Subversion Repositories Kolibri OS

Rev

Go to most recent revision | Blame | Last modification | View Log | RSS feed

  1. /*
  2.  * This file is part of FFmpeg.
  3.  *
  4.  * FFmpeg is free software; you can redistribute it and/or
  5.  * modify it under the terms of the GNU Lesser General Public
  6.  * License as published by the Free Software Foundation; either
  7.  * version 2.1 of the License, or (at your option) any later version.
  8.  *
  9.  * FFmpeg is distributed in the hope that it will be useful,
  10.  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  11.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  12.  * Lesser General Public License for more details.
  13.  *
  14.  * You should have received a copy of the GNU Lesser General Public
  15.  * License along with FFmpeg; if not, write to the Free Software
  16.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  17.  */
  18.  
  19. /**
  20.  * @file
  21.  * @brief IntraX8 (J-Frame) subdecoder, used by WMV2 and VC-1
  22.  */
  23.  
  24. #include "libavutil/avassert.h"
  25. #include "avcodec.h"
  26. #include "error_resilience.h"
  27. #include "get_bits.h"
  28. #include "mpegvideo.h"
  29. #include "msmpeg4data.h"
  30. #include "intrax8huf.h"
  31. #include "intrax8.h"
  32. #include "intrax8dsp.h"
  33.  
  34. #define MAX_TABLE_DEPTH(table_bits, max_bits) ((max_bits+table_bits-1)/table_bits)
  35.  
  36. #define DC_VLC_BITS 9
  37. #define AC_VLC_BITS 9
  38. #define OR_VLC_BITS 7
  39.  
  40. #define DC_VLC_MTD MAX_TABLE_DEPTH(DC_VLC_BITS, MAX_DC_VLC_BITS)
  41. #define AC_VLC_MTD MAX_TABLE_DEPTH(AC_VLC_BITS, MAX_AC_VLC_BITS)
  42. #define OR_VLC_MTD MAX_TABLE_DEPTH(OR_VLC_BITS, MAX_OR_VLC_BITS)
  43.  
  44. static VLC j_ac_vlc[2][2][8];  //[quant<13],[intra/inter],[select]
  45. static VLC j_dc_vlc[2][8];     //[quant], [select]
  46. static VLC j_orient_vlc[2][4]; //[quant], [select]
  47.  
  48. static av_cold void x8_vlc_init(void){
  49.     int i;
  50.     int offset = 0;
  51.     int sizeidx = 0;
  52.     static const uint16_t sizes[8*4 + 8*2 + 2 + 4] = {
  53.         576, 548, 582, 618, 546, 616, 560, 642,
  54.         584, 582, 704, 664, 512, 544, 656, 640,
  55.         512, 648, 582, 566, 532, 614, 596, 648,
  56.         586, 552, 584, 590, 544, 578, 584, 624,
  57.  
  58.         528, 528, 526, 528, 536, 528, 526, 544,
  59.         544, 512, 512, 528, 528, 544, 512, 544,
  60.  
  61.         128, 128, 128, 128, 128, 128};
  62.  
  63.     static VLC_TYPE table[28150][2];
  64.  
  65. #define  init_ac_vlc(dst,src) \
  66.     dst.table = &table[offset]; \
  67.     dst.table_allocated = sizes[sizeidx]; \
  68.     offset += sizes[sizeidx++]; \
  69.        init_vlc(&dst, \
  70.               AC_VLC_BITS,77, \
  71.               &src[1],4,2, \
  72.               &src[0],4,2, \
  73.               INIT_VLC_USE_NEW_STATIC)
  74. //set ac tables
  75.     for(i=0;i<8;i++){
  76.         init_ac_vlc( j_ac_vlc[0][0][i], x8_ac0_highquant_table[i][0] );
  77.         init_ac_vlc( j_ac_vlc[0][1][i], x8_ac1_highquant_table[i][0] );
  78.         init_ac_vlc( j_ac_vlc[1][0][i], x8_ac0_lowquant_table [i][0] );
  79.         init_ac_vlc( j_ac_vlc[1][1][i], x8_ac1_lowquant_table [i][0] );
  80.     }
  81. #undef init_ac_vlc
  82.  
  83. //set dc tables
  84. #define init_dc_vlc(dst,src) \
  85.     dst.table = &table[offset]; \
  86.     dst.table_allocated = sizes[sizeidx]; \
  87.     offset += sizes[sizeidx++]; \
  88.         init_vlc(&dst, \
  89.         DC_VLC_BITS,34, \
  90.         &src[1],4,2, \
  91.         &src[0],4,2, \
  92.         INIT_VLC_USE_NEW_STATIC);
  93.     for(i=0;i<8;i++){
  94.         init_dc_vlc( j_dc_vlc[0][i], x8_dc_highquant_table[i][0]);
  95.         init_dc_vlc( j_dc_vlc[1][i], x8_dc_lowquant_table [i][0]);
  96.     }
  97. #undef init_dc_vlc
  98.  
