uint8_t interpfrm;

  329.     //@}

  330. 

  331.     /** Frame decoding info for Advanced profile */

  332.     //@{

  333.     uint8_t fcm; ///< 0->Progressive, 2->Frame-Interlace, 3->Field-Interlace

  334.     uint8_t numpanscanwin;

  335.     uint8_t tfcntr;

  336.     uint8_t rptfrm, tff, rff;

  337.     uint16_t topleftx;

  338.     uint16_t toplefty;

  339.     uint16_t bottomrightx;

  340.     uint16_t bottomrighty;

  341.     uint8_t uvsamp;

  342.     uint8_t postproc;

  343.     int hrd_num_leaky_buckets;

  344.     uint8_t bit_rate_exponent;

  345.     uint8_t buffer_size_exponent;

  346. //    BitPlane ac_pred_plane;       ///< AC prediction flags bitplane

  347. //    BitPlane over_flags_plane;    ///< Overflags bitplane

  348.     uint8_t condover;

  349.     uint16_t *hrd_rate, *hrd_buffer;

  350.     uint8_t *hrd_fullness;

  351.     uint8_t range_mapy_flag;

  352.     uint8_t range_mapuv_flag;

  353.     uint8_t range_mapy;

  354.     uint8_t range_mapuv;

  355.     //@}

  356. } VC1Context;

  357. 

  358. /**

  359.  * Get unary code of limited length

  360.  * @fixme FIXME Slow and ugly

  361.  * @param gb GetBitContext

  362.  * @param[in] stop The bitstop value (unary code of 1's or 0's)

  363.  * @param[in] len Maximum length

  364.  * @return Unary length/index

  365.  */

  366. static int get_prefix(GetBitContext *gb, int stop, int len)

  367. {

  368. #if 1

  369.     int i;

  370. 

  371.     for(i = 0; i < len && get_bits1(gb) != stop; i++);

  372.     return i;

  373. /*  int i = 0, tmp = !stop;

  374. 

  375.   while (i != len && tmp != stop)

  376.   {

  377.     tmp = get_bits(gb, 1);

  378.     i++;

  379.   }

  380.   if (i == len && tmp != stop) return len+1;

  381.   return i;*/

  382. #else

  383.   unsigned int buf;

  384.   int log;

  385. 

  386.   OPEN_READER(re, gb);

  387.   UPDATE_CACHE(re, gb);

  388.   buf=GET_CACHE(re, gb); //Still not sure

  389.   if (stop) buf = ~buf;

  390. 

  391.   log= av_log2(-buf); //FIXME: -?

  392.   if (log < limit){

  393.     LAST_SKIP_BITS(re, gb, log+1);

  394.     CLOSE_READER(re, gb);

  395.     return log;

  396.   }

  397. 

  398.   LAST_SKIP_BITS(re, gb, limit);

  399.   CLOSE_READER(re, gb);

  400.   return limit;

  401. #endif

  402. }

  403. 

  404. static inline int decode210(GetBitContext *gb){

  405.     int n;

  406.     n = get_bits1(gb);

  407.     if (n == 1)

  408.         return 0;

  409.     else

  410.         return 2 - get_bits1(gb);

  411. }

  412. 

  413. /**

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

  415.  * @param v The VC1Context to initialize

  416.  * @return Status

  417.  */

  418. static int vc1_init_common(VC1Context *v)

  419. {

  420.     static int done = 0;

  421.     int i = 0;

  422. 

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

  424. 

  425.     /* VLC tables */

  426.     if(!done)

  427.     {

  428.         done = 1;

  429.         init_vlc(&vc1_bfraction_vlc, VC1_BFRACTION_VLC_BITS, 23,

  430.                  vc1_bfraction_bits, 1, 1,

  431.                  vc1_bfraction_codes, 1, 1, 1);

  432.         init_vlc(&vc1_norm2_vlc, VC1_NORM2_VLC_BITS, 4,

  433.                  vc1_norm2_bits, 1, 1,

  434.                  vc1_norm2_codes, 1, 1, 1);

  435.         init_vlc(&vc1_norm6_vlc, VC1_NORM6_VLC_BITS, 64,

  436.                  vc1_norm6_bits, 1, 1,

  437.                  vc1_norm6_codes, 2, 2, 1);

  438.         init_vlc(&vc1_imode_vlc, VC1_IMODE_VLC_BITS, 7,

  439.                  vc1_imode_bits, 1, 1,

  440.                  vc1_imode_codes, 1, 1, 1);

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

  442.         {

  443.             init_vlc(&vc1_ttmb_vlc[i], VC1_TTMB_VLC_BITS, 16,

  444.                      vc1_ttmb_bits[i], 1, 1,

  445.                      vc1_ttmb_codes[i], 2, 2, 1);

  446.             init_vlc(&vc1_ttblk_vlc[i], VC1_TTBLK_VLC_BITS, 8,

  447.                      vc1_ttblk_bits[i], 1, 1,

  448.                      vc1_ttblk_codes[i], 1, 1, 1);

  449.             init_vlc(&vc1_subblkpat_vlc[i], VC1_SUBBLKPAT_VLC_BITS, 15,

  450.                      vc1_subblkpat_bits[i], 1, 1,

  451.                      vc1_subblkpat_codes[i], 1, 1, 1);

  452.         }

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

  454.         {

  455.             init_vlc(&vc1_4mv_block_pattern_vlc[i], VC1_4MV_BLOCK_PATTERN_VLC_BITS, 16,

  456.                      vc1_4mv_block_pattern_bits[i], 1, 1,

  457.                      vc1_4mv_block_pattern_codes[i], 1, 1, 1);

  458.             init_vlc(&vc1_cbpcy_p_vlc[i], VC1_CBPCY_P_VLC_BITS, 64,

  459.                      vc1_cbpcy_p_bits[i], 1, 1,

  460.                      vc1_cbpcy_p_codes[i], 2, 2, 1);

  461.             init_vlc(&vc1_mv_diff_vlc[i], VC1_MV_DIFF_VLC_BITS, 73,

  462.                      vc1_mv_diff_bits[i], 1, 1,

  463.                      vc1_mv_diff_codes[i], 2, 2, 1);

  464.         }

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

  466.             init_vlc(&vc1_ac_coeff_table[i], AC_VLC_BITS, vc1_ac_sizes[i],

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

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

  469.         init_vlc(&ff_msmp4_mb_i_vlc, MB_INTRA_VLC_BITS, 64,

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

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

  472.     }

  473. 

  474.     /* Other defaults */

  475.     v->pq = -1;

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

  477. 

  478.     return 0;

  479. }

  480. 

  481. /***********************************************************************/

  482. /**

  483.  * @defgroup bitplane VC9 Bitplane decoding

  484.  * @see 8.7, p56

  485.  * @{

  486.  */

  487. 

  488. /** @addtogroup bitplane

  489.  * Imode types

  490.  * @{

  491.  */

  492. enum Imode {

  493.     IMODE_RAW,

  494.     IMODE_NORM2,

  495.     IMODE_DIFF2,

  496.     IMODE_NORM6,

  497.     IMODE_DIFF6,

  498.     IMODE_ROWSKIP,

  499.     IMODE_COLSKIP

  500. };

  501. /** @} */ //imode defines

  502. 

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

  504.  * @param plane Buffer to store decoded bits

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

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

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

  508.  */

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

  510.     int x, y;

  511. 

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

  513.         if (!get_bits(gb, 1)) //rowskip

  514.             memset(plane, 0, width);

  515.         else

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

  517.                 plane[x] = get_bits(gb, 1);

  518.         plane += stride;

  519.     }

  520. }

  521. 

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

  523.  * @param plane Buffer to store decoded bits

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

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

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

  527.  * @fixme FIXME: Optimize

  528.  */

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

  530.     int x, y;

  531. 

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

  533.         if (!get_bits(gb, 1)) //colskip

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

  535.                 plane[y*stride] = 0;

  536.         else

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

  538.                 plane[y*stride] = get_bits(gb, 1);

  539.         plane ++;

  540.     }

  541. }

  542. 

  543. /** Decode a bitplane's bits

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

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

  546.  * @return Status

  547.  * @fixme FIXME: Optimize

  548.  * @todo TODO: Decide if a struct is needed

  549.  */

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

  551. {

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

  553. 

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

  555.     uint8_t invert, *planep = data;

  556.     int width, height, stride;

  557. 

  558.     width = v->s.mb_width;

  559.     height = v->s.mb_height;

  560.     stride = v->s.mb_stride;

  561.     invert = get_bits(gb, 1);

  562.     imode = get_vlc2(gb, vc1_imode_vlc.table, VC1_IMODE_VLC_BITS, 1);

  563. 

  564.     *raw_flag = 0;

  565.     switch (imode)

  566.     {

  567.     case IMODE_RAW:

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

  569.         *raw_flag = 1; //invert ignored

  570.         return invert;

  571.     case IMODE_DIFF2:

  572.     case IMODE_NORM2:

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

  574.         {

  575.             *planep++ = get_bits(gb, 1);

  576.             offset = 1;

  577.         }

  578.         else offset = 0;

  579.         // decode bitplane as one long line

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

  581.             code = get_vlc2(gb, vc1_norm2_vlc.table, VC1_NORM2_VLC_BITS, 1);

  582.             *planep++ = code & 1;

  583.             offset++;

  584.             if(offset == width) {

  585.                 offset = 0;

  586.                 planep += stride - width;

  587.             }

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

  589.             offset++;

  590.             if(offset == width) {

  591.                 offset = 0;

  592.                 planep += stride - width;

  593.             }

  594.         }

  595.         break;

  596.     case IMODE_DIFF6:

  597.     case IMODE_NORM6:

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

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

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

  601.                     code = get_vlc2(gb, vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);

  602.                     if(code < 0){

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

  604.                         return -1;

  605.                     }

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

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

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

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

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

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

  612.                 }

  613.                 planep += stride * 3;

  614.             }

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

  616.         } else { // 3x2

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

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

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

  620.                     code = get_vlc2(gb, vc1_norm6_vlc.table, VC1_NORM6_VLC_BITS, 2);

  621.                     if(code < 0){

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

  623.                         return -1;

  624.                     }

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

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

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

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

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

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

  631.                 }

  632.                 planep += stride * 2;

  633.             }

  634.             x = width % 3;

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

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

  637.         }

  638.         break;

  639.     case IMODE_ROWSKIP:

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

  641.         break;

  642.     case IMODE_COLSKIP:

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

  644.         break;

  645.     default: break;

  646.     }

  647. 

  648.     /* Applying diff operator */

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

  650.     {

  651.         planep = data;

  652.         planep[0] ^= invert;

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

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

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

  656.         {

  657.             planep += stride;

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

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

  660.             {

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

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

  663.             }

  664.         }

  665.     }

  666.     else if (invert)

  667.     {

  668.         planep = data;

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

  670.     }

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

  672. }

  673. 

  674. /** @} */ //Bitplane group

  675. 

  676. /***********************************************************************/

  677. /** VOP Dquant decoding

  678.  * @param v VC-1 Context

  679.  */

  680. static int vop_dquant_decoding(VC1Context *v)

  681. {

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

  683.     int pqdiff;

  684. 

  685.     //variable size

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

  687.     {

  688.         pqdiff = get_bits(gb, 3);

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

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

  691.     }

  692.     else

  693.     {

  694.         v->dquantfrm = get_bits(gb, 1);

  695.         if ( v->dquantfrm )

  696.         {

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

  698.             switch (v->dqprofile)

  699.             {

  700.             case DQPROFILE_SINGLE_EDGE:

  701.             case DQPROFILE_DOUBLE_EDGES:

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

  703.                 break;

  704.             case DQPROFILE_ALL_MBS:

  705.                 v->dqbilevel = get_bits(gb, 1);

  706.             default: break; //Forbidden ?

  707.             }

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

  709.             {

  710.                 pqdiff = get_bits(gb, 3);

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

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

  713.             }

  714.         }

  715.     }

  716.     return 0;

  717. }

  718. 

  719. 

  720. /** Do inverse transform

  721.  */

  722. static void vc1_inv_trans(DCTELEM block[64], int M, int N)

  723. {

  724.     int i;

  725.     register int t1,t2,t3,t4,t5,t6,t7,t8;

  726.     DCTELEM *src, *dst;

  727. 

  728.     src = block;

  729.     dst = block;

  730.     if(M==4){

  731.         for(i = 0; i < N; i++){

  732.             t1 = 17 * (src[0] + src[2]);

  733.             t2 = 17 * (src[0] - src[2]);

  734.             t3 = 22 * src[1];

  735.             t4 = 22 * src[3];

  736.             t5 = 10 * src[1];

  737.             t6 = 10 * src[3];

  738. 

  739.             dst[0] = (t1 + t3 + t6 + 4) >> 3;

  740.             dst[1] = (t2 - t4 + t5 + 4) >> 3;

  741.             dst[2] = (t2 + t4 - t5 + 4) >> 3;

  742.             dst[3] = (t1 - t3 - t6 + 4) >> 3;

  743. 

  744.             src += 8;

  745.             dst += 8;

  746.         }

  747.     }else{

  748.         for(i = 0; i < N; i++){

  749.             t1 = 12 * (src[0] + src[4]);

  750.             t2 = 12 * (src[0] - src[4]);

  751.             t3 = 16 * src[2] +  6 * src[6];

  752.             t4 =  6 * src[2] - 16 * src[6];

  753. 

  754.             t5 = t1 + t3;

  755.             t6 = t2 + t4;

  756.             t7 = t2 - t4;

  757.             t8 = t1 - t3;

  758. 

  759.             t1 = 16 * src[1] + 15 * src[3] +  9 * src[5] +  4 * src[7];

  760.             t2 = 15 * src[1] -  4 * src[3] - 16 * src[5] -  9 * src[7];

  761.             t3 =  9 * src[1] - 16 * src[3] +  4 * src[5] + 15 * src[7];

  762.             t4 =  4 * src[1] -  9 * src[3] + 15 * src[5] - 16 * src[7];

  763. 

  764.             dst[0] = (t5 + t1 + 4) >> 3;

  765.             dst[1] = (t6 + t2 + 4) >> 3;

  766.             dst[2] = (t7 + t3 + 4) >> 3;

  767.             dst[3] = (t8 + t4 + 4) >> 3;

  768.             dst[4] = (t8 - t4 + 4) >> 3;

  769.             dst[5] = (t7 - t3 + 4) >> 3;

  770.             dst[6] = (t6 - t2 + 4) >> 3;

  771.             dst[7] = (t5 - t1 + 4) >> 3;

  772. 

