yading@10: /*
yading@10:  * This file is part of FFmpeg.
yading@10:  *
yading@10:  * FFmpeg is free software; you can redistribute it and/or
yading@10:  * modify it under the terms of the GNU Lesser General Public
yading@10:  * License as published by the Free Software Foundation; either
yading@10:  * version 2.1 of the License, or (at your option) any later version.
yading@10:  *
yading@10:  * FFmpeg is distributed in the hope that it will be useful,
yading@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
yading@10:  * Lesser General Public License for more details.
yading@10:  *
yading@10:  * You should have received a copy of the GNU Lesser General Public
yading@10:  * License along with FFmpeg; if not, write to the Free Software
yading@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@10:  */
yading@10: 
yading@10: /**
yading@10:  * @file
yading@10:  * @brief IntraX8 (J-Frame) subdecoder, used by WMV2 and VC-1
yading@10:  */
yading@10: 
yading@10: #include "libavutil/avassert.h"
yading@10: #include "avcodec.h"
yading@10: #include "error_resilience.h"
yading@10: #include "get_bits.h"
yading@10: #include "mpegvideo.h"
yading@10: #include "msmpeg4data.h"
yading@10: #include "intrax8huf.h"
yading@10: #include "intrax8.h"
yading@10: #include "intrax8dsp.h"
yading@10: 
yading@10: #define MAX_TABLE_DEPTH(table_bits, max_bits) ((max_bits+table_bits-1)/table_bits)
yading@10: 
yading@10: #define DC_VLC_BITS 9
yading@10: #define AC_VLC_BITS 9
yading@10: #define OR_VLC_BITS 7
yading@10: 
yading@10: #define DC_VLC_MTD MAX_TABLE_DEPTH(DC_VLC_BITS, MAX_DC_VLC_BITS)
yading@10: #define AC_VLC_MTD MAX_TABLE_DEPTH(AC_VLC_BITS, MAX_AC_VLC_BITS)
yading@10: #define OR_VLC_MTD MAX_TABLE_DEPTH(OR_VLC_BITS, MAX_OR_VLC_BITS)
yading@10: 
yading@10: static VLC j_ac_vlc[2][2][8];  //[quant<13],[intra/inter],[select]
yading@10: static VLC j_dc_vlc[2][8];     //[quant], [select]
yading@10: static VLC j_orient_vlc[2][4]; //[quant], [select]
yading@10: 
yading@10: static av_cold void x8_vlc_init(void){
yading@10:     int i;
yading@10:     int offset = 0;
yading@10:     int sizeidx = 0;
yading@10:     static const uint16_t sizes[8*4 + 8*2 + 2 + 4] = {
yading@10:         576, 548, 582, 618, 546, 616, 560, 642,
yading@10:         584, 582, 704, 664, 512, 544, 656, 640,
yading@10:         512, 648, 582, 566, 532, 614, 596, 648,
yading@10:         586, 552, 584, 590, 544, 578, 584, 624,
yading@10: 
yading@10:         528, 528, 526, 528, 536, 528, 526, 544,
yading@10:         544, 512, 512, 528, 528, 544, 512, 544,
yading@10: 
yading@10:         128, 128, 128, 128, 128, 128};
yading@10: 
yading@10:     static VLC_TYPE table[28150][2];
yading@10: 
yading@10: #define  init_ac_vlc(dst,src) \
yading@10:     dst.table = &table[offset]; \
yading@10:     dst.table_allocated = sizes[sizeidx]; \
yading@10:     offset += sizes[sizeidx++]; \
yading@10:        init_vlc(&dst, \
yading@10:               AC_VLC_BITS,77, \
yading@10:               &src[1],4,2, \
yading@10:               &src[0],4,2, \
yading@10:               INIT_VLC_USE_NEW_STATIC)
yading@10: //set ac tables
yading@10:     for(i=0;i<8;i++){
yading@10:         init_ac_vlc( j_ac_vlc[0][0][i], x8_ac0_highquant_table[i][0] );
yading@10:         init_ac_vlc( j_ac_vlc[0][1][i], x8_ac1_highquant_table[i][0] );
yading@10:         init_ac_vlc( j_ac_vlc[1][0][i], x8_ac0_lowquant_table [i][0] );
yading@10:         init_ac_vlc( j_ac_vlc[1][1][i], x8_ac1_lowquant_table [i][0] );
yading@10:     }
yading@10: #undef init_ac_vlc
yading@10: 
yading@10: //set dc tables
yading@10: #define init_dc_vlc(dst,src) \