  99. //set orient tables
  100. #define init_or_vlc(dst,src) \
  101.     dst.table = &table[offset]; \
  102.     dst.table_allocated = sizes[sizeidx]; \
  103.     offset += sizes[sizeidx++]; \
  104.     init_vlc(&dst, \
  105.     OR_VLC_BITS,12, \
  106.     &src[1],4,2, \
  107.     &src[0],4,2, \
  108.     INIT_VLC_USE_NEW_STATIC);
  109.     for(i=0;i<2;i++){
  110.         init_or_vlc( j_orient_vlc[0][i], x8_orient_highquant_table[i][0]);
  111.     }
  112.     for(i=0;i<4;i++){
  113.         init_or_vlc( j_orient_vlc[1][i], x8_orient_lowquant_table [i][0])
  114.     }
  115.     if (offset != sizeof(table)/sizeof(VLC_TYPE)/2)
  116.         av_log(NULL, AV_LOG_ERROR, "table size %i does not match needed %i\n", (int)(sizeof(table)/sizeof(VLC_TYPE)/2), offset);
  117. }
  118. #undef init_or_vlc
  119.  
  120. static void x8_reset_vlc_tables(IntraX8Context * w){
  121.     memset(w->j_dc_vlc,0,sizeof(w->j_dc_vlc));
  122.     memset(w->j_ac_vlc,0,sizeof(w->j_ac_vlc));
  123.     w->j_orient_vlc=NULL;
  124. }
  125.  
  126. static inline void x8_select_ac_table(IntraX8Context * const w , int mode){
  127.     MpegEncContext * const s= w->s;
  128.     int table_index;
  129.  
  130.     av_assert2(mode<4);
  131.  
  132.     if( w->j_ac_vlc[mode] ) return;
  133.  
  134.     table_index = get_bits(&s->gb, 3);
  135.     w->j_ac_vlc[mode] = &j_ac_vlc[w->quant<13][mode>>1][table_index];//2 modes use same tables
  136.     av_assert2(w->j_ac_vlc[mode]);
  137. }
  138.  
  139. static inline int x8_get_orient_vlc(IntraX8Context * w){
  140.     MpegEncContext * const s= w->s;
  141.     int table_index;
  142.  
  143.     if(!w->j_orient_vlc ){
  144.         table_index = get_bits(&s->gb, 1+(w->quant<13) );
  145.         w->j_orient_vlc = &j_orient_vlc[w->quant<13][table_index];
  146.     }
  147.  
  148.     return get_vlc2(&s->gb, w->j_orient_vlc->table, OR_VLC_BITS, OR_VLC_MTD);
  149. }
  150.  
  151. #define extra_bits(eb) (eb)
  152. #define extra_run   (0xFF<<8)
  153. #define extra_level (0x00<<8)
  154. #define   run_offset(r)    ((r)<<16)
  155. #define level_offset(l)    ((l)<<24)
  156. static const uint32_t ac_decode_table[]={
  157.     /*46*/ extra_bits(3) |  extra_run  | run_offset(16) | level_offset( 0),
  158.     /*47*/ extra_bits(3) |  extra_run  | run_offset(24) | level_offset( 0),
  159.     /*48*/ extra_bits(2) |  extra_run  | run_offset( 4) | level_offset( 1),
  160.     /*49*/ extra_bits(3) |  extra_run  | run_offset( 8) | level_offset( 1),
  161.  
  162.     /*50*/ extra_bits(5) |  extra_run  | run_offset(32) | level_offset( 0),
  163.     /*51*/ extra_bits(4) |  extra_run  | run_offset(16) | level_offset( 1),
  164.  
  165.     /*52*/ extra_bits(2) | extra_level | run_offset( 0) | level_offset( 4),
  166.     /*53*/ extra_bits(2) | extra_level | run_offset( 0) | level_offset( 8),
  167.     /*54*/ extra_bits(2) | extra_level | run_offset( 0) | level_offset(12),
  168.     /*55*/ extra_bits(3) | extra_level | run_offset( 0) | level_offset(16),
  169.     /*56*/ extra_bits(3) | extra_level | run_offset( 0) | level_offset(24),
  170.  
  171.     /*57*/ extra_bits(2) | extra_level | run_offset( 1) | level_offset( 3),
  172.     /*58*/ extra_bits(3) | extra_level | run_offset( 1) | level_offset( 7),
  173.  