  773.             src += 8;

  774.             dst += 8;

  775.         }

  776.     }

  777. 

  778.     src = block;

  779.     dst = block;

  780.     if(N==4){

  781.         for(i = 0; i < M; i++){

  782.             t1 = 17 * (src[ 0] + src[16]);

  783.             t2 = 17 * (src[ 0] - src[16]);

  784.             t3 = 22 * src[ 8];

  785.             t4 = 22 * src[24];

  786.             t5 = 10 * src[ 8];

  787.             t6 = 10 * src[24];

  788. 

  789.             dst[ 0] = (t1 + t3 + t6 + 64) >> 7;

  790.             dst[ 8] = (t2 - t4 + t5 + 64) >> 7;

  791.             dst[16] = (t2 + t4 - t5 + 64) >> 7;

  792.             dst[24] = (t1 - t3 - t6 + 64) >> 7;

  793. 

  794.             src ++;

  795.             dst ++;

  796.         }

  797.     }else{

  798.         for(i = 0; i < M; i++){

  799.             t1 = 12 * (src[ 0] + src[32]);

  800.             t2 = 12 * (src[ 0] - src[32]);

  801.             t3 = 16 * src[16] +  6 * src[48];

  802.             t4 =  6 * src[16] - 16 * src[48];

  803. 

  804.             t5 = t1 + t3;

  805.             t6 = t2 + t4;

  806.             t7 = t2 - t4;

  807.             t8 = t1 - t3;

  808. 

  809.             t1 = 16 * src[ 8] + 15 * src[24] +  9 * src[40] +  4 * src[56];

  810.             t2 = 15 * src[ 8] -  4 * src[24] - 16 * src[40] -  9 * src[56];

  811.             t3 =  9 * src[ 8] - 16 * src[24] +  4 * src[40] + 15 * src[56];

  812.             t4 =  4 * src[ 8] -  9 * src[24] + 15 * src[40] - 16 * src[56];

  813. 

  814.             dst[ 0] = (t5 + t1 + 64) >> 7;

  815.             dst[ 8] = (t6 + t2 + 64) >> 7;

  816.             dst[16] = (t7 + t3 + 64) >> 7;

  817.             dst[24] = (t8 + t4 + 64) >> 7;

  818.             dst[32] = (t8 - t4 + 64 + 1) >> 7;

  819.             dst[40] = (t7 - t3 + 64 + 1) >> 7;

  820.             dst[48] = (t6 - t2 + 64 + 1) >> 7;

  821.             dst[56] = (t5 - t1 + 64 + 1) >> 7;

  822. 

  823.             src++;

  824.             dst++;

  825.         }

  826.     }

  827. }

  828. 

  829. /** Apply overlap transform to vertical edge

  830.  * @todo optimize

  831.  * @todo move to DSPContext

  832.  */

  833. static void vc1_v_overlap(uint8_t* src, int stride)

  834. {

  835.     int i;

  836.     int a, b, c, d;

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

  838.         a = src[-2*stride];

  839.         b = src[-stride];

  840.         c = src[0];

  841.         d = src[stride];

  842. 

  843.         src[-2*stride] = clip_uint8((7*a + d + 3) >> 3);

  844.         src[-stride] = clip_uint8((-a + 7*b + c + d + 3) >> 3);

  845.         src[0] = clip_uint8((a + b + 7*c - d + 3) >> 3);

  846.         src[stride] = clip_uint8((a + 7*d + 3) >> 3);

  847.         src++;

  848.     }

  849. }

  850. 

  851. /** Apply overlap transform to horizontal edge

  852.  * @todo optimize

  853.  * @todo move to DSPContext

  854.  */

  855. static void vc1_h_overlap(uint8_t* src, int stride)

  856. {

  857.     int i;

  858.     int a, b, c, d;

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

  860.         a = src[-2];

  861.         b = src[-1];

  862.         c = src[0];

  863.         d = src[1];

  864. 

  865.         src[-2] = clip_uint8((7*a + d + 3) >> 3);

  866.         src[-1] = clip_uint8((-a + 7*b + c + d + 3) >> 3);

  867.         src[0] = clip_uint8((a + b + 7*c - d + 3) >> 3);

  868.         src[1] = clip_uint8((a + 7*d + 3) >> 3);

  869.         src += stride;

  870.     }

  871. }

  872. 

  873. /** Put block onto picture

  874.  * @todo move to DSPContext

  875.  */

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

  877. {

  878.     uint8_t *Y;

  879.     int ys, us, vs;

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

  881. 

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

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

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

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

  886. 

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

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

  889.     Y += ys * 8;

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

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

  892. 

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

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

  895. }

  896. 

  897. /** Do motion compensation over 1 macroblock

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

  899.  */

  900. static void vc1_mc_1mv(VC1Context *v)

  901. {

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

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

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

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

  906. 

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

  908. 

  909.     mx = s->mv[0][0][0];

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

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

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

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

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

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

  916. 

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

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

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

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

  921. 

  922.     src_x   = clip(  src_x, -16, s->mb_width  * 16);

  923.     src_y   = clip(  src_y, -16, s->mb_height * 16);

  924.     uvsrc_x = clip(uvsrc_x,  -8, s->mb_width  *  8);

  925.     uvsrc_y = clip(uvsrc_y,  -8, s->mb_height *  8);

  926. 

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

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

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

  930. 

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

  932.        || (unsigned)src_x > s->h_edge_pos - (mx&3) - 16

  933.        || (unsigned)src_y > s->v_edge_pos - (my&3) - 16){

  934.         uint8_t *uvbuf= s->edge_emu_buffer + 18 * s->linesize;

  935. 

  936.         ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 16+1, 16+1,

  937.                             src_x, src_y, s->h_edge_pos, s->v_edge_pos);

  938.         srcY = s->edge_emu_buffer;

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

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

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

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

  943.         srcU = uvbuf;

  944.         srcV = uvbuf + 16;

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

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

  947.             int i, j;

  948.             uint8_t *src, *src2;

  949. 

  950.             src = srcY;

  951.             for(j = 0; j < 17; j++) {

  952.                 for(i = 0; i < 17; i++) src[i] = v->luty[src[i]];

  953.                 src += s->linesize;

  954.             }

  955.             src = srcU; src2 = srcV;

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

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

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

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

  960.                 }

  961.                 src += s->uvlinesize;

  962.                 src2 += s->uvlinesize;

  963.             }

  964.         }

  965.     }

  966. 

  967.     if(v->fastuvmc) {

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

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

  970.     }

  971. 

  972.     if(!s->quarter_sample) { // hpel mc

  973.         mx >>= 1;

  974.         my >>= 1;

  975.         dxy = ((my & 1) << 1) | (mx & 1);

  976. 

  977.         if(!v->rnd)

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

  979.         else

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

  981.     } else {

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

  983. 

  984.         if(!v->rnd)

  985.             dsp->put_qpel_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize);

  986.         else

  987.             dsp->put_no_rnd_qpel_pixels_tab[0][dxy](s->dest[0], srcY, s->linesize);

  988.     }

  989.     /* Chroma MC always uses qpel blilinear */

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

  991.     if(!v->rnd){

  992.         dsp->put_qpel_pixels_tab[1][uvdxy](s->dest[1], srcU, s->uvlinesize);

  993.         dsp->put_qpel_pixels_tab[1][uvdxy](s->dest[2], srcV, s->uvlinesize);

  994.     }else{

  995.         dsp->put_no_rnd_qpel_pixels_tab[1][uvdxy](s->dest[1], srcU, s->uvlinesize);

  996.         dsp->put_no_rnd_qpel_pixels_tab[1][uvdxy](s->dest[2], srcV, s->uvlinesize);

  997.     }

  998. }

  999. 

  1000. /** Do motion compensation for 4-MV macroblock - luminance block

  1001.  */

  1002. static void vc1_mc_4mv_luma(VC1Context *v, int n)

  1003. {

  1004.     MpegEncContext *s = &v->s;

  1005.     DSPContext *dsp = &v->s.dsp;

  1006.     uint8_t *srcY;

  1007.     int dxy, mx, my, src_x, src_y;

  1008.     int off;

  1009. 

  1010.     if(!v->s.last_picture.data[0])return;

  1011.     mx = s->mv[0][n][0];

  1012.     my = s->mv[0][n][1];

  1013.     srcY = s->last_picture.data[0];

  1014. 

  1015.     off = s->linesize * 4 * (n&2) + (n&1) * 8;

  1016. 

  1017.     src_x = s->mb_x * 16 + (n&1) * 8 + (mx >> 2);

  1018.     src_y = s->mb_y * 16 + (n&2) * 4 + (my >> 2);

  1019. 

  1020.     src_x   = clip(  src_x, -16, s->mb_width  * 16);

  1021.     src_y   = clip(  src_y, -16, s->mb_height * 16);

  1022. 

  1023.     srcY += src_y * s->linesize + src_x;

  1024. 

  1025.     if((unsigned)src_x > s->h_edge_pos - (mx&3) - 16

  1026.        || (unsigned)src_y > s->v_edge_pos - (my&3) - 16){

  1027.         ff_emulated_edge_mc(s->edge_emu_buffer, srcY, s->linesize, 16+1, 16+1,

  1028.                             src_x, src_y, s->h_edge_pos, s->v_edge_pos);

  1029.         srcY = s->edge_emu_buffer;

  1030.     }

  1031. 

  1032.     if(!s->quarter_sample) { // hpel mc

  1033.         mx >>= 1;

  1034.         my >>= 1;

  1035.         dxy = ((my & 1) << 1) | (mx & 1);

  1036. 

  1037.         if(!v->rnd)

  1038.             dsp->put_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);

  1039.         else

  1040.             dsp->put_no_rnd_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize, 8);

  1041.     } else {

  1042.         dxy = ((my & 3) << 2) | (mx & 3);

  1043. 

  1044.         if(!v->rnd)

  1045.             dsp->put_qpel_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize);

  1046.         else

  1047.             dsp->put_no_rnd_qpel_pixels_tab[1][dxy](s->dest[0] + off, srcY, s->linesize);

  1048.     }

  1049. }

  1050. 

  1051. static inline int median4(int a, int b, int c, int d)

  1052. {

  1053.     if(a < b) {

  1054.         if(c < d) return (FFMIN(b, d) + FFMAX(a, c)) / 2;

  1055.         else      return (FFMIN(b, c) + FFMAX(a, d)) / 2;

  1056.     } else {

  1057.         if(c < d) return (FFMIN(a, d) + FFMAX(b, c)) / 2;

  1058.         else      return (FFMIN(a, c) + FFMAX(b, d)) / 2;

  1059.     }

  1060. }

  1061. 

  1062. 

  1063. /** Do motion compensation for 4-MV macroblock - both chroma blocks

  1064.  */

  1065. static void vc1_mc_4mv_chroma(VC1Context *v)

  1066. {

  1067.     MpegEncContext *s = &v->s;

  1068.     DSPContext *dsp = &v->s.dsp;

  1069.     uint8_t *srcU, *srcV;

  1070.     int uvdxy, uvmx, uvmy, uvsrc_x, uvsrc_y;

  1071.     int i, idx, tx = 0, ty = 0;

  1072.     int mvx[4], mvy[4], intra[4];

  1073.     static const int count[16] = { 0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4};

  1074. 

  1075.     if(!v->s.last_picture.data[0])return;

  1076. 

  1077.     for(i = 0; i < 4; i++) {

  1078.         mvx[i] = s->mv[0][i][0];

  1079.         mvy[i] = s->mv[0][i][1];

  1080.         intra[i] = v->mb_type[0][s->block_index[i]];

  1081.     }

  1082. 

  1083.     /* calculate chroma MV vector from four luma MVs */

  1084.     idx = (intra[3] << 3) | (intra[2] << 2) | (intra[1] << 1) | intra[0];

  1085.     if(!idx) { // all blocks are inter

  1086.         tx = median4(mvx[0], mvx[1], mvx[2], mvx[3]);

  1087.         ty = median4(mvy[0], mvy[1], mvy[2], mvy[3]);

  1088.     } else if(count[idx] == 1) { // 3 inter blocks

  1089.         switch(idx) {

  1090.         case 0x1:

  1091.             tx = mid_pred(mvx[1], mvx[2], mvx[3]);

  1092.             ty = mid_pred(mvy[1], mvy[2], mvy[3]);

  1093.             break;

  1094.         case 0x2:

  1095.             tx = mid_pred(mvx[0], mvx[2], mvx[3]);

  1096.             ty = mid_pred(mvy[0], mvy[2], mvy[3]);

  1097.             break;

  1098.         case 0x4:

  1099.             tx = mid_pred(mvx[0], mvx[1], mvx[3]);

  1100.             ty = mid_pred(mvy[0], mvy[1], mvy[3]);

  1101.             break;

  1102.         case 0x8:

  1103.             tx = mid_pred(mvx[0], mvx[1], mvx[2]);

  1104.             ty = mid_pred(mvy[0], mvy[1], mvy[2]);

  1105.             break;

  1106.         }

  1107.     } else if(count[idx] == 2) {

  1108.         int t1 = 0, t2 = 0;

  1109.         for(i=0; i<3;i++) if(!intra[i]) {t1 = i; break;}

  1110.         for(i= t1+1; i<4; i++)if(!intra[i]) {t2 = i; break;}

  1111.         tx = (mvx[t1] + mvx[t2]) / 2;

  1112.         ty = (mvy[t1] + mvy[t2]) / 2;

  1113.     } else

  1114.         return; //no need to do MC for inter blocks

  1115. 

  1116.     uvmx = (tx + ((tx&3) == 3)) >> 1;

  1117.     uvmy = (ty + ((ty&3) == 3)) >> 1;

  1118. 

  1119.     uvsrc_x = s->mb_x * 8 + (uvmx >> 2);

  1120.     uvsrc_y = s->mb_y * 8 + (uvmy >> 2);

  1121. 