yading@10:     dst.table = &table[offset]; \
yading@10:     dst.table_allocated = sizes[sizeidx]; \
yading@10:     offset += sizes[sizeidx++]; \
yading@10:         init_vlc(&dst, \
yading@10:         DC_VLC_BITS,34, \
yading@10:         &src[1],4,2, \
yading@10:         &src[0],4,2, \
yading@10:         INIT_VLC_USE_NEW_STATIC);
yading@10:     for(i=0;i<8;i++){
yading@10:         init_dc_vlc( j_dc_vlc[0][i], x8_dc_highquant_table[i][0]);
yading@10:         init_dc_vlc( j_dc_vlc[1][i], x8_dc_lowquant_table [i][0]);
yading@10:     }
yading@10: #undef init_dc_vlc
yading@10: 
yading@10: //set orient tables
yading@10: #define init_or_vlc(dst,src) \
yading@10:     dst.table = &table[offset]; \
yading@10:     dst.table_allocated = sizes[sizeidx]; \
yading@10:     offset += sizes[sizeidx++]; \
yading@10:     init_vlc(&dst, \
yading@10:     OR_VLC_BITS,12, \
yading@10:     &src[1],4,2, \
yading@10:     &src[0],4,2, \
yading@10:     INIT_VLC_USE_NEW_STATIC);
yading@10:     for(i=0;i<2;i++){
yading@10:         init_or_vlc( j_orient_vlc[0][i], x8_orient_highquant_table[i][0]);
yading@10:     }
yading@10:     for(i=0;i<4;i++){
yading@10:         init_or_vlc( j_orient_vlc[1][i], x8_orient_lowquant_table [i][0])
yading@10:     }
yading@10:     if (offset != sizeof(table)/sizeof(VLC_TYPE)/2)
yading@10:         av_log(NULL, AV_LOG_ERROR, "table size %i does not match needed %i\n", (int)(sizeof(table)/sizeof(VLC_TYPE)/2), offset);
yading@10: }
yading@10: #undef init_or_vlc
yading@10: 
yading@10: static void x8_reset_vlc_tables(IntraX8Context * w){
yading@10:     memset(w->j_dc_vlc,0,sizeof(w->j_dc_vlc));
yading@10:     memset(w->j_ac_vlc,0,sizeof(w->j_ac_vlc));
yading@10:     w->j_orient_vlc=NULL;
yading@10: }
yading@10: 
yading@10: static inline void x8_select_ac_table(IntraX8Context * const w , int mode){
yading@10:     MpegEncContext * const s= w->s;
yading@10:     int table_index;
yading@10: 
yading@10:     av_assert2(mode<4);
yading@10: 
yading@10:     if( w->j_ac_vlc[mode] ) return;
yading@10: 
yading@10:     table_index = get_bits(&s->gb, 3);
yading@10:     w->j_ac_vlc[mode] = &j_ac_vlc[w->quant<13][mode>>1][table_index];//2 modes use same tables
yading@10:     av_assert2(w->j_ac_vlc[mode]);
yading@10: }
yading@10: 
yading@10: static inline int x8_get_orient_vlc(IntraX8Context * w){
yading@10:     MpegEncContext * const s= w->s;
yading@10:     int table_index;
yading@10: 
yading@10:     if(!w->j_orient_vlc ){
yading@10:         table_index = get_bits(&s->gb, 1+(w->quant<13) );
yading@10:         w->j_orient_vlc = &j_orient_vlc[w->quant<13][table_index];
yading@10:     }
yading@10: 
yading@10:     return get_vlc2(&s->gb, w->j_orient_vlc->table, OR_VLC_BITS, OR_VLC_MTD);
yading@10: }
yading@10: 
yading@10: #define extra_bits(eb) (eb)
yading@10: #define extra_run   (0xFF<<8)
yading@10: #define extra_level (0x00<<8)
yading@10: #define   run_offset(r)    ((r)<<16)
yading@10: #define level_offset(l)    ((l)<<24)
yading@10: static const uint32_t ac_decode_table[]={
yading@10:     /*46*/ extra_bits(3) |  extra_run  | run_offset(16) | level_offset( 0),
yading@10:     /*47*/ extra_bits(3) |  extra_run  | run_offset(24) | level_offset( 0),
yading@10:     /*48*/ extra_bits(2) |  extra_run  | run_offset( 4) | level_offset( 1),
yading@10:     /*49*/ extra_bits(3) |  extra_run  | run_offset( 8) | level_offset( 1),
yading@10: 
yading@10:     /*50*/ extra_bits(5) |  extra_run  | run_offset(32) | level_offset( 0),
yading@10:     /*51*/ extra_bits(4) |  extra_run  | run_offset(16) | level_offset( 1),
yading@10: 