  174.     /*59*/ extra_bits(2) |  extra_run  | run_offset(16) | level_offset( 0),
  175.     /*60*/ extra_bits(2) |  extra_run  | run_offset(20) | level_offset( 0),
  176.     /*61*/ extra_bits(2) |  extra_run  | run_offset(24) | level_offset( 0),
  177.     /*62*/ extra_bits(2) |  extra_run  | run_offset(28) | level_offset( 0),
  178.     /*63*/ extra_bits(4) |  extra_run  | run_offset(32) | level_offset( 0),
  179.     /*64*/ extra_bits(4) |  extra_run  | run_offset(48) | level_offset( 0),
  180.  
  181.     /*65*/ extra_bits(2) |  extra_run  | run_offset( 4) | level_offset( 1),
  182.     /*66*/ extra_bits(3) |  extra_run  | run_offset( 8) | level_offset( 1),
  183.     /*67*/ extra_bits(4) |  extra_run  | run_offset(16) | level_offset( 1),
  184.  
  185.     /*68*/ extra_bits(2) | extra_level | run_offset( 0) | level_offset( 4),
  186.     /*69*/ extra_bits(3) | extra_level | run_offset( 0) | level_offset( 8),
  187.     /*70*/ extra_bits(4) | extra_level | run_offset( 0) | level_offset(16),
  188.  
  189.     /*71*/ extra_bits(2) | extra_level | run_offset( 1) | level_offset( 3),
  190.     /*72*/ extra_bits(3) | extra_level | run_offset( 1) | level_offset( 7),
  191. };
  192. //extra_bits = 3bits; extra_run/level = 1 bit; run_offset = 6bits; level_offset = 5 bits;
  193. #undef extra_bits
  194. #undef extra_run
  195. #undef extra_level
  196. #undef run_offset
  197. #undef level_offset
  198.  
  199. static void x8_get_ac_rlf(IntraX8Context * const w, const int mode,
  200.                      int * const run, int * const level, int * const final){
  201.     MpegEncContext *  const s= w->s;
  202.     int i,e;
  203.  
  204. //    x8_select_ac_table(w,mode);
  205.     i = get_vlc2(&s->gb, w->j_ac_vlc[mode]->table, AC_VLC_BITS, AC_VLC_MTD);
  206.  
  207.     if(i<46){ //[0-45]
  208.         int t,l;
  209.         if(i<0){
  210.             (*level)=(*final)=//prevent 'may be used unilitialized'
  211.             (*run)=64;//this would cause error exit in the ac loop
  212.             return;
  213.         }
  214.  
  215.         (*final) = t = (i>22);
  216.         i-=23*t;
  217. /*
  218.   i== 0-15 r=0-15 l=0 ;r=i& %01111
  219.   i==16-19 r=0-3  l=1 ;r=i& %00011
  220.   i==20-21 r=0-1  l=2 ;r=i& %00001
  221.   i==22    r=0    l=3 ;r=i& %00000
  222. l=lut_l[i/2]={0,0,0,0,0,0,0,0,1,1,2,3}[i>>1];// 11 10'01 01'00 00'00 00'00 00'00 00 => 0xE50000
  223. t=lut_mask[l]={0x0f,0x03,0x01,0x00}[l]; as i<256 the higher bits do not matter */
  224.         l=(0xE50000>>(i&(0x1E)))&3;/*0x1E or (~1) or ((i>>1)<<1)*/
  225.         t=(0x01030F>>(l<<3));
  226.  
  227.         (*run)   = i&t;
  228.         (*level) = l;
  229.     }else if(i<73){//[46-72]
  230.         uint32_t sm;
  231.         uint32_t mask;
  232.  
  233.         i-=46;
  234.         sm=ac_decode_table[i];
  235.  
  236.         e=get_bits(&s->gb,sm&0xF);sm>>=8;//3bits
  237.         mask=sm&0xff;sm>>=8;             //1bit
  238.  
  239.         (*run)  =(sm&0xff) + (e&( mask));//6bits
  240.         (*level)=(sm>>8)   + (e&(~mask));//5bits
  241.         (*final)=i>(58-46);
  242.     }else if(i<75){//[73-74]
  243.         static const uint8_t crazy_mix_runlevel[32]={
  244.         0x22,0x32,0x33,0x53,0x23,0x42,0x43,0x63,
  245.         0x24,0x52,0x34,0x73,0x25,0x62,0x44,0x83,
  246.         0x26,0x72,0x35,0x54,0x27,0x82,0x45,0x64,
  247.         0x28,0x92,0x36,0x74,0x29,0xa2,0x46,0x84};
  248.  
  249.         (*final)=!(i&1);
  250.         e=get_bits(&s->gb,5);//get the extra bits
  251.         (*run)  =crazy_mix_runlevel[e]>>4;
  252.         (*level)=crazy_mix_runlevel[e]&0x0F;
  253.     }else{
  254.         (*level)=get_bits( &s->gb, 7-3*(i&1));
  255.         (*run)  =get_bits( &s->gb, 6);
  256.         (*final)=get_bits1(&s->gb);
  257.     }
  258.     return;
  259. }
  260.  