  1122.     uvsrc_x = clip(uvsrc_x,  -8, s->mb_width  *  8);

  1123.     uvsrc_y = clip(uvsrc_y,  -8, s->mb_height *  8);

  1124.     srcU = s->last_picture.data[1] + uvsrc_y * s->uvlinesize + uvsrc_x;

  1125.     srcV = s->last_picture.data[2] + uvsrc_y * s->uvlinesize + uvsrc_x;

  1126.     if((unsigned)uvsrc_x > (s->h_edge_pos >> 1) - 9

  1127.        || (unsigned)uvsrc_y > (s->v_edge_pos >> 1) - 9){

  1128.         ff_emulated_edge_mc(s->edge_emu_buffer     , srcU, s->uvlinesize, 8+1, 8+1,

  1129.                             uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);

  1130.         ff_emulated_edge_mc(s->edge_emu_buffer + 16, srcV, s->uvlinesize, 8+1, 8+1,

  1131.                             uvsrc_x, uvsrc_y, s->h_edge_pos >> 1, s->v_edge_pos >> 1);

  1132.         srcU = s->edge_emu_buffer;

  1133.         srcV = s->edge_emu_buffer + 16;

  1134.     }

  1135. 

  1136.     if(v->fastuvmc) {

  1137.         uvmx = uvmx + ((uvmx<0)?(uvmx&1):-(uvmx&1));

  1138.         uvmy = uvmy + ((uvmy<0)?(uvmy&1):-(uvmy&1));

  1139.     }

  1140. 

  1141.     /* Chroma MC always uses qpel blilinear */

  1142.     uvdxy = ((uvmy & 3) << 2) | (uvmx & 3);

  1143.     if(!v->rnd){

  1144.         dsp->put_qpel_pixels_tab[1][uvdxy](s->dest[1], srcU, s->uvlinesize);

  1145.         dsp->put_qpel_pixels_tab[1][uvdxy](s->dest[2], srcV, s->uvlinesize);

  1146.     }else{

  1147.         dsp->put_no_rnd_qpel_pixels_tab[1][uvdxy](s->dest[1], srcU, s->uvlinesize);

  1148.         dsp->put_no_rnd_qpel_pixels_tab[1][uvdxy](s->dest[2], srcV, s->uvlinesize);

  1149.     }

  1150. }

  1151. 

  1152. /**

  1153.  * Decode Simple/Main Profiles sequence header

  1154.  * @see Figure 7-8, p16-17

  1155.  * @param avctx Codec context

  1156.  * @param gb GetBit context initialized from Codec context extra_data

  1157.  * @return Status

  1158.  */

  1159. static int decode_sequence_header(AVCodecContext *avctx, GetBitContext *gb)

  1160. {

  1161.     VC1Context *v = avctx->priv_data;

  1162. 

  1163.     av_log(avctx, AV_LOG_INFO, "Header: %0X\n", show_bits(gb, 32));

  1164.     v->profile = get_bits(gb, 2);

  1165.     if (v->profile == 2)

  1166.     {

  1167.         av_log(avctx, AV_LOG_ERROR, "Profile value 2 is forbidden (and WMV3 Complex Profile is unsupported)\n");

  1168.         return -1;

  1169.     }

  1170. 

  1171.     if (v->profile == PROFILE_ADVANCED)

  1172.     {

  1173.         v->level = get_bits(gb, 3);

  1174.         if(v->level >= 5)

  1175.         {

  1176.             av_log(avctx, AV_LOG_ERROR, "Reserved LEVEL %i\n",v->level);

  1177.         }

  1178.         v->chromaformat = get_bits(gb, 2);

  1179.         if (v->chromaformat != 1)

  1180.         {

  1181.             av_log(avctx, AV_LOG_ERROR,

  1182.                    "Only 4:2:0 chroma format supported\n");

  1183.             return -1;

  1184.         }

  1185.     }

  1186.     else

  1187.     {

  1188.         v->res_sm = get_bits(gb, 2); //reserved

  1189.         if (v->res_sm)

  1190.         {

  1191.             av_log(avctx, AV_LOG_ERROR,

  1192.                    "Reserved RES_SM=%i is forbidden\n", v->res_sm);

  1193.             return -1;

  1194.         }

  1195.     }

  1196. 

  1197.     // (fps-2)/4 (->30)

  1198.     v->frmrtq_postproc = get_bits(gb, 3); //common

  1199.     // (bitrate-32kbps)/64kbps

  1200.     v->bitrtq_postproc = get_bits(gb, 5); //common

  1201.     v->s.loop_filter = get_bits(gb, 1); //common

  1202.     if(v->s.loop_filter == 1 && v->profile == PROFILE_SIMPLE)

  1203.     {

  1204.         av_log(avctx, AV_LOG_ERROR,

  1205.                "LOOPFILTER shell not be enabled in simple profile\n");

  1206.     }

  1207. 

  1208.     if (v->profile < PROFILE_ADVANCED)

  1209.     {

  1210.         v->res_x8 = get_bits(gb, 1); //reserved

  1211.         if (v->res_x8)

  1212.         {

  1213.             av_log(avctx, AV_LOG_ERROR,

  1214.                    "1 for reserved RES_X8 is forbidden\n");

  1215.             //return -1;

  1216.         }

  1217.         v->multires = get_bits(gb, 1);

  1218.         v->res_fasttx = get_bits(gb, 1);

  1219.         if (!v->res_fasttx)

  1220.         {

  1221.             av_log(avctx, AV_LOG_ERROR,

  1222.                    "0 for reserved RES_FASTTX is forbidden\n");

  1223.             //return -1;

  1224.         }

  1225.     }

  1226. 

  1227.     v->fastuvmc =  get_bits(gb, 1); //common

  1228.     if (!v->profile && !v->fastuvmc)

  1229.     {

  1230.         av_log(avctx, AV_LOG_ERROR,

  1231.                "FASTUVMC unavailable in Simple Profile\n");

  1232.         return -1;

  1233.     }

  1234.     v->extended_mv =  get_bits(gb, 1); //common

  1235.     if (!v->profile && v->extended_mv)

  1236.     {

  1237.         av_log(avctx, AV_LOG_ERROR,

  1238.                "Extended MVs unavailable in Simple Profile\n");

  1239.         return -1;

  1240.     }

  1241.     v->dquant =  get_bits(gb, 2); //common

  1242.     v->vstransform =  get_bits(gb, 1); //common

  1243. 

  1244.     if (v->profile < PROFILE_ADVANCED)

  1245.     {

  1246.         v->res_transtab = get_bits(gb, 1);

  1247.         if (v->res_transtab)

  1248.         {

  1249.             av_log(avctx, AV_LOG_ERROR,

  1250.                    "1 for reserved RES_TRANSTAB is forbidden\n");

  1251.             return -1;

  1252.         }

  1253.     }

  1254. 

  1255.     v->overlap = get_bits(gb, 1); //common

  1256. 

  1257.     if (v->profile < PROFILE_ADVANCED)

  1258.     {

  1259.         v->s.resync_marker = get_bits(gb, 1);

  1260.         v->rangered = get_bits(gb, 1);

  1261.         if (v->rangered && v->profile == PROFILE_SIMPLE)

  1262.         {

  1263.             av_log(avctx, AV_LOG_INFO,

  1264.                    "RANGERED should be set to 0 in simple profile\n");

  1265.         }

  1266.     }

  1267. 

  1268.     v->s.max_b_frames = avctx->max_b_frames = get_bits(gb, 3); //common

  1269.     v->quantizer_mode = get_bits(gb, 2); //common

  1270. 

  1271.     if (v->profile < PROFILE_ADVANCED)

  1272.     {

  1273.         v->finterpflag = get_bits(gb, 1); //common

  1274.         v->res_rtm_flag = get_bits(gb, 1); //reserved

  1275.         if (!v->res_rtm_flag)

  1276.         {

  1277.             av_log(avctx, AV_LOG_ERROR,

  1278.                    "0 for reserved RES_RTM_FLAG is forbidden\n");

  1279.             //return -1;

  1280.         }

  1281.         av_log(avctx, AV_LOG_DEBUG,

  1282.                "Profile %i:\nfrmrtq_postproc=%i, bitrtq_postproc=%i\n"

  1283.                "LoopFilter=%i, MultiRes=%i, FastUVMC=%i, Extended MV=%i\n"

  1284.                "Rangered=%i, VSTransform=%i, Overlap=%i, SyncMarker=%i\n"

  1285.                "DQuant=%i, Quantizer mode=%i, Max B frames=%i\n",

  1286.                v->profile, v->frmrtq_postproc, v->bitrtq_postproc,

  1287.                v->s.loop_filter, v->multires, v->fastuvmc, v->extended_mv,

  1288.                v->rangered, v->vstransform, v->overlap, v->s.resync_marker,

  1289.                v->dquant, v->quantizer_mode, avctx->max_b_frames

  1290.                );

  1291.         return 0;

  1292.     }

  1293.     return -1;

  1294. }

  1295. 

  1296. 

  1297. static int vc1_parse_frame_header(VC1Context *v, GetBitContext* gb)

  1298. {

  1299.     int pqindex, lowquant, status;

  1300. 

  1301.     if(v->finterpflag) v->interpfrm = get_bits(gb, 1);

  1302.     skip_bits(gb, 2); //framecnt unused

  1303.     v->rangeredfrm = 0;

  1304.     if (v->rangered) v->rangeredfrm = get_bits(gb, 1);

  1305.     v->s.pict_type = get_bits(gb, 1);

  1306.     if (v->s.avctx->max_b_frames) {

  1307.         if (!v->s.pict_type) {

  1308.             if (get_bits(gb, 1)) v->s.pict_type = I_TYPE;

  1309.             else v->s.pict_type = B_TYPE;

  1310.         } else v->s.pict_type = P_TYPE;

  1311.     } else v->s.pict_type = v->s.pict_type ? P_TYPE : I_TYPE;

  1312. 

  1313.     if(v->s.pict_type == I_TYPE)

  1314.         get_bits(gb, 7); // skip buffer fullness

  1315. 

  1316.     /* calculate RND */

  1317.     if(v->s.pict_type == I_TYPE)

  1318.         v->rnd = 1;

  1319.     if(v->s.pict_type == P_TYPE)

  1320.         v->rnd ^= 1;

  1321. 

  1322.     /* Quantizer stuff */

  1323.     pqindex = get_bits(gb, 5);

  1324.     if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)

  1325.         v->pq = pquant_table[0][pqindex];

  1326.     else

  1327.         v->pq = pquant_table[1][pqindex];

  1328. 

  1329.     v->pquantizer = 1;

  1330.     if (v->quantizer_mode == QUANT_FRAME_IMPLICIT)

  1331.         v->pquantizer = pqindex < 9;

  1332.     if (v->quantizer_mode == QUANT_NON_UNIFORM)

  1333.         v->pquantizer = 0;

  1334.     v->pqindex = pqindex;

  1335.     if (pqindex < 9) v->halfpq = get_bits(gb, 1);

  1336.     else v->halfpq = 0;

  1337.     if (v->quantizer_mode == QUANT_FRAME_EXPLICIT)

  1338.         v->pquantizer = get_bits(gb, 1);

  1339.     v->dquantfrm = 0;

  1340.     if (v->extended_mv == 1) v->mvrange = get_prefix(gb, 0, 3);

  1341.     v->k_x = v->mvrange + 9 + (v->mvrange >> 1); //k_x can be 9 10 12 13

  1342.     v->k_y = v->mvrange + 8; //k_y can be 8 9 10 11

  1343.     v->range_x = 1 << (v->k_x - 1);

  1344.     v->range_y = 1 << (v->k_y - 1);

  1345.     if (v->profile == PROFILE_ADVANCED)

  1346.     {

  1347.         if (v->postprocflag) v->postproc = get_bits(gb, 1);

  1348.     }

  1349.     else

  1350.         if (v->multires && v->s.pict_type != B_TYPE) v->respic = get_bits(gb, 2);

  1351. 

  1352. //av_log(v->s.avctx, AV_LOG_INFO, "%c Frame: QP=[%i]%i (+%i/2) %i\n",

  1353. //        (v->s.pict_type == P_TYPE) ? 'P' : ((v->s.pict_type == I_TYPE) ? 'I' : 'B'), pqindex, v->pq, v->halfpq, v->rangeredfrm);

  1354. 

  1355.     //TODO: complete parsing for P/B/BI frames

  1356.     switch(v->s.pict_type) {

  1357.     case P_TYPE:

  1358.         if (v->pq < 5) v->tt_index = 0;

  1359.         else if(v->pq < 13) v->tt_index = 1;

  1360.         else v->tt_index = 2;

  1361. 

  1362.         lowquant = (v->pq > 12) ? 0 : 1;

  1363.         v->mv_mode = mv_pmode_table[lowquant][get_prefix(gb, 1, 4)];

  1364.         if (v->mv_mode == MV_PMODE_INTENSITY_COMP)

  1365.         {

  1366.             int scale, shift, i;

  1367.             v->mv_mode2 = mv_pmode_table2[lowquant][get_prefix(gb, 1, 3)];

  1368.             v->lumscale = get_bits(gb, 6);

  1369.             v->lumshift = get_bits(gb, 6);

  1370.             /* fill lookup tables for intensity compensation */

  1371.             if(!v->lumscale) {

  1372.                 scale = -64;

  1373.                 shift = (255 - v->lumshift * 2) << 6;

  1374.                 if(v->lumshift > 31)

  1375.                     shift += 128 << 6;

  1376.             } else {

  1377.                 scale = v->lumscale + 32;

  1378.                 if(v->lumshift > 31)

  1379.                     shift = (v->lumshift - 64) << 6;

  1380.                 else

  1381.                     shift = v->lumshift << 6;

  1382.             }

  1383.             for(i = 0; i < 256; i++) {

  1384.                 v->luty[i] = clip_uint8((scale * i + shift + 32) >> 6);

  1385.                 v->lutuv[i] = clip_uint8((scale * (i - 128) + 128*64 + 32) >> 6);

  1386.             }

  1387.         }

  1388.         if(v->mv_mode == MV_PMODE_1MV_HPEL || v->mv_mode == MV_PMODE_1MV_HPEL_BILIN)

  1389.             v->s.quarter_sample = 0;

  1390.         else if(v->mv_mode == MV_PMODE_INTENSITY_COMP) {

  1391.             if(v->mv_mode2 == MV_PMODE_1MV_HPEL || v->mv_mode2 == MV_PMODE_1MV_HPEL_BILIN)

  1392.                 v->s.quarter_sample = 0;

  1393.             else

  1394.                 v->s.quarter_sample = 1;

  1395.         } else

  1396.             v->s.quarter_sample = 1;

  1397. 