yading@10:     /*52*/ extra_bits(2) | extra_level | run_offset( 0) | level_offset( 4),
yading@10:     /*53*/ extra_bits(2) | extra_level | run_offset( 0) | level_offset( 8),
yading@10:     /*54*/ extra_bits(2) | extra_level | run_offset( 0) | level_offset(12),
yading@10:     /*55*/ extra_bits(3) | extra_level | run_offset( 0) | level_offset(16),
yading@10:     /*56*/ extra_bits(3) | extra_level | run_offset( 0) | level_offset(24),
yading@10: 
yading@10:     /*57*/ extra_bits(2) | extra_level | run_offset( 1) | level_offset( 3),
yading@10:     /*58*/ extra_bits(3) | extra_level | run_offset( 1) | level_offset( 7),
yading@10: 
yading@10:     /*59*/ extra_bits(2) |  extra_run  | run_offset(16) | level_offset( 0),
yading@10:     /*60*/ extra_bits(2) |  extra_run  | run_offset(20) | level_offset( 0),
yading@10:     /*61*/ extra_bits(2) |  extra_run  | run_offset(24) | level_offset( 0),
yading@10:     /*62*/ extra_bits(2) |  extra_run  | run_offset(28) | level_offset( 0),
yading@10:     /*63*/ extra_bits(4) |  extra_run  | run_offset(32) | level_offset( 0),
yading@10:     /*64*/ extra_bits(4) |  extra_run  | run_offset(48) | level_offset( 0),
yading@10: 
yading@10:     /*65*/ extra_bits(2) |  extra_run  | run_offset( 4) | level_offset( 1),
yading@10:     /*66*/ extra_bits(3) |  extra_run  | run_offset( 8) | level_offset( 1),
yading@10:     /*67*/ extra_bits(4) |  extra_run  | run_offset(16) | level_offset( 1),
yading@10: 
yading@10:     /*68*/ extra_bits(2) | extra_level | run_offset( 0) | level_offset( 4),
yading@10:     /*69*/ extra_bits(3) | extra_level | run_offset( 0) | level_offset( 8),
yading@10:     /*70*/ extra_bits(4) | extra_level | run_offset( 0) | level_offset(16),
yading@10: 
yading@10:     /*71*/ extra_bits(2) | extra_level | run_offset( 1) | level_offset( 3),
yading@10:     /*72*/ extra_bits(3) | extra_level | run_offset( 1) | level_offset( 7),
yading@10: };
yading@10: //extra_bits = 3bits; extra_run/level = 1 bit; run_offset = 6bits; level_offset = 5 bits;
yading@10: #undef extra_bits
yading@10: #undef extra_run
yading@10: #undef extra_level
yading@10: #undef run_offset
yading@10: #undef level_offset
yading@10: 
yading@10: static void x8_get_ac_rlf(IntraX8Context * const w, const int mode,
yading@10:                      int * const run, int * const level, int * const final){
yading@10:     MpegEncContext *  const s= w->s;
yading@10:     int i,e;
yading@10: 
yading@10: //    x8_select_ac_table(w,mode);
yading@10:     i = get_vlc2(&s->gb, w->j_ac_vlc[mode]->table, AC_VLC_BITS, AC_VLC_MTD);
yading@10: 
yading@10:     if(i<46){ //[0-45]
yading@10:         int t,l;
yading@10:         if(i<0){
yading@10:             (*level)=(*final)=//prevent 'may be used unilitialized'
yading@10:             (*run)=64;//this would cause error exit in the ac loop
yading@10:             return;
yading@10:         }
yading@10: 
yading@10:         (*final) = t = (i>22);
yading@10:         i-=23*t;
yading@10: /*
yading@10:   i== 0-15 r=0-15 l=0 ;r=i& %01111
yading@10:   i==16-19 r=0-3  l=1 ;r=i& %00011
yading@10:   i==20-21 r=0-1  l=2 ;r=i& %00001
yading@10:   i==22    r=0    l=3 ;r=i& %00000
yading@10: 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
yading@10: t=lut_mask[l]={0x0f,0x03,0x01,0x00}[l]; as i<256 the higher bits do not matter */
yading@10:         l=(0xE50000>>(i&(0x1E)))&3;/*0x1E or (~1) or ((i>>1)<<1)*/
yading@10:         t=(0x01030F>>(l<<3));
yading@10: 
yading@10:         (*run)   = i&t;
yading@10:         (*level) = l;
yading@10:     }else if(i<73){//[46-72]
yading@10:         uint32_t sm;