  261. //static const uint8_t dc_extra_sbits[]   ={0, 1,1, 1,1, 2,2, 3,3,   4,4,   5,5,   6,6,    7,7    };
  262. static const uint8_t dc_index_offset[]  ={ 0, 1,2, 3,4, 5,7, 9,13, 17,25, 33,49, 65,97, 129,193};
  263.  
  264. static int x8_get_dc_rlf(IntraX8Context * const w,int const mode, int * const level, int * const final){
  265.     MpegEncContext * const s= w->s;
  266.     int i,e,c;
  267.  
  268.     av_assert2(mode<3);
  269.     if( !w->j_dc_vlc[mode] ) {
  270.         int table_index;
  271.         table_index = get_bits(&s->gb, 3);
  272.         //4 modes, same table
  273.         w->j_dc_vlc[mode]= &j_dc_vlc[w->quant<13][table_index];
  274.     }
  275.  
  276.     i=get_vlc2(&s->gb, w->j_dc_vlc[mode]->table, DC_VLC_BITS, DC_VLC_MTD);
  277.  
  278.     /*(i>=17) {i-=17;final=1;}*/
  279.     c= i>16;
  280.     (*final)=c;
  281.     i-=17*c;
  282.  
  283.     if(i<=0){
  284.         (*level)=0;
  285.         return -i;
  286.     }
  287.     c=(i+1)>>1;//hackish way to calculate dc_extra_sbits[]
  288.     c-=c>1;
  289.  
  290.     e=get_bits(&s->gb,c);//get the extra bits
  291.     i=dc_index_offset[i]+(e>>1);
  292.  
  293.     e= -(e & 1);//0,0xffffff
  294.     (*level)= (i ^ e) - e;// (i^0)-0 , (i^0xff)-(-1)
  295.     return 0;
  296. }
  297. //end of huffman
  298.  
  299. static int x8_setup_spatial_predictor(IntraX8Context * const w, const int chroma){
  300.     MpegEncContext * const s= w->s;
  301.     int range;
  302.     int sum;
  303.     int quant;
  304.  
  305.     w->dsp.setup_spatial_compensation(s->dest[chroma], s->edge_emu_buffer,
  306.                                       s->current_picture.f.linesize[chroma>0],
  307.                                       &range, &sum, w->edges);
  308.     if(chroma){
  309.         w->orient=w->chroma_orient;
  310.         quant=w->quant_dc_chroma;
  311.     }else{
  312.         quant=w->quant;
  313.     }
  314.  
  315.     w->flat_dc=0;
  316.     if(range < quant || range < 3){
  317.         w->orient=0;
  318.         if(range < 3){//yep you read right, a +-1 idct error may break decoding!
  319.             w->flat_dc=1;
  320.             sum+=9;
  321.             w->predicted_dc = (sum*6899)>>17;//((1<<17)+9)/(8+8+1+2)=6899
  322.         }
  323.     }
  324.     if(chroma)
  325.         return 0;
  326.  
  327.     av_assert2(w->orient < 3);
  328.     if(range < 2*w->quant){
  329.         if( (w->edges&3) == 0){
  330.             if(w->orient==1) w->orient=11;
  331.             if(w->orient==2) w->orient=10;
  332.         }else{
  333.             w->orient=0;
  334.         }
  335.         w->raw_orient=0;
  336.     }else{
  337.         static const uint8_t prediction_table[3][12]={
  338.             {0,8,4, 10,11, 2,6,9,1,3,5,7},
  339.             {4,0,8, 11,10, 3,5,2,6,9,1,7},
  340.             {8,0,4, 10,11, 1,7,2,6,9,3,5}
  341.         };
  342.         w->raw_orient=x8_get_orient_vlc(w);
  343.         if(w->raw_orient<0) return -1;
  344.         av_assert2(w->raw_orient < 12 );
  345.         av_assert2(w->orient<3);
  346.         w->orient=prediction_table[w->orient][w->raw_orient];
  347.     }
  348.     return 0;
  349. }
  350.  
  351. static void x8_update_predictions(IntraX8Context * const w, const int orient, const int est_run ){
  352.     MpegEncContext * const s= w->s;
  353.  
  354.     w->prediction_table[s->mb_x*2+(s->mb_y&1)] = (est_run<<2) + 1*(orient==4) + 2*(orient==8);
  355. /*
  356.   y=2n+0 ->//0 2 4
  357.   y=2n+1 ->//1 3 5
  358. */
  359. }
  360. static void x8_get_prediction_chroma(IntraX8Context * const w){
  361.     MpegEncContext * const s= w->s;
  362.  
  363.     w->edges = 1*( !(s->mb_x>>1) );
  364.     w->edges|= 2*( !(s->mb_y>>1) );
  365.     w->edges|= 4*( s->mb_x >= (2*s->mb_width-1) );//mb_x for chroma would always be odd
  366.  