  1398.         if ((v->mv_mode == MV_PMODE_INTENSITY_COMP &&

  1399.                  v->mv_mode2 == MV_PMODE_MIXED_MV)

  1400.                 || v->mv_mode == MV_PMODE_MIXED_MV)

  1401.         {

  1402.             status = bitplane_decoding(v->mv_type_mb_plane, &v->mv_type_is_raw, v);

  1403.             if (status < 0) return -1;

  1404.             av_log(v->s.avctx, AV_LOG_DEBUG, "MB MV Type plane encoding: "

  1405.                    "Imode: %i, Invert: %i\n", status>>1, status&1);

  1406.         } else {

  1407.             v->mv_type_is_raw = 0;

  1408.             memset(v->mv_type_mb_plane, 0, v->s.mb_stride * v->s.mb_height);

  1409.         }

  1410.         status = bitplane_decoding(v->s.mbskip_table, &v->skip_is_raw, v);

  1411.         if (status < 0) return -1;

  1412.         av_log(v->s.avctx, AV_LOG_DEBUG, "MB Skip plane encoding: "

  1413.                "Imode: %i, Invert: %i\n", status>>1, status&1);

  1414. 

  1415.         /* Hopefully this is correct for P frames */

  1416.         v->s.mv_table_index = get_bits(gb, 2); //but using vc1_ tables

  1417.         v->cbpcy_vlc = &vc1_cbpcy_p_vlc[get_bits(gb, 2)];

  1418. 

  1419.         if (v->dquant)

  1420.         {

  1421.             av_log(v->s.avctx, AV_LOG_DEBUG, "VOP DQuant info\n");

  1422.             vop_dquant_decoding(v);

  1423.         }

  1424. 

  1425.         v->ttfrm = 0; //FIXME Is that so ?

  1426.         if (v->vstransform)

  1427.         {

  1428.             v->ttmbf = get_bits(gb, 1);

  1429.             if (v->ttmbf)

  1430.             {

  1431.                 v->ttfrm = ttfrm_to_tt[get_bits(gb, 2)];

  1432.             }

  1433.         } else {

  1434.             v->ttmbf = 1;

  1435.             v->ttfrm = TT_8X8;

  1436.         }

  1437.         break;

  1438.     case B_TYPE:

  1439.         break;

  1440.     }

  1441. 

  1442.     /* AC Syntax */

  1443.     v->c_ac_table_index = decode012(gb);

  1444.     if (v->s.pict_type == I_TYPE || v->s.pict_type == BI_TYPE)

  1445.     {

  1446.         v->y_ac_table_index = decode012(gb);

  1447.     }

  1448.     /* DC Syntax */

  1449.     v->s.dc_table_index = get_bits(gb, 1);

  1450. 

  1451.     return 0;

  1452. }

  1453. 

  1454. /***********************************************************************/

  1455. /**

  1456.  * @defgroup block VC-1 Block-level functions

  1457.  * @see 7.1.4, p91 and 8.1.1.7, p(1)04

  1458.  * @todo TODO: Integrate to MpegEncContext facilities

  1459.  * @{

  1460.  */

  1461. 

  1462. /**

  1463.  * @def GET_MQUANT

  1464.  * @brief Get macroblock-level quantizer scale

  1465.  * @warning XXX: qdiff to the frame quant, not previous quant ?

  1466.  * @fixme XXX: Don't know how to initialize mquant otherwise in last case

  1467.  */

  1468. #define GET_MQUANT()                                           \

  1469.   if (v->dquantfrm)                                            \

  1470.   {                                                            \

  1471.     int edges = 0;                                             \

  1472.     if (v->dqprofile == DQPROFILE_ALL_MBS)                     \

  1473.     {                                                          \

  1474.       if (v->dqbilevel)                                        \

  1475.       {                                                        \

  1476.         mquant = (get_bits(gb, 1)) ? v->altpq : v->pq;         \

  1477.       }                                                        \

  1478.       else                                                     \

  1479.       {                                                        \

  1480.         mqdiff = get_bits(gb, 3);                              \

  1481.         if (mqdiff != 7) mquant = v->pq + mqdiff;              \

  1482.         else mquant = get_bits(gb, 5);                         \

  1483.       }                                                        \

  1484.     }                                                          \

  1485.     if(v->dqprofile == DQPROFILE_SINGLE_EDGE)                  \

  1486.         edges = 1 << v->dqsbedge;                              \

  1487.     else if(v->dqprofile == DQPROFILE_DOUBLE_EDGES)            \

  1488.         edges = (3 << v->dqsbedge) % 15;                       \

  1489.     else if(v->dqprofile == DQPROFILE_FOUR_EDGES)              \

  1490.         edges = 15;                                            \

  1491.     if((edges&1) && !s->mb_x)                                  \

  1492.         mquant = v->altpq;                                     \

  1493.     if((edges&2) && s->first_slice_line)                       \

  1494.         mquant = v->altpq;                                     \

  1495.     if((edges&4) && s->mb_x == (s->mb_width - 1))              \

  1496.         mquant = v->altpq;                                     \

  1497.     if((edges&8) && s->mb_y == (s->mb_height - 1))             \

  1498.         mquant = v->altpq;                                     \

  1499.   }

  1500. 

  1501. /**

  1502.  * @def GET_MVDATA(_dmv_x, _dmv_y)

  1503.  * @brief Get MV differentials

  1504.  * @see MVDATA decoding from 8.3.5.2, p(1)20

  1505.  * @param _dmv_x Horizontal differential for decoded MV

  1506.  * @param _dmv_y Vertical differential for decoded MV

  1507.  * @todo TODO: Use MpegEncContext arrays to store them

  1508.  */

  1509. #define GET_MVDATA(_dmv_x, _dmv_y)                                  \

  1510.   index = 1 + get_vlc2(gb, vc1_mv_diff_vlc[s->mv_table_index].table,\

  1511.                        VC1_MV_DIFF_VLC_BITS, 2);                    \

  1512.   if (index > 36)                                                   \

  1513.   {                                                                 \

  1514.     mb_has_coeffs = 1;                                              \

  1515.     index -= 37;                                                    \

  1516.   }                                                                 \

  1517.   else mb_has_coeffs = 0;                                           \

  1518.   s->mb_intra = 0;                                                  \

  1519.   if (!index) { _dmv_x = _dmv_y = 0; }                              \

  1520.   else if (index == 35)                                             \

  1521.   {                                                                 \

  1522.     _dmv_x = get_bits(gb, v->k_x - 1 + s->quarter_sample);          \

  1523.     _dmv_y = get_bits(gb, v->k_y - 1 + s->quarter_sample);          \

  1524.   }                                                                 \

  1525.   else if (index == 36)                                             \

  1526.   {                                                                 \

  1527.     _dmv_x = 0;                                                     \

  1528.     _dmv_y = 0;                                                     \

  1529.     s->mb_intra = 1;                                                \

  1530.   }                                                                 \

  1531.   else                                                              \

  1532.   {                                                                 \

  1533.     index1 = index%6;                                               \

  1534.     if (!s->quarter_sample && index1 == 5) val = 1;                 \

  1535.     else                                   val = 0;                 \

  1536.     if(size_table[index1] - val > 0)                                \

  1537.         val = get_bits(gb, size_table[index1] - val);               \

  1538.     else                                   val = 0;                 \

  1539.     sign = 0 - (val&1);                                             \

  1540.     _dmv_x = (sign ^ ((val>>1) + offset_table[index1])) - sign;     \

  1541.                                                                     \

  1542.     index1 = index/6;                                               \

  1543.     if (!s->quarter_sample && index1 == 5) val = 1;                 \

  1544.     else                                   val = 0;                 \

  1545.     if(size_table[index1] - val > 0)                                \

  1546.         val = get_bits(gb, size_table[index1] - val);               \

  1547.     else                                   val = 0;                 \

  1548.     sign = 0 - (val&1);                                             \

  1549.     _dmv_y = (sign ^ ((val>>1) + offset_table[index1])) - sign;     \

  1550.   }

  1551. 

  1552. /** Predict and set motion vector

  1553.  */

  1554. static inline void vc1_pred_mv(MpegEncContext *s, int n, int dmv_x, int dmv_y, int mv1, int r_x, int r_y, uint8_t* is_intra)

  1555. {

  1556.     int xy, wrap, off = 0;

  1557.     int16_t *A, *B, *C;

  1558.     int px, py;

  1559.     int sum;

  1560. 

  1561.     /* scale MV difference to be quad-pel */

  1562.     dmv_x <<= 1 - s->quarter_sample;

  1563.     dmv_y <<= 1 - s->quarter_sample;

  1564. 

  1565.     wrap = s->b8_stride;

  1566.     xy = s->block_index[n];

  1567. 

  1568.     if(s->mb_intra){

  1569.         s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = 0;

  1570.         s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = 0;

  1571.         if(mv1) { /* duplicate motion data for 1-MV block */

  1572.             s->current_picture.motion_val[0][xy + 1][0] = 0;

  1573.             s->current_picture.motion_val[0][xy + 1][1] = 0;

  1574.             s->current_picture.motion_val[0][xy + wrap][0] = 0;

  1575.             s->current_picture.motion_val[0][xy + wrap][1] = 0;

  1576.             s->current_picture.motion_val[0][xy + wrap + 1][0] = 0;

  1577.             s->current_picture.motion_val[0][xy + wrap + 1][1] = 0;

  1578.         }

  1579.         return;

  1580.     }

  1581. 

  1582.     C = s->current_picture.motion_val[0][xy - 1];

  1583.     A = s->current_picture.motion_val[0][xy - wrap];

  1584.     if(mv1)

  1585.         off = (s->mb_x == (s->mb_width - 1)) ? -1 : 2;

  1586.     else {

  1587.         //in 4-MV mode different blocks have different B predictor position

  1588.         switch(n){

  1589.         case 0:

  1590.             off = (s->mb_x > 0) ? -1 : 1;

  1591.             break;

  1592.         case 1:

  1593.             off = (s->mb_x == (s->mb_width - 1)) ? -1 : 1;

  1594.             break;

  1595.         case 2:

  1596.             off = 1;

  1597.             break;

  1598.         case 3:

  1599.             off = -1;

  1600.         }

  1601.     }

  1602.     B = s->current_picture.motion_val[0][xy - wrap + off];

  1603. 

  1604.     if(!s->first_slice_line || (n==2 || n==3)) { // predictor A is not out of bounds

  1605.         if(s->mb_width == 1) {

  1606.             px = A[0];

  1607.             py = A[1];

  1608.         } else {

  1609.             px = mid_pred(A[0], B[0], C[0]);

  1610.             py = mid_pred(A[1], B[1], C[1]);

  1611.         }

  1612.     } else if(s->mb_x || (n==1 || n==3)) { // predictor C is not out of bounds

  1613.         px = C[0];

  1614.         py = C[1];

  1615.     } else {

  1616.         px = py = 0;

  1617.     }

  1618.     /* Pullback MV as specified in 8.3.5.3.4 */

  1619.     {

  1620.         int qx, qy, X, Y;

  1621.         qx = (s->mb_x << 6) + ((n==1 || n==3) ? 32 : 0);

  1622.         qy = (s->mb_y << 6) + ((n==2 || n==3) ? 32 : 0);

  1623.         X = (s->mb_width << 6) - 4;

  1624.         Y = (s->mb_height << 6) - 4;

  1625.         if(mv1) {

  1626.             if(qx + px < -60) px = -60 - qx;

  1627.             if(qy + py < -60) py = -60 - qy;

  1628.         } else {

  1629.             if(qx + px < -28) px = -28 - qx;

  1630.             if(qy + py < -28) py = -28 - qy;

  1631.         }

  1632.         if(qx + px > X) px = X - qx;

  1633.         if(qy + py > Y) py = Y - qy;

  1634.     }

  1635.     /* Calculate hybrid prediction as specified in 8.3.5.3.5 */

  1636.     if((!s->first_slice_line || (n==2 || n==3)) && (s->mb_x || (n==1 || n==3))) {

  1637.         if(is_intra[xy - wrap])

  1638.             sum = ABS(px) + ABS(py);

  1639.         else

  1640.             sum = ABS(px - A[0]) + ABS(py - A[1]);

  1641.         if(sum > 32) {

  1642.             if(get_bits1(&s->gb)) {

  1643.                 px = A[0];

  1644.                 py = A[1];

  1645.             } else {

  1646.                 px = C[0];

  1647.                 py = C[1];

  1648.             }

  1649.         } else {

  1650.             if(is_intra[xy - 1])

  1651.                 sum = ABS(px) + ABS(py);

  1652.             else

  1653.                 sum = ABS(px - C[0]) + ABS(py - C[1]);

  1654.             if(sum > 32) {

  1655.                 if(get_bits1(&s->gb)) {

  1656.                     px = A[0];

  1657.                     py = A[1];

  1658.                 } else {

  1659.                     px = C[0];

  1660.                     py = C[1];

  1661.                 }

  1662.             }

  1663.         }

  1664.     }

  1665.     /* store MV using signed modulus of MV range defined in 4.11 */

  1666.     s->mv[0][n][0] = s->current_picture.motion_val[0][xy][0] = ((px + dmv_x + r_x) & ((r_x << 1) - 1)) - r_x;

  1667.     s->mv[0][n][1] = s->current_picture.motion_val[0][xy][1] = ((py + dmv_y + r_y) & ((r_y << 1) - 1)) - r_y;

  1668.     if(mv1) { /* duplicate motion data for 1-MV block */

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

  1670.         s->current_picture.motion_val[0][xy + 1][1] = s->current_picture.motion_val[0][xy][1];

  1671.         s->current_picture.motion_val[0][xy + wrap][0] = s->current_picture.motion_val[0][xy][0];

  1672.         s->current_picture.motion_val[0][xy + wrap][1] = s->current_picture.motion_val[0][xy][1];

  1673.         s->current_picture.motion_val[0][xy + wrap + 1][0] = s->current_picture.motion_val[0][xy][0];

  1674.         s->current_picture.motion_val[0][xy + wrap + 1][1] = s->current_picture.motion_val[0][xy][1];

  1675.     }

  1676. }

  1677. 

  1678. /** Get predicted DC value for I-frames only

  1679.  * prediction dir: left=0, top=1

  1680.  * @param s MpegEncContext

  1681.  * @param[in] n block index in the current MB

  1682.  * @param dc_val_ptr Pointer to DC predictor

  1683.  * @param dir_ptr Prediction direction for use in AC prediction

  1684.  */

  1685. static inline int vc1_i_pred_dc(MpegEncContext *s, int overlap, int pq, int n,

  1686.                               int16_t **dc_val_ptr, int *dir_ptr)

  1687. {

  1688.     int a, b, c, wrap, pred, scale;

  1689.     int16_t *dc_val;

  1690.     static const uint16_t dcpred[32] = {

  1691.     -1, 1024,  512,  341,  256,  205,  171,  146,  128,

  1692.          114,  102,   93,   85,   79,   73,   68,   64,

  1693.           60,   57,   54,   51,   49,   47,   45,   43,

  1694.           41,   39,   38,   37,   35,   34,   33

  1695.     };

  1696. 