yading@10:         uint32_t mask;
yading@10: 
yading@10:         i-=46;
yading@10:         sm=ac_decode_table[i];
yading@10: 
yading@10:         e=get_bits(&s->gb,sm&0xF);sm>>=8;//3bits
yading@10:         mask=sm&0xff;sm>>=8;             //1bit
yading@10: 
yading@10:         (*run)  =(sm&0xff) + (e&( mask));//6bits
yading@10:         (*level)=(sm>>8)   + (e&(~mask));//5bits
yading@10:         (*final)=i>(58-46);
yading@10:     }else if(i<75){//[73-74]
yading@10:         static const uint8_t crazy_mix_runlevel[32]={
yading@10:         0x22,0x32,0x33,0x53,0x23,0x42,0x43,0x63,
yading@10:         0x24,0x52,0x34,0x73,0x25,0x62,0x44,0x83,
yading@10:         0x26,0x72,0x35,0x54,0x27,0x82,0x45,0x64,
yading@10:         0x28,0x92,0x36,0x74,0x29,0xa2,0x46,0x84};
yading@10: 
yading@10:         (*final)=!(i&1);
yading@10:         e=get_bits(&s->gb,5);//get the extra bits
yading@10:         (*run)  =crazy_mix_runlevel[e]>>4;
yading@10:         (*level)=crazy_mix_runlevel[e]&0x0F;
yading@10:     }else{
yading@10:         (*level)=get_bits( &s->gb, 7-3*(i&1));
yading@10:         (*run)  =get_bits( &s->gb, 6);
yading@10:         (*final)=get_bits1(&s->gb);
yading@10:     }
yading@10:     return;
yading@10: }
yading@10: 
yading@10: //static const uint8_t dc_extra_sbits[]   ={0, 1,1, 1,1, 2,2, 3,3,   4,4,   5,5,   6,6,    7,7    };
yading@10: static const uint8_t dc_index_offset[]  ={ 0, 1,2, 3,4, 5,7, 9,13, 17,25, 33,49, 65,97, 129,193};
yading@10: 
yading@10: static int x8_get_dc_rlf(IntraX8Context * const w,int const mode, int * const level, int * const final){
yading@10:     MpegEncContext * const s= w->s;
yading@10:     int i,e,c;
yading@10: 
yading@10:     av_assert2(mode<3);
yading@10:     if( !w->j_dc_vlc[mode] ) {
yading@10:         int table_index;
yading@10:         table_index = get_bits(&s->gb, 3);
yading@10:         //4 modes, same table
yading@10:         w->j_dc_vlc[mode]= &j_dc_vlc[w->quant<13][table_index];
yading@10:     }
yading@10: 
yading@10:     i=get_vlc2(&s->gb, w->j_dc_vlc[mode]->table, DC_VLC_BITS, DC_VLC_MTD);
yading@10: 
yading@10:     /*(i>=17) {i-=17;final=1;}*/
yading@10:     c= i>16;
yading@10:     (*final)=c;
yading@10:     i-=17*c;
yading@10: 
yading@10:     if(i<=0){
yading@10:         (*level)=0;
yading@10:         return -i;
yading@10:     }
yading@10:     c=(i+1)>>1;//hackish way to calculate dc_extra_sbits[]
yading@10:     c-=c>1;
yading@10: 
yading@10:     e=get_bits(&s->gb,c);//get the extra bits
yading@10:     i=dc_index_offset[i]+(e>>1);
yading@10: 
yading@10:     e= -(e & 1);//0,0xffffff
yading@10:     (*level)= (i ^ e) - e;// (i^0)-0 , (i^0xff)-(-1)
yading@10:     return 0;
yading@10: }
yading@10: //end of huffman
yading@10: 
yading@10: static int x8_setup_spatial_predictor(IntraX8Context * const w, const int chroma){
yading@10:     MpegEncContext * const s= w->s;
yading@10:     int range;
yading@10:     int sum;
yading@10:     int quant;
yading@10: 
yading@10:     w->dsp.setup_spatial_compensation(s->dest[chroma], s->edge_emu_buffer,
yading@10:                                       s->current_picture.f.linesize[chroma>0],
yading@10:                                       &range, &sum, w->edges);
yading@10:     if(chroma){
yading@10:         w->orient=w->chroma_orient;
yading@10:         quant=w->quant_dc_chroma;
yading@10:     }else{
yading@10:         quant=w->quant;
yading@10:     }
yading@10: 
yading@10:     w->flat_dc=0;
yading@10:     if(range < quant || range < 3){
yading@10:         w->orient=0;
yading@10:         if(range < 3){//yep you read right, a +-1 idct error may break decoding!