  367.     w->raw_orient=0;
  368.     if(w->edges&3){//lut_co[8]={inv,4,8,8, inv,4,8,8}<- =>{1,1,0,0;1,1,0,0} => 0xCC
  369.         w->chroma_orient=4<<((0xCC>>w->edges)&1);
  370.         return;
  371.     }
  372.     w->chroma_orient = (w->prediction_table[2*s->mb_x-2] & 0x03)<<2;//block[x-1][y|1-1)]
  373. }
  374.  
  375. static void x8_get_prediction(IntraX8Context * const w){
  376.     MpegEncContext * const s= w->s;
  377.     int a,b,c,i;
  378.  
  379.     w->edges = 1*( !s->mb_x );
  380.     w->edges|= 2*( !s->mb_y );
  381.     w->edges|= 4*( s->mb_x >= (2*s->mb_width-1) );
  382.  
  383.     switch(w->edges&3){
  384.         case 0:
  385.             break;
  386.         case 1:
  387.             //take the one from the above block[0][y-1]
  388.             w->est_run = w->prediction_table[!(s->mb_y&1)]>>2;
  389.             w->orient  = 1;
  390.             return;
  391.         case 2:
  392.             //take the one from the previous block[x-1][0]
  393.             w->est_run = w->prediction_table[2*s->mb_x-2]>>2;
  394.             w->orient  = 2;
  395.             return;
  396.         case 3:
  397.             w->est_run = 16;
  398.             w->orient  = 0;
  399.             return;
  400.     }
  401.     //no edge cases
  402.     b= w->prediction_table[2*s->mb_x   + !(s->mb_y&1) ];//block[x  ][y-1]
  403.     a= w->prediction_table[2*s->mb_x-2 +  (s->mb_y&1) ];//block[x-1][y  ]
  404.     c= w->prediction_table[2*s->mb_x-2 + !(s->mb_y&1) ];//block[x-1][y-1]
  405.  
  406.     w->est_run = FFMIN(b,a);
  407.     /* This condition has nothing to do with w->edges, even if it looks
  408.        similar it would trigger if e.g. x=3;y=2;
  409.        I guess somebody wrote something wrong and it became standard. */
  410.     if( (s->mb_x & s->mb_y) != 0 ) w->est_run=FFMIN(c,w->est_run);
  411.     w->est_run>>=2;
  412.  
  413.     a&=3;
  414.     b&=3;
  415.     c&=3;
  416.  
  417.     i=( 0xFFEAF4C4>>(2*b+8*a) )&3;
  418.     if(i!=3) w->orient=i;
  419.     else     w->orient=( 0xFFEAD8>>(2*c+8*(w->quant>12)) )&3;
  420. /*
  421. lut1[b][a]={
  422. ->{0, 1, 0, pad},
  423.   {0, 1, X, pad},
  424.   {2, 2, 2, pad}}
  425.    pad 2   2  2; pad X  1  0; pad 0  1  0 <-
  426. -> 11 10 '10 10 '11 11'01 00 '11 00'01 00=>0xEAF4C4
  427.  
  428. lut2[q>12][c]={
  429.   ->{0,2,1,pad},
  430.     {2,2,2,pad}}
  431.    pad 2  2  2; pad 1  2  0 <-
  432. -> 11 10'10 10 '11 01'10 00=>0xEAD8
  433. */
  434. }
  435.  
  436.  
  437. static void x8_ac_compensation(IntraX8Context * const w, int const direction, int const dc_level){
  438.     MpegEncContext * const s= w->s;
  439.     int t;
  440. #define B(x,y)  s->block[0][w->idct_permutation[(x)+(y)*8]]
  441. #define T(x)  ((x) * dc_level + 0x8000) >> 16;
  442.     switch(direction){
  443.     case 0:
  444.         t = T(3811);//h
  445.         B(1,0) -= t;
  446.         B(0,1) -= t;
  447.  
  448.         t = T(487);//e
  449.         B(2,0) -= t;
  450.         B(0,2) -= t;
  451.  
  452.         t = T(506);//f
  453.         B(3,0) -= t;
  454.         B(0,3) -= t;
  455.  
  456.         t = T(135);//c
  457.         B(4,0) -= t;
  458.         B(0,4) -= t;
  459.         B(2,1) += t;
  460.         B(1,2) += t;
  461.         B(3,1) += t;
  462.         B(1,3) += t;
  463.  
  464.         t = T(173);//d
  465.         B(5,0) -= t;
  466.         B(0,5) -= t;
  467.  
  468.         t = T(61);//b
  469.         B(6,0) -= t;
  470.         B(0,6) -= t;
  471.         B(5,1) += t;
  472.         B(1,5) += t;
  473.  