  1697.     /* find prediction - wmv3_dc_scale always used here in fact */

  1698.     if (n < 4)     scale = s->y_dc_scale;

  1699.     else           scale = s->c_dc_scale;

  1700. 

  1701.     wrap = s->block_wrap[n];

  1702.     dc_val= s->dc_val[0] + s->block_index[n];

  1703. 

  1704.     /* B A

  1705.      * C X

  1706.      */

  1707.     c = dc_val[ - 1];

  1708.     b = dc_val[ - 1 - wrap];

  1709.     a = dc_val[ - wrap];

  1710. 

  1711.     if (pq < 9 || !overlap)

  1712.     {

  1713.         /* Set outer values */

  1714.         if (s->first_slice_line && (n!=2 && n!=3)) b=a=dcpred[scale];

  1715.         if (s->mb_x == 0 && (n!=1 && n!=3)) b=c=dcpred[scale];

  1716.     }

  1717.     else

  1718.     {

  1719.         /* Set outer values */

  1720.         if (s->first_slice_line && (n!=2 && n!=3)) b=a=0;

  1721.         if (s->mb_x == 0 && (n!=1 && n!=3)) b=c=0;

  1722.     }

  1723. 

  1724.     if (abs(a - b) <= abs(b - c)) {

  1725.         pred = c;

  1726.         *dir_ptr = 1;//left

  1727.     } else {

  1728.         pred = a;

  1729.         *dir_ptr = 0;//top

  1730.     }

  1731. 

  1732.     /* update predictor */

  1733.     *dc_val_ptr = &dc_val[0];

  1734.     return pred;

  1735. }

  1736. 

  1737. 

  1738. /** Get predicted DC value

  1739.  * prediction dir: left=0, top=1

  1740.  * @param s MpegEncContext

  1741.  * @param[in] n block index in the current MB

  1742.  * @param dc_val_ptr Pointer to DC predictor

  1743.  * @param dir_ptr Prediction direction for use in AC prediction

  1744.  */

  1745. static inline int vc1_pred_dc(MpegEncContext *s, int overlap, int pq, int n,

  1746.                               int a_avail, int c_avail,

  1747.                               int16_t **dc_val_ptr, int *dir_ptr)

  1748. {

  1749.     int a, b, c, wrap, pred, scale;

  1750.     int16_t *dc_val;

  1751.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  1752.     int q1, q2 = 0;

  1753. 

  1754.     /* find prediction - wmv3_dc_scale always used here in fact */

  1755.     if (n < 4)     scale = s->y_dc_scale;

  1756.     else           scale = s->c_dc_scale;

  1757. 

  1758.     wrap = s->block_wrap[n];

  1759.     dc_val= s->dc_val[0] + s->block_index[n];

  1760. 

  1761.     /* B A

  1762.      * C X

  1763.      */

  1764.     c = dc_val[ - 1];

  1765.     b = dc_val[ - 1 - wrap];

  1766.     a = dc_val[ - wrap];

  1767.     /* scale predictors if needed */

  1768.     q1 = s->current_picture.qscale_table[mb_pos];

  1769.     if(c_avail && (n!= 1 && n!=3)) {

  1770.         q2 = s->current_picture.qscale_table[mb_pos - 1];

  1771.         if(q2 && q2 != q1)

  1772.             c = (c * s->y_dc_scale_table[q2] * vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;

  1773.     }

  1774.     if(a_avail && (n!= 2 && n!=3)) {

  1775.         q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];

  1776.         if(q2 && q2 != q1)

  1777.             a = (a * s->y_dc_scale_table[q2] * vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;

  1778.     }

  1779.     if(a_avail && c_avail && (n!=3)) {

  1780.         int off = mb_pos;

  1781.         if(n != 1) off--;

  1782.         if(n != 2) off -= s->mb_stride;

  1783.         q2 = s->current_picture.qscale_table[off];

  1784.         if(q2 && q2 != q1)

  1785.             b = (b * s->y_dc_scale_table[q2] * vc1_dqscale[s->y_dc_scale_table[q1] - 1] + 0x20000) >> 18;

  1786.     }

  1787. 

  1788.     if(a_avail && c_avail) {

  1789.         if(abs(a - b) <= abs(b - c)) {

  1790.             pred = c;

  1791.             *dir_ptr = 1;//left

  1792.         } else {

  1793.             pred = a;

  1794.             *dir_ptr = 0;//top

  1795.         }

  1796.     } else if(a_avail) {

  1797.         pred = a;

  1798.         *dir_ptr = 0;//top

  1799.     } else if(c_avail) {

  1800.         pred = c;

  1801.         *dir_ptr = 1;//left

  1802.     } else {

  1803.         pred = 0;

  1804.         *dir_ptr = 1;//left

  1805.     }

  1806. 

  1807.     /* update predictor */

  1808.     *dc_val_ptr = &dc_val[0];

  1809.     return pred;

  1810. }

  1811. 

  1812. 

  1813. /**

  1814.  * @defgroup std_mb VC1 Macroblock-level functions in Simple/Main Profiles

  1815.  * @see 7.1.4, p91 and 8.1.1.7, p(1)04

  1816.  * @todo TODO: Integrate to MpegEncContext facilities

  1817.  * @{

  1818.  */

  1819. 

  1820. static inline int vc1_coded_block_pred(MpegEncContext * s, int n, uint8_t **coded_block_ptr)

  1821. {

  1822.     int xy, wrap, pred, a, b, c;

  1823. 

  1824.     xy = s->block_index[n];

  1825.     wrap = s->b8_stride;

  1826. 

  1827.     /* B C

  1828.      * A X

  1829.      */

  1830.     a = s->coded_block[xy - 1       ];

  1831.     b = s->coded_block[xy - 1 - wrap];

  1832.     c = s->coded_block[xy     - wrap];

  1833. 

  1834.     if (b == c) {

  1835.         pred = a;

  1836.     } else {

  1837.         pred = c;

  1838.     }

  1839. 

  1840.     /* store value */

  1841.     *coded_block_ptr = &s->coded_block[xy];

  1842. 

  1843.     return pred;

  1844. }

  1845. 

  1846. /**

  1847.  * Decode one AC coefficient

  1848.  * @param v The VC1 context

  1849.  * @param last Last coefficient

  1850.  * @param skip How much zero coefficients to skip

  1851.  * @param value Decoded AC coefficient value

  1852.  * @see 8.1.3.4

  1853.  */

  1854. static void vc1_decode_ac_coeff(VC1Context *v, int *last, int *skip, int *value, int codingset)

  1855. {

  1856.     GetBitContext *gb = &v->s.gb;

  1857.     int index, escape, run = 0, level = 0, lst = 0;

  1858. 

  1859.     index = get_vlc2(gb, vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);

  1860.     if (index != vc1_ac_sizes[codingset] - 1) {

  1861.         run = vc1_index_decode_table[codingset][index][0];

  1862.         level = vc1_index_decode_table[codingset][index][1];

  1863.         lst = index >= vc1_last_decode_table[codingset];

  1864.         if(get_bits(gb, 1))

  1865.             level = -level;

  1866.     } else {

  1867.         escape = decode210(gb);

  1868.         if (escape != 2) {

  1869.             index = get_vlc2(gb, vc1_ac_coeff_table[codingset].table, AC_VLC_BITS, 3);

  1870.             run = vc1_index_decode_table[codingset][index][0];

  1871.             level = vc1_index_decode_table[codingset][index][1];

  1872.             lst = index >= vc1_last_decode_table[codingset];

  1873.             if(escape == 0) {

  1874.                 if(lst)

  1875.                     level += vc1_last_delta_level_table[codingset][run];

  1876.                 else

  1877.                     level += vc1_delta_level_table[codingset][run];

  1878.             } else {

  1879.                 if(lst)

  1880.                     run += vc1_last_delta_run_table[codingset][level] + 1;

  1881.                 else

  1882.                     run += vc1_delta_run_table[codingset][level] + 1;

  1883.             }

  1884.             if(get_bits(gb, 1))

  1885.                 level = -level;

  1886.         } else {

  1887.             int sign;

  1888.             lst = get_bits(gb, 1);

  1889.             if(v->s.esc3_level_length == 0) {

  1890.                 if(v->pq < 8 || v->dquantfrm) { // table 59

  1891.                     v->s.esc3_level_length = get_bits(gb, 3);

  1892.                     if(!v->s.esc3_level_length)

  1893.                         v->s.esc3_level_length = get_bits(gb, 2) + 8;

  1894.                 } else { //table 60

  1895.                     v->s.esc3_level_length = get_prefix(gb, 1, 6) + 2;

  1896.                 }

  1897.                 v->s.esc3_run_length = 3 + get_bits(gb, 2);

  1898.             }

  1899.             run = get_bits(gb, v->s.esc3_run_length);

  1900.             sign = get_bits(gb, 1);

  1901.             level = get_bits(gb, v->s.esc3_level_length);

  1902.             if(sign)

  1903.                 level = -level;

  1904.         }

  1905.     }

  1906. 

  1907.     *last = lst;

  1908.     *skip = run;

  1909.     *value = level;

  1910. }

  1911. 

  1912. /** Decode intra block in intra frames - should be faster than decode_intra_block

  1913.  * @param v VC1Context

  1914.  * @param block block to decode

  1915.  * @param coded are AC coeffs present or not

  1916.  * @param codingset set of VLC to decode data

  1917.  */

  1918. static int vc1_decode_i_block(VC1Context *v, DCTELEM block[64], int n, int coded, int codingset)

  1919. {

  1920.     GetBitContext *gb = &v->s.gb;

  1921.     MpegEncContext *s = &v->s;

  1922.     int dc_pred_dir = 0; /* Direction of the DC prediction used */

  1923.     int run_diff, i;

  1924.     int16_t *dc_val;

  1925.     int16_t *ac_val, *ac_val2;

  1926.     int dcdiff;

  1927. 

  1928.     /* Get DC differential */

  1929.     if (n < 4) {

  1930.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  1931.     } else {

  1932.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  1933.     }

  1934.     if (dcdiff < 0){

  1935.         av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");

  1936.         return -1;

  1937.     }

  1938.     if (dcdiff)

  1939.     {

  1940.         if (dcdiff == 119 /* ESC index value */)

  1941.         {

  1942.             /* TODO: Optimize */

  1943.             if (v->pq == 1) dcdiff = get_bits(gb, 10);

  1944.             else if (v->pq == 2) dcdiff = get_bits(gb, 9);

  1945.             else dcdiff = get_bits(gb, 8);

  1946.         }

  1947.         else

  1948.         {

  1949.             if (v->pq == 1)

  1950.                 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;

  1951.             else if (v->pq == 2)

  1952.                 dcdiff = (dcdiff<<1) + get_bits(gb, 1) - 1;

  1953.         }

  1954.         if (get_bits(gb, 1))

  1955.             dcdiff = -dcdiff;

  1956.     }

  1957. 

  1958.     /* Prediction */

  1959.     dcdiff += vc1_i_pred_dc(&v->s, v->overlap, v->pq, n, &dc_val, &dc_pred_dir);

  1960.     *dc_val = dcdiff;

  1961. 

  1962.     /* Store the quantized DC coeff, used for prediction */

  1963.     if (n < 4) {

  1964.         block[0] = dcdiff * s->y_dc_scale;

  1965.     } else {

  1966.         block[0] = dcdiff * s->c_dc_scale;

  1967.     }

  1968.     /* Skip ? */

  1969.     run_diff = 0;

  1970.     i = 0;

  1971.     if (!coded) {

  1972.         goto not_coded;

  1973.     }

  1974. 

  1975.     //AC Decoding

  1976.     i = 1;

  1977. 

  1978.     {

  1979.         int last = 0, skip, value;

  1980.         const int8_t *zz_table;

  1981.         int scale;

  1982.         int k;

  1983. 

  1984.         scale = v->pq * 2 + v->halfpq;

  1985. 

  1986.         if(v->s.ac_pred) {

  1987.             if(!dc_pred_dir)

  1988.                 zz_table = vc1_horizontal_zz;

  1989.             else

  1990.                 zz_table = vc1_vertical_zz;

  1991.         } else

  1992.             zz_table = vc1_normal_zz;

  1993. 

  1994.         ac_val = s->ac_val[0][0] + s->block_index[n] * 16;

  1995.         ac_val2 = ac_val;

  1996.         if(dc_pred_dir) //left

  1997.             ac_val -= 16;

  1998.         else //top

  1999.             ac_val -= 16 * s->block_wrap[n];

  2000. 

  2001.         while (!last) {

  2002.             vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);

  2003.             i += skip;

  2004.             if(i > 63)

  2005.                 break;

  2006.             block[zz_table[i++]] = value;

  2007.         }

  2008. 

  2009.         /* apply AC prediction if needed */

  2010.         if(s->ac_pred) {

  2011.             if(dc_pred_dir) { //left

  2012.                 for(k = 1; k < 8; k++)

  2013.                     block[k << 3] += ac_val[k];

  2014.             } else { //top

  2015.                 for(k = 1; k < 8; k++)

  2016.                     block[k] += ac_val[k + 8];

  2017.             }

  2018.         }

  2019.         /* save AC coeffs for further prediction */

  2020.         for(k = 1; k < 8; k++) {

  2021.             ac_val2[k] = block[k << 3];

  2022.             ac_val2[k + 8] = block[k];

  2023.         }

  2024. 

  2025.         /* scale AC coeffs */

  2026.         for(k = 1; k < 64; k++)

  2027.             if(block[k]) {

  2028.                 block[k] *= scale;

  2029.                 if(!v->pquantizer)

  2030.                     block[k] += (block[k] < 0) ? -v->pq : v->pq;

  2031.             }

  2032. 

  2033.         if(s->ac_pred) i = 63;

  2034.     }

  2035. 

  2036. not_coded:

  2037.     if(!coded) {

  2038.         int k, scale;

  2039.         ac_val = s->ac_val[0][0] + s->block_index[n] * 16;

  2040.         ac_val2 = ac_val;

  2041. 