yading@10:             w->flat_dc=1;
yading@10:             sum+=9;
yading@10:             w->predicted_dc = (sum*6899)>>17;//((1<<17)+9)/(8+8+1+2)=6899
yading@10:         }
yading@10:     }
yading@10:     if(chroma)
yading@10:         return 0;
yading@10: 
yading@10:     av_assert2(w->orient < 3);
yading@10:     if(range < 2*w->quant){
yading@10:         if( (w->edges&3) == 0){
yading@10:             if(w->orient==1) w->orient=11;
yading@10:             if(w->orient==2) w->orient=10;
yading@10:         }else{
yading@10:             w->orient=0;
yading@10:         }
yading@10:         w->raw_orient=0;
yading@10:     }else{
yading@10:         static const uint8_t prediction_table[3][12]={
yading@10:             {0,8,4, 10,11, 2,6,9,1,3,5,7},
yading@10:             {4,0,8, 11,10, 3,5,2,6,9,1,7},
yading@10:             {8,0,4, 10,11, 1,7,2,6,9,3,5}
yading@10:         };
yading@10:         w->raw_orient=x8_get_orient_vlc(w);
yading@10:         if(w->raw_orient<0) return -1;
yading@10:         av_assert2(w->raw_orient < 12 );
yading@10:         av_assert2(w->orient<3);
yading@10:         w->orient=prediction_table[w->orient][w->raw_orient];
yading@10:     }
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static void x8_update_predictions(IntraX8Context * const w, const int orient, const int est_run ){
yading@10:     MpegEncContext * const s= w->s;
yading@10: 
yading@10:     w->prediction_table[s->mb_x*2+(s->mb_y&1)] = (est_run<<2) + 1*(orient==4) + 2*(orient==8);
yading@10: /*
yading@10:   y=2n+0 ->//0 2 4
yading@10:   y=2n+1 ->//1 3 5
yading@10: */
yading@10: }
yading@10: static void x8_get_prediction_chroma(IntraX8Context * const w){
yading@10:     MpegEncContext * const s= w->s;
yading@10: 
yading@10:     w->edges = 1*( !(s->mb_x>>1) );
yading@10:     w->edges|= 2*( !(s->mb_y>>1) );
yading@10:     w->edges|= 4*( s->mb_x >= (2*s->mb_width-1) );//mb_x for chroma would always be odd
yading@10: 
yading@10:     w->raw_orient=0;
yading@10:     if(w->edges&3){//lut_co[8]={inv,4,8,8, inv,4,8,8}<- =>{1,1,0,0;1,1,0,0} => 0xCC
yading@10:         w->chroma_orient=4<<((0xCC>>w->edges)&1);
yading@10:         return;
yading@10:     }
yading@10:     w->chroma_orient = (w->prediction_table[2*s->mb_x-2] & 0x03)<<2;//block[x-1][y|1-1)]
yading@10: }
yading@10: 
yading@10: static void x8_get_prediction(IntraX8Context * const w){
yading@10:     MpegEncContext * const s= w->s;
yading@10:     int a,b,c,i;
yading@10: 
yading@10:     w->edges = 1*( !s->mb_x );
yading@10:     w->edges|= 2*( !s->mb_y );
yading@10:     w->edges|= 4*( s->mb_x >= (2*s->mb_width-1) );
yading@10: 
yading@10:     switch(w->edges&3){
yading@10:         case 0:
yading@10:             break;
yading@10:         case 1:
yading@10:             //take the one from the above block[0][y-1]
yading@10:             w->est_run = w->prediction_table[!(s->mb_y&1)]>>2;
yading@10:             w->orient  = 1;
yading@10:             return;
yading@10:         case 2:
yading@10:             //take the one from the previous block[x-1][0]
yading@10:             w->est_run = w->prediction_table[2*s->mb_x-2]>>2;
yading@10:             w->orient  = 2;
yading@10:             return;
yading@10:         case 3:
yading@10:             w->est_run = 16;
yading@10:             w->orient  = 0;
yading@10:             return;
yading@10:     }
yading@10:     //no edge cases
yading@10:     b= w->prediction_table[2*s->mb_x   + !(s->mb_y&1) ];//block[x  ][y-1]
yading@10:     a= w->prediction_table[2*s->mb_x-2 +  (s->mb_y&1) ];//block[x-1][y  ]
yading@10:     c= w->prediction_table[2*s->mb_x-2 + !(s->mb_y&1) ];//block[x-1][y-1]
yading@10: 
yading@10:     w->est_run = FFMIN(b,a);
yading@10:     /* This condition has nothing to do with w->edges, even if it looks
yading@10:        similar it would trigger if e.g. x=3;y=2;
yading@10:        I guess somebody wrote something wrong and it became standard. */
yading@10:     if( (s->mb_x & s->mb_y) != 0 ) w->est_run=FFMIN(c,w->est_run);
yading@10:     w->est_run>>=2;
yading@10: 
yading@10:     a&=3;
yading@10:     b&=3;
yading@10:     c&=3;
yading@10: 
yading@10:     i=( 0xFFEAF4C4>>(2*b+8*a) )&3;
yading@10:     if(i!=3) w->orient=i;
yading@10:     else     w->orient=( 0xFFEAD8>>(2*c+8*(w->quant>12)) )&3;
yading@10: /*
yading@10: lut1[b][a]={
yading@10: ->{0, 1, 0, pad},
yading@10:   {0, 1, X, pad},
yading@10:   {2, 2, 2, pad}}