  474.         t = T(42); //a
  475.         B(7,0) -= t;
  476.         B(0,7) -= t;
  477.         B(4,1) += t;
  478.         B(1,4) += t;
  479.         B(4,4) += t;
  480.  
  481.         t = T(1084);//g
  482.         B(1,1) += t;
  483.  
  484.         s->block_last_index[0] = FFMAX(s->block_last_index[0], 7*8);
  485.         break;
  486.     case 1:
  487.         B(0,1) -= T(6269);
  488.         B(0,3) -= T( 708);
  489.         B(0,5) -= T( 172);
  490.         B(0,7) -= T(  73);
  491.  
  492.         s->block_last_index[0] = FFMAX(s->block_last_index[0], 7*8);
  493.         break;
  494.     case 2:
  495.         B(1,0) -= T(6269);
  496.         B(3,0) -= T( 708);
  497.         B(5,0) -= T( 172);
  498.         B(7,0) -= T(  73);
  499.  
  500.         s->block_last_index[0] = FFMAX(s->block_last_index[0], 7);
  501.         break;
  502.     }
  503. #undef B
  504. #undef T
  505. }
  506.  
  507. static void dsp_x8_put_solidcolor(uint8_t const pix, uint8_t * dst, int const linesize){
  508.     int k;
  509.     for(k=0;k<8;k++){
  510.         memset(dst,pix,8);
  511.         dst+=linesize;
  512.     }
  513. }
  514.  
  515. static const int16_t quant_table[64] = {
  516.     256, 256, 256, 256,  256, 256, 259, 262,
  517.     265, 269, 272, 275,  278, 282, 285, 288,
  518.     292, 295, 299, 303,  306, 310, 314, 317,
  519.     321, 325, 329, 333,  337, 341, 345, 349,
  520.     353, 358, 362, 366,  371, 375, 379, 384,
  521.     389, 393, 398, 403,  408, 413, 417, 422,
  522.     428, 433, 438, 443,  448, 454, 459, 465,
  523.     470, 476, 482, 488,  493, 499, 505, 511
  524. };
  525.  
  526. static int x8_decode_intra_mb(IntraX8Context* const w, const int chroma){
  527.     MpegEncContext * const s= w->s;
  528.  
  529.     uint8_t * scantable;
  530.     int final,run,level;
  531.     int ac_mode,dc_mode,est_run,dc_level;
  532.     int pos,n;
  533.     int zeros_only;
  534.     int use_quant_matrix;
  535.     int sign;
  536.  
  537.     av_assert2(w->orient<12);
  538.     s->dsp.clear_block(s->block[0]);
  539.  
  540.     if(chroma){
  541.         dc_mode=2;
  542.     }else{
  543.         dc_mode=!!w->est_run;//0,1
  544.     }
  545.  
  546.     if(x8_get_dc_rlf(w, dc_mode, &dc_level, &final)) return -1;
  547.     n=0;
  548.     zeros_only=0;
  549.     if(!final){//decode ac
  550.         use_quant_matrix=w->use_quant_matrix;
  551.         if(chroma){
  552.             ac_mode = 1;
  553.             est_run = 64;//not used
  554.         }else{
  555.             if (w->raw_orient < 3){
  556.                 use_quant_matrix = 0;
  557.             }
  558.             if(w->raw_orient > 4){
  559.                 ac_mode = 0;
  560.                 est_run = 64;
  561.             }else{
  562.                 if(w->est_run > 1){
  563.                     ac_mode = 2;
  564.                     est_run=w->est_run;
  565.                 }else{
  566.                     ac_mode = 3;
  567.                     est_run = 64;
  568.                 }
  569.             }
  570.         }
  571.         x8_select_ac_table(w,ac_mode);
  572.         /*scantable_selector[12]={0,2,0,1,1,1,0,2,2,0,1,2};<-
  573.         -> 10'01' 00'10' 10'00' 01'01' 01'00' 10'00 =>0x928548 */
  574.         scantable = w->scantable[ (0x928548>>(2*w->orient))&3 ].permutated;
  575.         pos=0;
  576.         do {
  577.             n++;
  578.             if( n >= est_run ){
  579.                 ac_mode=3;
  580.                 x8_select_ac_table(w,3);
  581.             }
  582.  
  583.             x8_get_ac_rlf(w,ac_mode,&run,&level,&final);
  584.  
  585.             pos+=run+1;
  586.             if(pos>63){
  587.                 //this also handles vlc error in x8_get_ac_rlf
  588.                 return -1;
  589.             }
  590.             level= (level+1) * w->dquant;
  591.             level+= w->qsum;
  592.  
  593.             sign = - get_bits1(&s->gb);
  594.             level = (level ^ sign) - sign;
  595.  
  596.             if(use_quant_matrix){
  597.                 level = (level*quant_table[pos])>>8;
  598.             }
  599.             s->block[0][ scantable[pos] ]=level;
  600.         }while(!final);
  601.  