  2042.         scale = v->pq * 2 + v->halfpq;

  2043.         memset(ac_val2, 0, 16 * 2);

  2044.         if(dc_pred_dir) {//left

  2045.             ac_val -= 16;

  2046.             if(s->ac_pred)

  2047.                 memcpy(ac_val2, ac_val, 8 * 2);

  2048.         } else {//top

  2049.             ac_val -= 16 * s->block_wrap[n];

  2050.             if(s->ac_pred)

  2051.                 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);

  2052.         }

  2053. 

  2054.         /* apply AC prediction if needed */

  2055.         if(s->ac_pred) {

  2056.             if(dc_pred_dir) { //left

  2057.                 for(k = 1; k < 8; k++) {

  2058.                     block[k << 3] = ac_val[k] * scale;

  2059.                     if(!v->pquantizer && block[k << 3])

  2060.                         block[k << 3] += (block[k << 3] < 0) ? -v->pq : v->pq;

  2061.                 }

  2062.             } else { //top

  2063.                 for(k = 1; k < 8; k++) {

  2064.                     block[k] = ac_val[k + 8] * scale;

  2065.                     if(!v->pquantizer && block[k])

  2066.                         block[k] += (block[k] < 0) ? -v->pq : v->pq;

  2067.                 }

  2068.             }

  2069.             i = 63;

  2070.         }

  2071.     }

  2072.     s->block_last_index[n] = i;

  2073. 

  2074.     return 0;

  2075. }

  2076. 

  2077. /** Decode intra block in inter frames - more generic version than vc1_decode_i_block

  2078.  * @param v VC1Context

  2079.  * @param block block to decode

  2080.  * @param coded are AC coeffs present or not

  2081.  * @param mquant block quantizer

  2082.  * @param codingset set of VLC to decode data

  2083.  */

  2084. static int vc1_decode_intra_block(VC1Context *v, DCTELEM block[64], int n, int coded, int mquant, int codingset)

  2085. {

  2086.     GetBitContext *gb = &v->s.gb;

  2087.     MpegEncContext *s = &v->s;

  2088.     int dc_pred_dir = 0; /* Direction of the DC prediction used */

  2089.     int run_diff, i;

  2090.     int16_t *dc_val;

  2091.     int16_t *ac_val, *ac_val2;

  2092.     int dcdiff;

  2093.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  2094.     int a_avail = v->a_avail, c_avail = v->c_avail;

  2095.     int use_pred = s->ac_pred;

  2096.     int scale;

  2097.     int q1, q2 = 0;

  2098. 

  2099.     /* XXX: Guard against dumb values of mquant */

  2100.     mquant = (mquant < 1) ? 0 : ( (mquant>31) ? 31 : mquant );

  2101. 

  2102.     /* Set DC scale - y and c use the same */

  2103.     s->y_dc_scale = s->y_dc_scale_table[mquant];

  2104.     s->c_dc_scale = s->c_dc_scale_table[mquant];

  2105. 

  2106.     /* Get DC differential */

  2107.     if (n < 4) {

  2108.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_luma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2109.     } else {

  2110.         dcdiff = get_vlc2(&s->gb, ff_msmp4_dc_chroma_vlc[s->dc_table_index].table, DC_VLC_BITS, 3);

  2111.     }

  2112.     if (dcdiff < 0){

  2113.         av_log(s->avctx, AV_LOG_ERROR, "Illegal DC VLC\n");

  2114.         return -1;

  2115.     }

  2116.     if (dcdiff)

  2117.     {

  2118.         if (dcdiff == 119 /* ESC index value */)

  2119.         {

  2120.             /* TODO: Optimize */

  2121.             if (mquant == 1) dcdiff = get_bits(gb, 10);

  2122.             else if (mquant == 2) dcdiff = get_bits(gb, 9);

  2123.             else dcdiff = get_bits(gb, 8);

  2124.         }

  2125.         else

  2126.         {

  2127.             if (mquant == 1)

  2128.                 dcdiff = (dcdiff<<2) + get_bits(gb, 2) - 3;

  2129.             else if (mquant == 2)

  2130.                 dcdiff = (dcdiff<<1) + get_bits(gb, 1) - 1;

  2131.         }

  2132.         if (get_bits(gb, 1))

  2133.             dcdiff = -dcdiff;

  2134.     }

  2135. 

  2136.     /* Prediction */

  2137.     dcdiff += vc1_pred_dc(&v->s, v->overlap, mquant, n, a_avail, c_avail, &dc_val, &dc_pred_dir);

  2138.     *dc_val = dcdiff;

  2139. 

  2140.     /* Store the quantized DC coeff, used for prediction */

  2141. 

  2142.     if (n < 4) {

  2143.         block[0] = dcdiff * s->y_dc_scale;

  2144.     } else {

  2145.         block[0] = dcdiff * s->c_dc_scale;

  2146.     }

  2147.     /* Skip ? */

  2148.     run_diff = 0;

  2149.     i = 0;

  2150. 

  2151.     //AC Decoding

  2152.     i = 1;

  2153. 

  2154.     /* check if AC is needed at all and adjust direction if needed */

  2155.     if(!a_avail) dc_pred_dir = 1;

  2156.     if(!c_avail) dc_pred_dir = 0;

  2157.     if(!a_avail && !c_avail) use_pred = 0;

  2158.     ac_val = s->ac_val[0][0] + s->block_index[n] * 16;

  2159.     ac_val2 = ac_val;

  2160. 

  2161.     scale = mquant * 2 + v->halfpq;

  2162. 

  2163.     if(dc_pred_dir) //left

  2164.         ac_val -= 16;

  2165.     else //top

  2166.         ac_val -= 16 * s->block_wrap[n];

  2167. 

  2168.     q1 = s->current_picture.qscale_table[mb_pos];

  2169.     if(dc_pred_dir && c_avail) q2 = s->current_picture.qscale_table[mb_pos - 1];

  2170.     if(!dc_pred_dir && a_avail) q2 = s->current_picture.qscale_table[mb_pos - s->mb_stride];

  2171.     if(n && n<4) q2 = q1;

  2172. 

  2173.     if(coded) {

  2174.         int last = 0, skip, value;

  2175.         const int8_t *zz_table;

  2176.         int k;

  2177. 

  2178.         zz_table = vc1_simple_progressive_8x8_zz;

  2179. 

  2180.         while (!last) {

  2181.             vc1_decode_ac_coeff(v, &last, &skip, &value, codingset);

  2182.             i += skip;

  2183.             if(i > 63)

  2184.                 break;

  2185.             block[zz_table[i++]] = value;

  2186.         }

  2187. 

  2188.         /* apply AC prediction if needed */

  2189.         if(use_pred) {

  2190.             /* scale predictors if needed*/

  2191.             if(q2 && q1!=q2) {

  2192.                 q1 = q1 * 2 - 1;

  2193.                 q2 = q2 * 2 - 1;

  2194. 

  2195.                 if(dc_pred_dir) { //left

  2196.                     for(k = 1; k < 8; k++)

  2197.                         block[k << 3] += (ac_val[k] * q2 * vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2198.                 } else { //top

  2199.                     for(k = 1; k < 8; k++)

  2200.                         block[k] += (ac_val[k + 8] * q2 * vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2201.                 }

  2202.             } else {

  2203.                 if(dc_pred_dir) { //left

  2204.                     for(k = 1; k < 8; k++)

  2205.                         block[k << 3] += ac_val[k];

  2206.                 } else { //top

  2207.                     for(k = 1; k < 8; k++)

  2208.                         block[k] += ac_val[k + 8];

  2209.                 }

  2210.             }

  2211.         }

  2212.         /* save AC coeffs for further prediction */

  2213.         for(k = 1; k < 8; k++) {

  2214.             ac_val2[k] = block[k << 3];

  2215.             ac_val2[k + 8] = block[k];

  2216.         }

  2217. 

  2218.         /* scale AC coeffs */

  2219.         for(k = 1; k < 64; k++)

  2220.             if(block[k]) {

  2221.                 block[k] *= scale;

  2222.                 if(!v->pquantizer)

  2223.                     block[k] += (block[k] < 0) ? -mquant : mquant;

  2224.             }

  2225. 

  2226.         if(use_pred) i = 63;

  2227.     } else { // no AC coeffs

  2228.         int k;

  2229. 

  2230.         memset(ac_val2, 0, 16 * 2);

  2231.         if(dc_pred_dir) {//left

  2232.             if(use_pred) {

  2233.                 memcpy(ac_val2, ac_val, 8 * 2);

  2234.                 if(q2 && q1!=q2) {

  2235.                     q1 = q1 * 2 - 1;

  2236.                     q2 = q2 * 2 - 1;

  2237.                     for(k = 1; k < 8; k++)

  2238.                         ac_val2[k] = (ac_val2[k] * q2 * vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2239.                 }

  2240.             }

  2241.         } else {//top

  2242.             if(use_pred) {

  2243.                 memcpy(ac_val2 + 8, ac_val + 8, 8 * 2);

  2244.                 if(q2 && q1!=q2) {

  2245.                     q1 = q1 * 2 - 1;

  2246.                     q2 = q2 * 2 - 1;

  2247.                     for(k = 1; k < 8; k++)

  2248.                         ac_val2[k + 8] = (ac_val2[k + 8] * q2 * vc1_dqscale[q1 - 1] + 0x20000) >> 18;

  2249.                 }

  2250.             }

  2251.         }

  2252. 

  2253.         /* apply AC prediction if needed */

  2254.         if(use_pred) {

  2255.             if(dc_pred_dir) { //left

  2256.                 for(k = 1; k < 8; k++) {

  2257.                     block[k << 3] = ac_val2[k] * scale;

  2258.                     if(!v->pquantizer && block[k << 3])

  2259.                         block[k << 3] += (block[k << 3] < 0) ? -mquant : mquant;

  2260.                 }

  2261.             } else { //top

  2262.                 for(k = 1; k < 8; k++) {

  2263.                     block[k] = ac_val2[k + 8] * scale;

  2264.                     if(!v->pquantizer && block[k])

  2265.                         block[k] += (block[k] < 0) ? -mquant : mquant;

  2266.                 }

  2267.             }

  2268.             i = 63;

  2269.         }

  2270.     }

  2271.     s->block_last_index[n] = i;

  2272. 

  2273.     return 0;

  2274. }

  2275. 

  2276. /** Decode P block

  2277.  */

  2278. static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquant, int ttmb, int first_block)

  2279. {

  2280.     MpegEncContext *s = &v->s;

  2281.     GetBitContext *gb = &s->gb;

  2282.     int i, j;

  2283.     int subblkpat = 0;

  2284.     int scale, off, idx, last, skip, value;

  2285.     int ttblk = ttmb & 7;

  2286. 

  2287.     if(ttmb == -1) {

  2288.         ttblk = ttblk_to_tt[v->tt_index][get_vlc2(gb, vc1_ttblk_vlc[v->tt_index].table, VC1_TTBLK_VLC_BITS, 1)];

  2289.     }

  2290.     if(ttblk == TT_4X4) {

  2291.         subblkpat = ~(get_vlc2(gb, vc1_subblkpat_vlc[v->tt_index].table, VC1_SUBBLKPAT_VLC_BITS, 1) + 1);

  2292.     }

  2293.     if((ttblk != TT_8X8 && ttblk != TT_4X4) && (v->ttmbf || (ttmb != -1 && (ttmb & 8) && !first_block))) {

  2294.         subblkpat = decode012(gb);

  2295.         if(subblkpat) subblkpat ^= 3; //swap decoded pattern bits

  2296.         if(ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) ttblk = TT_8X4;

  2297.         if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) ttblk = TT_4X8;

  2298.     }

  2299.     scale = 2 * mquant + v->halfpq;

  2300. 

  2301.     // convert transforms like 8X4_TOP to generic TT and SUBBLKPAT

  2302.     if(ttblk == TT_8X4_TOP || ttblk == TT_8X4_BOTTOM) {

  2303.         subblkpat = 2 - (ttblk == TT_8X4_TOP);

  2304.         ttblk = TT_8X4;

  2305.     }

  2306.     if(ttblk == TT_4X8_RIGHT || ttblk == TT_4X8_LEFT) {

  2307.         subblkpat = 2 - (ttblk == TT_4X8_LEFT);

  2308.         ttblk = TT_4X8;

  2309.     }

  2310.     switch(ttblk) {

  2311.     case TT_8X8:

  2312.         i = 0;

  2313.         last = 0;

  2314.         while (!last) {

  2315.             vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  2316.             i += skip;

  2317.             if(i > 63)

  2318.                 break;

  2319.             idx = vc1_simple_progressive_8x8_zz[i++];

  2320.             block[idx] = value * scale;

  2321.             if(!v->pquantizer)

  2322.                 block[idx] += (block[idx] < 0) ? -mquant : mquant;

  2323.         }

  2324.         vc1_inv_trans(block, 8, 8);

  2325.         break;

  2326.     case TT_4X4:

  2327.         for(j = 0; j < 4; j++) {

  2328.             last = subblkpat & (1 << (3 - j));

  2329.             i = 0;

  2330.             off = (j & 1) * 4 + (j & 2) * 16;

  2331.             while (!last) {

  2332.                 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  2333.                 i += skip;

  2334.                 if(i > 15)

  2335.                     break;

  2336.                 idx = vc1_simple_progressive_4x4_zz[i++];

  2337.                 block[idx + off] = value * scale;

  2338.                 if(!v->pquantizer)

  2339.                     block[idx + off] += (block[idx + off] < 0) ? -mquant : mquant;

  2340.             }

  2341.             if(!(subblkpat & (1 << (3 - j))))

  2342.                 vc1_inv_trans(block + off, 4, 4);

  2343.         }

  2344.         break;

  2345.     case TT_8X4:

  2346.         for(j = 0; j < 2; j++) {

  2347.             last = subblkpat & (1 << (1 - j));

  2348.             i = 0;

  2349.             off = j * 32;

  2350.             while (!last) {

  2351.                 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  2352.                 i += skip;

  2353.                 if(i > 31)

  2354.                     break;

  2355.                 idx = vc1_simple_progressive_8x4_zz[i++];

  2356.                 block[idx + off] = value * scale;

  2357.                 if(!v->pquantizer)

  2358.                     block[idx + off] += (block[idx + off] < 0) ? -mquant : mquant;

  2359.             }

  2360.             if(!(subblkpat & (1 << (1 - j))))

  2361.                 vc1_inv_trans(block + off, 8, 4);

  2362.         }

  2363.         break;

  2364.     case TT_4X8:

  2365.         for(j = 0; j < 2; j++) {

  2366.             last = subblkpat & (1 << (1 - j));

  2367.             i = 0;

  2368.             off = j * 4;

  2369.             while (!last) {

  2370.                 vc1_decode_ac_coeff(v, &last, &skip, &value, v->codingset2);

  2371.                 i += skip;

  2372.                 if(i > 31)

  2373.                     break;

  2374.                 idx = vc1_simple_progressive_4x8_zz[i++];

  2375.                 block[idx + off] = value * scale;

  2376.                 if(!v->pquantizer)

  2377.                     block[idx + off] += (block[idx + off] < 0) ? -mquant : mquant;

  2378.             }

  2379.             if(!(subblkpat & (1 << (1 - j))))

  2380.                 vc1_inv_trans(block + off, 4, 8);

  2381.         }

  2382.         break;

  2383.     }

  2384.     return 0;

  2385. }

  2386. 