yading@10:    pad 2   2  2; pad X  1  0; pad 0  1  0 <-
yading@10: -> 11 10 '10 10 '11 11'01 00 '11 00'01 00=>0xEAF4C4
yading@10: 
yading@10: lut2[q>12][c]={
yading@10:   ->{0,2,1,pad},
yading@10:     {2,2,2,pad}}
yading@10:    pad 2  2  2; pad 1  2  0 <-
yading@10: -> 11 10'10 10 '11 01'10 00=>0xEAD8
yading@10: */
yading@10: }
yading@10: 
yading@10: 
yading@10: static void x8_ac_compensation(IntraX8Context * const w, int const direction, int const dc_level){
yading@10:     MpegEncContext * const s= w->s;
yading@10:     int t;
yading@10: #define B(x,y)  s->block[0][s->dsp.idct_permutation[(x)+(y)*8]]
yading@10: #define T(x)  ((x) * dc_level + 0x8000) >> 16;
yading@10:     switch(direction){
yading@10:     case 0:
yading@10:         t = T(3811);//h
yading@10:         B(1,0) -= t;
yading@10:         B(0,1) -= t;
yading@10: 
yading@10:         t = T(487);//e
yading@10:         B(2,0) -= t;
yading@10:         B(0,2) -= t;
yading@10: 
yading@10:         t = T(506);//f
yading@10:         B(3,0) -= t;
yading@10:         B(0,3) -= t;
yading@10: 
yading@10:         t = T(135);//c
yading@10:         B(4,0) -= t;
yading@10:         B(0,4) -= t;
yading@10:         B(2,1) += t;
yading@10:         B(1,2) += t;
yading@10:         B(3,1) += t;
yading@10:         B(1,3) += t;
yading@10: 
yading@10:         t = T(173);//d
yading@10:         B(5,0) -= t;
yading@10:         B(0,5) -= t;
yading@10: 
yading@10:         t = T(61);//b
yading@10:         B(6,0) -= t;
yading@10:         B(0,6) -= t;
yading@10:         B(5,1) += t;
yading@10:         B(1,5) += t;
yading@10: 
yading@10:         t = T(42); //a
yading@10:         B(7,0) -= t;
yading@10:         B(0,7) -= t;
yading@10:         B(4,1) += t;
yading@10:         B(1,4) += t;
yading@10:         B(4,4) += t;
yading@10: 
yading@10:         t = T(1084);//g
yading@10:         B(1,1) += t;
yading@10: 
yading@10:         s->block_last_index[0] = FFMAX(s->block_last_index[0], 7*8);
yading@10:         break;
yading@10:     case 1:
yading@10:         B(0,1) -= T(6269);
yading@10:         B(0,3) -= T( 708);
yading@10:         B(0,5) -= T( 172);
yading@10:         B(0,7) -= T(  73);
yading@10: 
yading@10:         s->block_last_index[0] = FFMAX(s->block_last_index[0], 7*8);
yading@10:         break;
yading@10:     case 2:
yading@10:         B(1,0) -= T(6269);
yading@10:         B(3,0) -= T( 708);
yading@10:         B(5,0) -= T( 172);
yading@10:         B(7,0) -= T(  73);
yading@10: 
yading@10:         s->block_last_index[0] = FFMAX(s->block_last_index[0], 7);
yading@10:         break;
yading@10:     }
yading@10: #undef B
yading@10: #undef T
yading@10: }
yading@10: 
yading@10: static void dsp_x8_put_solidcolor(uint8_t const pix, uint8_t * dst, int const linesize){
yading@10:     int k;
yading@10:     for(k=0;k<8;k++){
yading@10:         memset(dst,pix,8);
yading@10:         dst+=linesize;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static const int16_t quant_table[64] = {
yading@10:     256, 256, 256, 256,  256, 256, 259, 262,
yading@10:     265, 269, 272, 275,  278, 282, 285, 288,
yading@10:     292, 295, 299, 303,  306, 310, 314, 317,
yading@10:     321, 325, 329, 333,  337, 341, 345, 349,
yading@10:     353, 358, 362, 366,  371, 375, 379, 384,
yading@10:     389, 393, 398, 403,  408, 413, 417, 422,
yading@10:     428, 433, 438, 443,  448, 454, 459, 465,
yading@10:     470, 476, 482, 488,  493, 499, 505, 511
yading@10: };
yading@10: 
yading@10: static int x8_decode_intra_mb(IntraX8Context* const w, const int chroma){
yading@10:     MpegEncContext * const s= w->s;
yading@10: 
yading@10:     uint8_t * scantable;
yading@10:     int final,run,level;
yading@10:     int ac_mode,dc_mode,est_run,dc_level;
yading@10:     int pos,n;
yading@10:     int zeros_only;
yading@10:     int use_quant_matrix;
yading@10:     int sign;
yading@10: 
yading@10:     av_assert2(w->orient<12);
yading@10:     s->dsp.clear_block(s->block[0]);
yading@10: 
yading@10:     if(chroma){
yading@10:         dc_mode=2;
yading@10:     }else{
yading@10:         dc_mode=!!w->est_run;//0,1
yading@10:     }
yading@10: 
yading@10:     if(x8_get_dc_rlf(w, dc_mode, &dc_level, &final)) return -1;
yading@10:     n=0;
yading@10:     zeros_only=0;
yading@10:     if(!final){//decode ac
yading@10:         use_quant_matrix=w->use_quant_matrix;
yading@10:         if(chroma){