  602.         s->block_last_index[0]=pos;
  603.     }else{//DC only
  604.         s->block_last_index[0]=0;
  605.         if(w->flat_dc && ((unsigned)(dc_level+1)) < 3){//[-1;1]
  606.             int32_t divide_quant= !chroma ? w->divide_quant_dc_luma:
  607.                                             w->divide_quant_dc_chroma;
  608.             int32_t dc_quant    = !chroma ? w->quant:
  609.                                             w->quant_dc_chroma;
  610.  
  611.             //original intent dc_level+=predicted_dc/quant; but it got lost somewhere in the rounding
  612.             dc_level+= (w->predicted_dc*divide_quant + (1<<12) )>>13;
  613.  
  614.             dsp_x8_put_solidcolor( av_clip_uint8((dc_level*dc_quant+4)>>3),
  615.                                    s->dest[chroma], s->current_picture.f.linesize[!!chroma]);
  616.  
  617.             goto block_placed;
  618.         }
  619.         zeros_only = (dc_level == 0);
  620.     }
  621.     if(!chroma){
  622.         s->block[0][0] = dc_level*w->quant;
  623.     }else{
  624.         s->block[0][0] = dc_level*w->quant_dc_chroma;
  625.     }
  626.  
  627.     //there is !zero_only check in the original, but dc_level check is enough
  628.     if( (unsigned int)(dc_level+1) >= 3 && (w->edges&3) != 3 ){
  629.         int direction;
  630.         /*ac_comp_direction[orient] = { 0, 3, 3, 1, 1, 0, 0, 0, 2, 2, 2, 1 };<-
  631.         -> 01'10' 10'10' 00'00' 00'01' 01'11' 11'00 =>0x6A017C */
  632.         direction= (0x6A017C>>(w->orient*2))&3;
  633.         if (direction != 3){
  634.             x8_ac_compensation(w, direction, s->block[0][0]);//modify block_last[]
  635.         }
  636.     }
  637.  
  638.     if(w->flat_dc){
  639.         dsp_x8_put_solidcolor(w->predicted_dc, s->dest[chroma], s->current_picture.f.linesize[!!chroma]);
  640.     }else{
  641.         w->dsp.spatial_compensation[w->orient]( s->edge_emu_buffer,
  642.                                             s->dest[chroma],
  643.                                             s->current_picture.f.linesize[!!chroma] );
  644.     }
  645.     if(!zeros_only)
  646.         w->wdsp.idct_add (s->dest[chroma],
  647.                           s->current_picture.f.linesize[!!chroma],
  648.                           s->block[0] );
  649.  
  650. block_placed:
  651.  
  652.     if(!chroma){
  653.         x8_update_predictions(w,w->orient,n);
  654.     }
  655.  
  656.     if(s->loop_filter){
  657.         uint8_t* ptr = s->dest[chroma];
  658.         int linesize = s->current_picture.f.linesize[!!chroma];
  659.  
  660.         if(!( (w->edges&2) || ( zeros_only && (w->orient|4)==4 ) )){
  661.             w->dsp.h_loop_filter(ptr, linesize, w->quant);
  662.         }
  663.         if(!( (w->edges&1) || ( zeros_only && (w->orient|8)==8 ) )){
  664.             w->dsp.v_loop_filter(ptr, linesize, w->quant);
  665.         }
  666.     }
  667.     return 0;
  668. }
  669.  
  670. static void x8_init_block_index(MpegEncContext *s){ //FIXME maybe merge with ff_*
  671. //not s->linesize as this would be wrong for field pics
  672. //not that IntraX8 has interlacing support ;)
  673.     const int linesize   = s->current_picture.f.linesize[0];
  674.     const int uvlinesize = s->current_picture.f.linesize[1];
  675.  
  676.     s->dest[0] = s->current_picture.f.data[0];
  677.     s->dest[1] = s->current_picture.f.data[1];
  678.     s->dest[2] = s->current_picture.f.data[2];
  679.  
  680.     s->dest[0] +=   s->mb_y        *   linesize << 3;
  681.     s->dest[1] += ( s->mb_y&(~1) ) * uvlinesize << 2;//chroma blocks are on add rows
  682.     s->dest[2] += ( s->mb_y&(~1) ) * uvlinesize << 2;
  683. }
  684.  
  685. /**
  686.  * Initialize IntraX8 frame decoder.
  687.  * Requires valid MpegEncContext with valid s->mb_width before calling.
  688.  * @param w pointer to IntraX8Context
  689.  * @param s pointer to MpegEncContext of the parent codec
  690.  */
  691. av_cold void ff_intrax8_common_init(IntraX8Context * w, MpegEncContext * const s){
  692.  
  693.     w->s=s;
  694.     x8_vlc_init();
  695.     av_assert0(s->mb_width>0);
  696.     w->prediction_table=av_mallocz(s->mb_width*2*2);//two rows, 2 blocks per cannon mb
  697.  