  2387. 

  2388. /** Decode one P-frame MB (in Simple/Main profile)

  2389.  * @todo TODO: Extend to AP

  2390.  * @fixme FIXME: DC value for inter blocks not set

  2391.  */

  2392. static int vc1_decode_p_mb(VC1Context *v)

  2393. {

  2394.     MpegEncContext *s = &v->s;

  2395.     GetBitContext *gb = &s->gb;

  2396.     int i, j;

  2397.     int mb_pos = s->mb_x + s->mb_y * s->mb_stride;

  2398.     int cbp; /* cbp decoding stuff */

  2399.     int mqdiff, mquant; /* MB quantization */

  2400.     int ttmb = v->ttfrm; /* MB Transform type */

  2401.     int status;

  2402. 

  2403.     static const int size_table[6] = { 0, 2, 3, 4, 5, 8 },

  2404.       offset_table[6] = { 0, 1, 3, 7, 15, 31 };

  2405.     int mb_has_coeffs = 1; /* last_flag */

  2406.     int dmv_x, dmv_y; /* Differential MV components */

  2407.     int index, index1; /* LUT indices */

  2408.     int val, sign; /* temp values */

  2409.     int first_block = 1;

  2410.     int dst_idx, off;

  2411.     int skipped, fourmv;

  2412. 

  2413.     mquant = v->pq; /* Loosy initialization */

  2414. 

  2415.     if (v->mv_type_is_raw)

  2416.         fourmv = get_bits1(gb);

  2417.     else

  2418.         fourmv = v->mv_type_mb_plane[mb_pos];

  2419.     if (v->skip_is_raw)

  2420.         skipped = get_bits1(gb);

  2421.     else

  2422.         skipped = v->s.mbskip_table[mb_pos];

  2423. 

  2424.     s->dsp.clear_blocks(s->block[0]);

  2425. 

  2426.     if (!fourmv) /* 1MV mode */

  2427.     {

  2428.         if (!skipped)

  2429.         {

  2430.             GET_MVDATA(dmv_x, dmv_y);

  2431. 

  2432.             s->current_picture.mb_type[mb_pos] = s->mb_intra ? MB_TYPE_INTRA : MB_TYPE_16x16;

  2433.             vc1_pred_mv(s, 0, dmv_x, dmv_y, 1, v->range_x, v->range_y, v->mb_type[0]);

  2434. 

  2435.             /* FIXME Set DC val for inter block ? */

  2436.             if (s->mb_intra && !mb_has_coeffs)

  2437.             {

  2438.                 GET_MQUANT();

  2439.                 s->ac_pred = get_bits(gb, 1);

  2440.                 cbp = 0;

  2441.             }

  2442.             else if (mb_has_coeffs)

  2443.             {

  2444.                 if (s->mb_intra) s->ac_pred = get_bits(gb, 1);

  2445.                 cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  2446.                 GET_MQUANT();

  2447.             }

  2448.             else

  2449.             {

  2450.                 mquant = v->pq;

  2451.                 cbp = 0;

  2452.             }

  2453.             s->current_picture.qscale_table[mb_pos] = mquant;

  2454. 

  2455.             if (!v->ttmbf && !s->mb_intra && mb_has_coeffs)

  2456.                 ttmb = get_vlc2(gb, vc1_ttmb_vlc[v->tt_index].table,

  2457.                                 VC1_TTMB_VLC_BITS, 2);

  2458.             if(!s->mb_intra) vc1_mc_1mv(v);

  2459.             dst_idx = 0;

  2460.             for (i=0; i<6; i++)

  2461.             {

  2462.                 s->dc_val[0][s->block_index[i]] = 0;

  2463.                 dst_idx += i >> 2;

  2464.                 val = ((cbp >> (5 - i)) & 1);

  2465.                 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  2466.                 v->mb_type[0][s->block_index[i]] = s->mb_intra;

  2467.                 if(s->mb_intra) {

  2468.                     /* check if prediction blocks A and C are available */

  2469.                     v->a_avail = v->c_avail = 0;

  2470.                     if(i == 2 || i == 3 || !s->first_slice_line)

  2471.                         v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  2472.                     if(i == 1 || i == 3 || s->mb_x)

  2473.                         v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  2474. 

  2475.                     vc1_decode_intra_block(v, s->block[i], i, val, mquant, (i&4)?v->codingset2:v->codingset);

  2476.                     vc1_inv_trans(s->block[i], 8, 8);

  2477.                     for(j = 0; j < 64; j++) s->block[i][j] += 128;

  2478.                     s->dsp.put_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));

  2479.                     /* TODO: proper loop filtering */

  2480.                     if(v->pq >= 9 && v->overlap) {

  2481.                         if(v->a_avail)

  2482.                             vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));

  2483.                         if(v->c_avail)

  2484.                             vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));

  2485.                     }

  2486.                 } else if(val) {

  2487.                     vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block);

  2488.                     if(!v->ttmbf && ttmb < 8) ttmb = -1;

  2489.                     first_block = 0;

  2490.                     s->dsp.add_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize);

  2491.                 }

  2492.             }

  2493.         }

  2494.         else //Skipped

  2495.         {

  2496.             s->mb_intra = 0;

  2497.             for(i = 0; i < 6; i++) {

  2498.                 v->mb_type[0][s->block_index[i]] = 0;

  2499.                 s->dc_val[0][s->block_index[i]] = 0;

  2500.             }

  2501.             s->current_picture.mb_type[mb_pos] = MB_TYPE_SKIP;

  2502.             s->current_picture.qscale_table[mb_pos] = 0;

  2503.             vc1_pred_mv(s, 0, 0, 0, 1, v->range_x, v->range_y, v->mb_type[0]);

  2504.             vc1_mc_1mv(v);

  2505.             return 0;

  2506.         }

  2507.     } //1MV mode

  2508.     else //4MV mode

  2509.     {

  2510.         if (!skipped /* unskipped MB */)

  2511.         {

  2512.             int intra_count = 0, coded_inter = 0;

  2513.             int is_intra[6], is_coded[6];

  2514.             /* Get CBPCY */

  2515.             cbp = get_vlc2(&v->s.gb, v->cbpcy_vlc->table, VC1_CBPCY_P_VLC_BITS, 2);

  2516.             for (i=0; i<6; i++)

  2517.             {

  2518.                 val = ((cbp >> (5 - i)) & 1);

  2519.                 s->dc_val[0][s->block_index[i]] = 0;

  2520.                 s->mb_intra = 0;

  2521.                 if(i < 4) {

  2522.                     dmv_x = dmv_y = 0;

  2523.                     s->mb_intra = 0;

  2524.                     mb_has_coeffs = 0;

  2525.                     if(val) {

  2526.                         GET_MVDATA(dmv_x, dmv_y);

  2527.                     }

  2528.                     vc1_pred_mv(s, i, dmv_x, dmv_y, 0, v->range_x, v->range_y, v->mb_type[0]);

  2529.                     if(!s->mb_intra) vc1_mc_4mv_luma(v, i);

  2530.                     intra_count += s->mb_intra;

  2531.                     is_intra[i] = s->mb_intra;

  2532.                     is_coded[i] = mb_has_coeffs;

  2533.                 }

  2534.                 if(i&4){

  2535.                     is_intra[i] = (intra_count >= 3);

  2536.                     is_coded[i] = val;

  2537.                 }

  2538.                 if(i == 4) vc1_mc_4mv_chroma(v);

  2539.                 v->mb_type[0][s->block_index[i]] = is_intra[i];

  2540.                 if(!coded_inter) coded_inter = !is_intra[i] & is_coded[i];

  2541.             }

  2542.             // if there are no coded blocks then don't do anything more

  2543.             if(!intra_count && !coded_inter) return;

  2544.             dst_idx = 0;

  2545.             GET_MQUANT();

  2546.             s->current_picture.qscale_table[mb_pos] = mquant;

  2547.             /* test if block is intra and has pred */

  2548.             {

  2549.                 int intrapred = 0;

  2550.                 for(i=0; i<6; i++)

  2551.                     if(is_intra[i]) {

  2552.                         if(((!s->first_slice_line || (i==2 || i==3)) && v->mb_type[0][s->block_index[i] - s->block_wrap[i]])

  2553.                             || ((s->mb_x || (i==1 || i==3)) && v->mb_type[0][s->block_index[i] - 1])) {

  2554.                             intrapred = 1;

  2555.                             break;

  2556.                         }

  2557.                     }

  2558.                 if(intrapred)s->ac_pred = get_bits(gb, 1);

  2559.                 else s->ac_pred = 0;

  2560.             }

  2561.             if (!v->ttmbf && coded_inter)

  2562.                 ttmb = get_vlc2(gb, vc1_ttmb_vlc[v->tt_index].table, VC1_TTMB_VLC_BITS, 12);

  2563.             for (i=0; i<6; i++)

  2564.             {

  2565.                 dst_idx += i >> 2;

  2566.                 off = (i & 4) ? 0 : ((i & 1) * 8 + (i & 2) * 4 * s->linesize);

  2567.                 s->mb_intra = is_intra[i];

  2568.                 if (is_intra[i]) {

  2569.                     /* check if prediction blocks A and C are available */

  2570.                     v->a_avail = v->c_avail = 0;

  2571.                     if(i == 2 || i == 3 || !s->first_slice_line)

  2572.                         v->a_avail = v->mb_type[0][s->block_index[i] - s->block_wrap[i]];

  2573.                     if(i == 1 || i == 3 || s->mb_x)

  2574.                         v->c_avail = v->mb_type[0][s->block_index[i] - 1];

  2575. 

  2576.                     vc1_decode_intra_block(v, s->block[i], i, is_coded[i], mquant, (i&4)?v->codingset2:v->codingset);

  2577.                     vc1_inv_trans(s->block[i], 8, 8);

  2578.                     for(j = 0; j < 64; j++) s->block[i][j] += 128;

  2579.                     s->dsp.put_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize);

  2580.                     /* TODO: proper loop filtering */

  2581.                     if(v->pq >= 9 && v->overlap) {

  2582.                         if(v->a_avail)

  2583.                             vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));

  2584.                         if(v->c_avail)

  2585.                             vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));

  2586.                     }

  2587.                 } else if(is_coded[i]) {

  2588.                     status = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block);

  2589.                     if(!v->ttmbf && ttmb < 8) ttmb = -1;

  2590.                     first_block = 0;

  2591.                     s->dsp.add_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize);

  2592.                 }

  2593.             }

  2594.             return status;

  2595.         }

  2596.         else //Skipped MB

  2597.         {

  2598.             s->mb_intra = 0;

  2599.             for (i=0; i<6; i++) {

  2600.                 v->mb_type[0][s->block_index[i]] = 0;

  2601.                 s->dc_val[0][s->block_index[i]] = 0;

  2602.             }

  2603.             for (i=0; i<4; i++)

  2604.             {

  2605.                 vc1_pred_mv(s, i, 0, 0, 0, v->range_x, v->range_y, v->mb_type[0]);

  2606.                 vc1_mc_4mv_luma(v, i);

  2607.             }

  2608.             vc1_mc_4mv_chroma(v);

  2609.             s->current_picture.qscale_table[mb_pos] = 0;

  2610.             return 0;

  2611.         }

  2612.     }

  2613. 

  2614.     /* Should never happen */

  2615.     return -1;

  2616. }

  2617. 

  2618. /** Decode blocks of I-frame

  2619.  */

  2620. static void vc1_decode_i_blocks(VC1Context *v)

  2621. {

  2622.     int k, j;

  2623.     MpegEncContext *s = &v->s;

  2624.     int cbp, val;

  2625.     uint8_t *coded_val;

  2626.     int mb_pos;

  2627. 

  2628.     /* select codingmode used for VLC tables selection */

  2629.     switch(v->y_ac_table_index){

  2630.     case 0:

  2631.         v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;

  2632.         break;

  2633.     case 1:

  2634.         v->codingset = CS_HIGH_MOT_INTRA;

  2635.         break;

  2636.     case 2:

  2637.         v->codingset = CS_MID_RATE_INTRA;

  2638.         break;

  2639.     }

  2640. 

  2641.     switch(v->c_ac_table_index){

  2642.     case 0:

  2643.         v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;

  2644.         break;

  2645.     case 1:

  2646.         v->codingset2 = CS_HIGH_MOT_INTER;

  2647.         break;

  2648.     case 2:

  2649.         v->codingset2 = CS_MID_RATE_INTER;

  2650.         break;

  2651.     }

  2652. 

  2653.     /* Set DC scale - y and c use the same */

  2654.     s->y_dc_scale = s->y_dc_scale_table[v->pq];

  2655.     s->c_dc_scale = s->c_dc_scale_table[v->pq];

  2656. 

  2657.     //do frame decode

  2658.     s->mb_x = s->mb_y = 0;

  2659.     s->mb_intra = 1;

  2660.     s->first_slice_line = 1;

  2661.     ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));

  2662.     for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {

  2663.         for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {

  2664.             ff_init_block_index(s);

  2665.             ff_update_block_index(s);

  2666.             s->dsp.clear_blocks(s->block[0]);

  2667.             mb_pos = s->mb_x + s->mb_y * s->mb_width;

  2668.             s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA;

  2669.             s->current_picture.qscale_table[mb_pos] = v->pq;

  2670. 

  2671.             // do actual MB decoding and displaying

  2672.             cbp = get_vlc2(&v->s.gb, ff_msmp4_mb_i_vlc.table, MB_INTRA_VLC_BITS, 2);

  2673.             v->s.ac_pred = get_bits(&v->s.gb, 1);

  2674. 

  2675.             for(k = 0; k < 6; k++) {

  2676.                 val = ((cbp >> (5 - k)) & 1);

  2677. 