yading@10:             ac_mode = 1;
yading@10:             est_run = 64;//not used
yading@10:         }else{
yading@10:             if (w->raw_orient < 3){
yading@10:                 use_quant_matrix = 0;
yading@10:             }
yading@10:             if(w->raw_orient > 4){
yading@10:                 ac_mode = 0;
yading@10:                 est_run = 64;
yading@10:             }else{
yading@10:                 if(w->est_run > 1){
yading@10:                     ac_mode = 2;
yading@10:                     est_run=w->est_run;
yading@10:                 }else{
yading@10:                     ac_mode = 3;
yading@10:                     est_run = 64;
yading@10:                 }
yading@10:             }
yading@10:         }
yading@10:         x8_select_ac_table(w,ac_mode);
yading@10:         /*scantable_selector[12]={0,2,0,1,1,1,0,2,2,0,1,2};<-
yading@10:         -> 10'01' 00'10' 10'00' 01'01' 01'00' 10'00 =>0x928548 */
yading@10:         scantable = w->scantable[ (0x928548>>(2*w->orient))&3 ].permutated;
yading@10:         pos=0;
yading@10:         do {
yading@10:             n++;
yading@10:             if( n >= est_run ){
yading@10:                 ac_mode=3;
yading@10:                 x8_select_ac_table(w,3);
yading@10:             }
yading@10: 
yading@10:             x8_get_ac_rlf(w,ac_mode,&run,&level,&final);
yading@10: 
yading@10:             pos+=run+1;
yading@10:             if(pos>63){
yading@10:                 //this also handles vlc error in x8_get_ac_rlf
yading@10:                 return -1;
yading@10:             }
yading@10:             level= (level+1) * w->dquant;
yading@10:             level+= w->qsum;
yading@10: 
yading@10:             sign = - get_bits1(&s->gb);
yading@10:             level = (level ^ sign) - sign;
yading@10: 
yading@10:             if(use_quant_matrix){
yading@10:                 level = (level*quant_table[pos])>>8;
yading@10:             }
yading@10:             s->block[0][ scantable[pos] ]=level;
yading@10:         }while(!final);
yading@10: 
yading@10:         s->block_last_index[0]=pos;
yading@10:     }else{//DC only
yading@10:         s->block_last_index[0]=0;
yading@10:         if(w->flat_dc && ((unsigned)(dc_level+1)) < 3){//[-1;1]
yading@10:             int32_t divide_quant= !chroma ? w->divide_quant_dc_luma:
yading@10:                                             w->divide_quant_dc_chroma;
yading@10:             int32_t dc_quant    = !chroma ? w->quant:
yading@10:                                             w->quant_dc_chroma;
yading@10: 
yading@10:             //original intent dc_level+=predicted_dc/quant; but it got lost somewhere in the rounding
yading@10:             dc_level+= (w->predicted_dc*divide_quant + (1<<12) )>>13;
yading@10: 
yading@10:             dsp_x8_put_solidcolor( av_clip_uint8((dc_level*dc_quant+4)>>3),
yading@10:                                    s->dest[chroma], s->current_picture.f.linesize[!!chroma]);
yading@10: 
yading@10:             goto block_placed;
yading@10:         }
yading@10:         zeros_only = (dc_level == 0);
yading@10:     }
yading@10:     if(!chroma){
yading@10:         s->block[0][0] = dc_level*w->quant;
yading@10:     }else{
yading@10:         s->block[0][0] = dc_level*w->quant_dc_chroma;
yading@10:     }
yading@10: 
yading@10:     //there is !zero_only check in the original, but dc_level check is enough
yading@10:     if( (unsigned int)(dc_level+1) >= 3 && (w->edges&3) != 3 ){
yading@10:         int direction;
yading@10:         /*ac_comp_direction[orient] = { 0, 3, 3, 1, 1, 0, 0, 0, 2, 2, 2, 1 };<-
yading@10:         -> 01'10' 10'10' 00'00' 00'01' 01'11' 11'00 =>0x6A017C */
yading@10:         direction= (0x6A017C>>(w->orient*2))&3;
yading@10:         if (direction != 3){
yading@10:             x8_ac_compensation(w, direction, s->block[0][0]);//modify block_last[]
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     if(w->flat_dc){
yading@10:         dsp_x8_put_solidcolor(w->predicted_dc, s->dest[chroma], s->current_picture.f.linesize[!!chroma]);
yading@10:     }else{
yading@10:         w->dsp.spatial_compensation[w->orient]( s->edge_emu_buffer,
yading@10:                                             s->dest[chroma],
yading@10:                                             s->current_picture.f.linesize[!!chroma] );
yading@10:     }
yading@10:     if(!zeros_only)
yading@10:         s->dsp.idct_add ( s->dest[chroma],