  698.     ff_wmv2dsp_init(&w->wdsp);
  699.     ff_init_scantable_permutation(w->idct_permutation,
  700.                                   w->wdsp.idct_perm);
  701.  
  702.     ff_init_scantable(w->idct_permutation, &w->scantable[0], ff_wmv1_scantable[0]);
  703.     ff_init_scantable(w->idct_permutation, &w->scantable[1], ff_wmv1_scantable[2]);
  704.     ff_init_scantable(w->idct_permutation, &w->scantable[2], ff_wmv1_scantable[3]);
  705.  
  706.     ff_intrax8dsp_init(&w->dsp);
  707. }
  708.  
  709. /**
  710.  * Destroy IntraX8 frame structure.
  711.  * @param w pointer to IntraX8Context
  712.  */
  713. av_cold void ff_intrax8_common_end(IntraX8Context * w)
  714. {
  715.     av_freep(&w->prediction_table);
  716. }
  717.  
  718. /**
  719.  * Decode single IntraX8 frame.
  720.  * The parent codec must fill s->loopfilter and s->gb (bitstream).
  721.  * The parent codec must call MPV_frame_start(), ff_er_frame_start() before calling this function.
  722.  * The parent codec must call ff_er_frame_end(), MPV_frame_end() after calling this function.
  723.  * This function does not use MPV_decode_mb().
  724.  * lowres decoding is theoretically impossible.
  725.  * @param w pointer to IntraX8Context
  726.  * @param dquant doubled quantizer, it would be odd in case of VC-1 halfpq==1.
  727.  * @param quant_offset offset away from zero
  728.  */
  729. int ff_intrax8_decode_picture(IntraX8Context * const w, int dquant, int quant_offset){
  730.     MpegEncContext * const s= w->s;
  731.     int mb_xy;
  732.     w->use_quant_matrix = get_bits1(&s->gb);
  733.  
  734.     w->dquant = dquant;
  735.     w->quant  = dquant >> 1;
  736.     w->qsum   = quant_offset;
  737.  
  738.     w->divide_quant_dc_luma = ((1<<16) + (w->quant>>1)) / w->quant;
  739.     if(w->quant < 5){
  740.         w->quant_dc_chroma =  w->quant;
  741.         w->divide_quant_dc_chroma = w->divide_quant_dc_luma;
  742.     }else{
  743.         w->quant_dc_chroma =  w->quant+((w->quant+3)>>3);
  744.         w->divide_quant_dc_chroma = ((1<<16) + (w->quant_dc_chroma>>1)) / w->quant_dc_chroma;
  745.     }
  746.     x8_reset_vlc_tables(w);
  747.  
  748.     s->resync_mb_x=0;
  749.     s->resync_mb_y=0;
  750.  
  751.     for(s->mb_y=0; s->mb_y < s->mb_height*2; s->mb_y++){
  752.         x8_init_block_index(s);
  753.         mb_xy=(s->mb_y>>1)*s->mb_stride;
  754.  
  755.         for(s->mb_x=0; s->mb_x < s->mb_width*2; s->mb_x++){
  756.             x8_get_prediction(w);
  757.             if(x8_setup_spatial_predictor(w,0)) goto error;
  758.             if(x8_decode_intra_mb(w,0)) goto error;
  759.  
  760.             if( s->mb_x & s->mb_y & 1 ){
  761.                 x8_get_prediction_chroma(w);
  762.  
  763.                 /*when setting up chroma, no vlc is read,
  764.                 so no error condition can be reached*/
  765.                 x8_setup_spatial_predictor(w,1);
  766.                 if(x8_decode_intra_mb(w,1)) goto error;
  767.  
  768.                 x8_setup_spatial_predictor(w,2);
  769.                 if(x8_decode_intra_mb(w,2)) goto error;
  770.  
  771.                 s->dest[1]+= 8;
  772.                 s->dest[2]+= 8;
  773.  
  774.                 /*emulate MB info in the relevant tables*/
  775.                 s->mbskip_table [mb_xy]=0;
  776.                 s->mbintra_table[mb_xy]=1;
  777.                 s->current_picture.qscale_table[mb_xy] = w->quant;
  778.                 mb_xy++;
  779.             }
  780.             s->dest[0]+= 8;
  781.         }
  782.         if(s->mb_y&1){
  783.             ff_mpeg_draw_horiz_band(s, (s->mb_y-1)*8, 16);
  784.         }
  785.     }
  786.  
  787. error:
  788.     ff_er_add_slice(&s->er, s->resync_mb_x, s->resync_mb_y,
  789.                         (s->mb_x>>1)-1, (s->mb_y>>1)-1,
  790.                         ER_MB_END );
  791.     return 0;
  792. }
  793.