  2678.                 if (k < 4) {

  2679.                     int pred = vc1_coded_block_pred(&v->s, k, &coded_val);

  2680.                     val = val ^ pred;

  2681.                     *coded_val = val;

  2682.                 }

  2683.                 cbp |= val << (5 - k);

  2684. 

  2685.                 vc1_decode_i_block(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2);

  2686. 

  2687.                 vc1_inv_trans(s->block[k], 8, 8);

  2688.                 if(v->pq >= 9 && v->overlap) {

  2689.                     for(j = 0; j < 64; j++) s->block[k][j] += 128;

  2690.                 }

  2691.             }

  2692. 

  2693.             vc1_put_block(v, s->block);

  2694.             if(v->pq >= 9 && v->overlap) { /* XXX: do proper overlapping insted of loop filter */

  2695.                 if(!s->first_slice_line) {

  2696.                     vc1_v_overlap(s->dest[0], s->linesize);

  2697.                     vc1_v_overlap(s->dest[0] + 8, s->linesize);

  2698.                     vc1_v_overlap(s->dest[1], s->uvlinesize);

  2699.                     vc1_v_overlap(s->dest[2], s->uvlinesize);

  2700.                 }

  2701.                 vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize);

  2702.                 vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);

  2703.                 if(s->mb_x) {

  2704.                     vc1_h_overlap(s->dest[0], s->linesize);

  2705.                     vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize);

  2706.                     vc1_h_overlap(s->dest[1], s->uvlinesize);

  2707.                     vc1_h_overlap(s->dest[2], s->uvlinesize);

  2708.                 }

  2709.                 vc1_h_overlap(s->dest[0] + 8, s->linesize);

  2710.                 vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize);

  2711.             }

  2712. 

  2713.             if(get_bits_count(&s->gb) > v->bits) {

  2714.                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits);

  2715.                 return;

  2716.             }

  2717.         }

  2718.         ff_draw_horiz_band(s, s->mb_y * 16, 16);

  2719.         s->first_slice_line = 0;

  2720.     }

  2721. }

  2722. 

  2723. static void vc1_decode_p_blocks(VC1Context *v)

  2724. {

  2725.     MpegEncContext *s = &v->s;

  2726. 

  2727.     /* select codingmode used for VLC tables selection */

  2728.     switch(v->c_ac_table_index){

  2729.     case 0:

  2730.         v->codingset = (v->pqindex <= 8) ? CS_HIGH_RATE_INTRA : CS_LOW_MOT_INTRA;

  2731.         break;

  2732.     case 1:

  2733.         v->codingset = CS_HIGH_MOT_INTRA;

  2734.         break;

  2735.     case 2:

  2736.         v->codingset = CS_MID_RATE_INTRA;

  2737.         break;

  2738.     }

  2739. 

  2740.     switch(v->c_ac_table_index){

  2741.     case 0:

  2742.         v->codingset2 = (v->pqindex <= 8) ? CS_HIGH_RATE_INTER : CS_LOW_MOT_INTER;

  2743.         break;

  2744.     case 1:

  2745.         v->codingset2 = CS_HIGH_MOT_INTER;

  2746.         break;

  2747.     case 2:

  2748.         v->codingset2 = CS_MID_RATE_INTER;

  2749.         break;

  2750.     }

  2751. 

  2752.     ff_er_add_slice(s, 0, 0, s->mb_width - 1, s->mb_height - 1, (AC_END|DC_END|MV_END));

  2753.     s->first_slice_line = 1;

  2754.     for(s->mb_y = 0; s->mb_y < s->mb_height; s->mb_y++) {

  2755.         for(s->mb_x = 0; s->mb_x < s->mb_width; s->mb_x++) {

  2756.             ff_init_block_index(s);

  2757.             ff_update_block_index(s);

  2758.             s->dsp.clear_blocks(s->block[0]);

  2759. 

  2760.             vc1_decode_p_mb(v);

  2761.             if(get_bits_count(&s->gb) > v->bits || get_bits_count(&s->gb) < 0) {

  2762.                 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i at %ix%i\n", get_bits_count(&s->gb), v->bits,s->mb_x,s->mb_y);

  2763.                 return;

  2764.             }

  2765.         }

  2766.         ff_draw_horiz_band(s, s->mb_y * 16, 16);

  2767.         s->first_slice_line = 0;

  2768.     }

  2769. }

  2770. 

  2771. static void vc1_decode_blocks(VC1Context *v)

  2772. {

  2773. 

  2774.     v->s.esc3_level_length = 0;

  2775. 

  2776.     switch(v->s.pict_type) {

  2777.     case I_TYPE:

  2778.         vc1_decode_i_blocks(v);

  2779.         break;

  2780.     case P_TYPE:

  2781.         vc1_decode_p_blocks(v);

  2782.         break;

  2783.     }

  2784. }

  2785. 

  2786. 

  2787. /** Initialize a VC1/WMV3 decoder

  2788.  * @todo TODO: Handle VC-1 IDUs (Transport level?)

  2789.  * @todo TODO: Decypher remaining bits in extra_data

  2790.  */

  2791. static int vc1_decode_init(AVCodecContext *avctx)

  2792. {

  2793.     VC1Context *v = avctx->priv_data;

  2794.     MpegEncContext *s = &v->s;

  2795.     GetBitContext gb;

  2796. 

  2797.     if (!avctx->extradata_size || !avctx->extradata) return -1;

  2798.     avctx->pix_fmt = PIX_FMT_YUV420P;

  2799.     v->s.avctx = avctx;

  2800.     avctx->flags |= CODEC_FLAG_EMU_EDGE;

  2801.     v->s.flags |= CODEC_FLAG_EMU_EDGE;

  2802. 

  2803.     if(ff_h263_decode_init(avctx) < 0)

  2804.         return -1;

  2805.     if (vc1_init_common(v) < 0) return -1;

  2806. 

  2807.     avctx->coded_width = avctx->width;

  2808.     avctx->coded_height = avctx->height;

  2809.     if (avctx->codec_id == CODEC_ID_WMV3)

  2810.     {

  2811.         int count = 0;

  2812. 

  2813.         // looks like WMV3 has a sequence header stored in the extradata

  2814.         // advanced sequence header may be before the first frame

  2815.         // the last byte of the extradata is a version number, 1 for the

  2816.         // samples we can decode

  2817. 

  2818.         init_get_bits(&gb, avctx->extradata, avctx->extradata_size*8);

  2819. 

  2820.         if (decode_sequence_header(avctx, &gb) < 0)

  2821.           return -1;

  2822. 

  2823.         count = avctx->extradata_size*8 - get_bits_count(&gb);

  2824.         if (count>0)

  2825.         {

  2826.             av_log(avctx, AV_LOG_INFO, "Extra data: %i bits left, value: %X\n",

  2827.                    count, get_bits(&gb, count));

  2828.         }

  2829.         else if (count < 0)

  2830.         {

  2831.             av_log(avctx, AV_LOG_INFO, "Read %i bits in overflow\n", -count);

  2832.         }

  2833.     }

  2834.     avctx->has_b_frames= !!(avctx->max_b_frames);

  2835. 

  2836.     s->mb_width = (avctx->coded_width+15)>>4;

  2837.     s->mb_height = (avctx->coded_height+15)>>4;

  2838. 

  2839.     /* Allocate mb bitplanes */

  2840.     v->mv_type_mb_plane = av_malloc(s->mb_stride * s->mb_height);

  2841. 

  2842.     /* allocate block type info in that way so it could be used with s->block_index[] */

  2843.     v->mb_type_base = av_malloc(s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride * (s->mb_height + 1) * 2);

  2844.     v->mb_type[0] = v->mb_type_base + s->b8_stride + 1;

  2845.     v->mb_type[1] = v->mb_type_base + s->b8_stride * (s->mb_height * 2 + 1) + s->mb_stride + 1;

  2846.     v->mb_type[2] = v->mb_type[1] + s->mb_stride * (s->mb_height + 1);

  2847. 

  2848.     /* Init coded blocks info */

  2849.     if (v->profile == PROFILE_ADVANCED)

  2850.     {

  2851. //        if (alloc_bitplane(&v->over_flags_plane, s->mb_width, s->mb_height) < 0)

  2852. //            return -1;

  2853. //        if (alloc_bitplane(&v->ac_pred_plane, s->mb_width, s->mb_height) < 0)

  2854. //            return -1;

  2855.     }

  2856. 

  2857.     return 0;

  2858. }

  2859. 

  2860. 

  2861. /** Decode a VC1/WMV3 frame

  2862.  * @todo TODO: Handle VC-1 IDUs (Transport level?)

  2863.  * @warning Initial try at using MpegEncContext stuff

  2864.  */

  2865. static int vc1_decode_frame(AVCodecContext *avctx,

  2866.                             void *data, int *data_size,

  2867.                             uint8_t *buf, int buf_size)

  2868. {

  2869.     VC1Context *v = avctx->priv_data;

  2870.     MpegEncContext *s = &v->s;

  2871.     AVFrame *pict = data;

  2872. 

  2873.     /* no supplementary picture */

  2874.     if (buf_size == 0) {

  2875.         /* special case for last picture */

  2876.         if (s->low_delay==0 && s->next_picture_ptr) {

  2877.             *pict= *(AVFrame*)s->next_picture_ptr;

  2878.             s->next_picture_ptr= NULL;

  2879. 

  2880.             *data_size = sizeof(AVFrame);

  2881.         }

  2882. 

  2883.         return 0;

  2884.     }

  2885. 

  2886.     //we need to set current_picture_ptr before reading the header, otherwise we cant store anyting im there

  2887.     if(s->current_picture_ptr==NULL || s->current_picture_ptr->data[0]){

  2888.         int i= ff_find_unused_picture(s, 0);

  2889.         s->current_picture_ptr= &s->picture[i];

  2890.     }

  2891. 

  2892.     avctx->has_b_frames= !s->low_delay;

  2893. 

  2894.     init_get_bits(&s->gb, buf, buf_size*8);

  2895.     // do parse frame header

  2896.     if(vc1_parse_frame_header(v, &s->gb) == -1)

  2897.         return -1;

  2898. 

  2899.     if(s->pict_type != I_TYPE && s->pict_type != P_TYPE)return -1;

  2900. 

  2901.     // for hurry_up==5

  2902.     s->current_picture.pict_type= s->pict_type;

  2903.     s->current_picture.key_frame= s->pict_type == I_TYPE;

  2904. 

  2905.     /* skip B-frames if we don't have reference frames */

  2906.     if(s->last_picture_ptr==NULL && (s->pict_type==B_TYPE || s->dropable)) return -1;//buf_size;

  2907.     /* skip b frames if we are in a hurry */

  2908.     if(avctx->hurry_up && s->pict_type==B_TYPE) return -1;//buf_size;

  2909.     if(   (avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type==B_TYPE)

  2910.        || (avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type!=I_TYPE)

  2911.        ||  avctx->skip_frame >= AVDISCARD_ALL)

  2912.         return buf_size;

  2913.     /* skip everything if we are in a hurry>=5 */

  2914.     if(avctx->hurry_up>=5) return -1;//buf_size;

  2915. 

  2916.     if(s->next_p_frame_damaged){

  2917.         if(s->pict_type==B_TYPE)

  2918.             return buf_size;

  2919.         else

  2920.             s->next_p_frame_damaged=0;

  2921.     }

  2922. 

  2923.     if(MPV_frame_start(s, avctx) < 0)

  2924.         return -1;

  2925. 

  2926.     ff_er_frame_start(s);

  2927. 

  2928.     v->bits = buf_size * 8;

  2929.     vc1_decode_blocks(v);

  2930. //av_log(s->avctx, AV_LOG_INFO, "Consumed %i/%i bits\n", get_bits_count(&s->gb), buf_size*8);

  2931. //  if(get_bits_count(&s->gb) > buf_size * 8)

  2932. //      return -1;

  2933.     ff_er_frame_end(s);

  2934. 

  2935.     MPV_frame_end(s);

  2936. 

  2937. assert(s->current_picture.pict_type == s->current_picture_ptr->pict_type);

  2938. assert(s->current_picture.pict_type == s->pict_type);

  2939.     if (s->pict_type == B_TYPE || s->low_delay) {

  2940.         *pict= *(AVFrame*)s->current_picture_ptr;

  2941.     } else if (s->last_picture_ptr != NULL) {

  2942.         *pict= *(AVFrame*)s->last_picture_ptr;

  2943.     }

  2944. 

  2945.     if(s->last_picture_ptr || s->low_delay){

  2946.         *data_size = sizeof(AVFrame);

  2947.         ff_print_debug_info(s, pict);

  2948.     }

  2949. 

  2950.     /* Return the Picture timestamp as the frame number */

  2951.     /* we substract 1 because it is added on utils.c    */

  2952.     avctx->frame_number = s->picture_number - 1;

  2953. 

  2954.     return buf_size;

  2955. }

  2956. 

  2957. 

  2958. /** Close a VC1/WMV3 decoder

  2959.  * @warning Initial try at using MpegEncContext stuff

  2960.  */

  2961. static int vc1_decode_end(AVCodecContext *avctx)

  2962. {

  2963.     VC1Context *v = avctx->priv_data;

  2964. 

  2965.     av_freep(&v->hrd_rate);

  2966.     av_freep(&v->hrd_buffer);

  2967.     MPV_common_end(&v->s);

  2968.     av_freep(&v->mv_type_mb_plane);

  2969.     av_freep(&v->mb_type_base);

  2970.     return 0;

  2971. }

  2972. 

  2973. 

  2974. AVCodec vc1_decoder = {

  2975.     "vc1",

  2976.     CODEC_TYPE_VIDEO,

  2977.     CODEC_ID_VC1,

  2978.     sizeof(VC1Context),

  2979.     vc1_decode_init,

  2980.     NULL,

  2981.     vc1_decode_end,

  2982.     vc1_decode_frame,

  2983.     CODEC_CAP_DELAY,

  2984.     NULL

  2985. };

  2986. 

  2987. AVCodec wmv3_decoder = {

  2988.     "wmv3",

  2989.     CODEC_TYPE_VIDEO,

  2990.     CODEC_ID_WMV3,

  2991.     sizeof(VC1Context),

  2992.     vc1_decode_init,

  2993.     NULL,

  2994.     vc1_decode_end,

  2995.     vc1_decode_frame,

  2996.     CODEC_CAP_DELAY,

  2997.     NULL

  2998. };