yading@10:                           s->current_picture.f.linesize[!!chroma],
yading@10:                           s->block[0] );
yading@10: 
yading@10: block_placed:
yading@10: 
yading@10:     if(!chroma){
yading@10:         x8_update_predictions(w,w->orient,n);
yading@10:     }
yading@10: 
yading@10:     if(s->loop_filter){
yading@10:         uint8_t* ptr = s->dest[chroma];
yading@10:         int linesize = s->current_picture.f.linesize[!!chroma];
yading@10: 
yading@10:         if(!( (w->edges&2) || ( zeros_only && (w->orient|4)==4 ) )){
yading@10:             w->dsp.h_loop_filter(ptr, linesize, w->quant);
yading@10:         }
yading@10:         if(!( (w->edges&1) || ( zeros_only && (w->orient|8)==8 ) )){
yading@10:             w->dsp.v_loop_filter(ptr, linesize, w->quant);
yading@10:         }
yading@10:     }
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static void x8_init_block_index(MpegEncContext *s){ //FIXME maybe merge with ff_*
yading@10: //not s->linesize as this would be wrong for field pics
yading@10: //not that IntraX8 has interlacing support ;)
yading@10:     const int linesize   = s->current_picture.f.linesize[0];
yading@10:     const int uvlinesize = s->current_picture.f.linesize[1];
yading@10: 
yading@10:     s->dest[0] = s->current_picture.f.data[0];
yading@10:     s->dest[1] = s->current_picture.f.data[1];
yading@10:     s->dest[2] = s->current_picture.f.data[2];
yading@10: 
yading@10:     s->dest[0] +=   s->mb_y        *   linesize << 3;
yading@10:     s->dest[1] += ( s->mb_y&(~1) ) * uvlinesize << 2;//chroma blocks are on add rows
yading@10:     s->dest[2] += ( s->mb_y&(~1) ) * uvlinesize << 2;
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * Initialize IntraX8 frame decoder.
yading@10:  * Requires valid MpegEncContext with valid s->mb_width before calling.
yading@10:  * @param w pointer to IntraX8Context
yading@10:  * @param s pointer to MpegEncContext of the parent codec
yading@10:  */
yading@10: av_cold void ff_intrax8_common_init(IntraX8Context * w, MpegEncContext * const s){
yading@10: 
yading@10:     w->s=s;
yading@10:     x8_vlc_init();
yading@10:     av_assert0(s->mb_width>0);
yading@10:     w->prediction_table=av_mallocz(s->mb_width*2*2);//two rows, 2 blocks per cannon mb
yading@10: 
yading@10:     ff_init_scantable(s->dsp.idct_permutation, &w->scantable[0], ff_wmv1_scantable[0]);
yading@10:     ff_init_scantable(s->dsp.idct_permutation, &w->scantable[1], ff_wmv1_scantable[2]);
yading@10:     ff_init_scantable(s->dsp.idct_permutation, &w->scantable[2], ff_wmv1_scantable[3]);
yading@10: 
yading@10:     ff_intrax8dsp_init(&w->dsp);
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * Destroy IntraX8 frame structure.
yading@10:  * @param w pointer to IntraX8Context
yading@10:  */
yading@10: av_cold void ff_intrax8_common_end(IntraX8Context * w)
yading@10: {
yading@10:     av_freep(&w->prediction_table);
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * Decode single IntraX8 frame.
yading@10:  * The parent codec must fill s->loopfilter and s->gb (bitstream).
yading@10:  * The parent codec must call MPV_frame_start(), ff_er_frame_start() before calling this function.
yading@10:  * The parent codec must call ff_er_frame_end(), MPV_frame_end() after calling this function.
yading@10:  * This function does not use MPV_decode_mb().
yading@10:  * lowres decoding is theoretically impossible.
yading@10:  * @param w pointer to IntraX8Context
yading@10:  * @param dquant doubled quantizer, it would be odd in case of VC-1 halfpq==1.
yading@10:  * @param quant_offset offset away from zero
yading@10:  */
yading@10: //FIXME extern uint8_t ff_wmv3_dc_scale_table[32];
yading@10: int ff_intrax8_decode_picture(IntraX8Context * const w, int dquant, int quant_offset){
yading@10:     MpegEncContext * const s= w->s;
yading@10:     int mb_xy;
yading@10:     w->use_quant_matrix = get_bits1(&s->gb);
yading@10: 
yading@10:     w->dquant = dquant;
yading@10:     w->quant  = dquant >> 1;
yading@10:     w->qsum   = quant_offset;
yading@10: 
yading@10:     w->divide_quant_dc_luma = ((1<<16) + (w->quant>>1)) / w->quant;
yading@10:     if(w->quant < 5){
yading@10:         w->quant_dc_chroma =  w->quant;
yading@10:         w->divide_quant_dc_chroma = w->divide_quant_dc_luma;
yading@10:     }else{
yading@10:         w->quant_dc_chroma =  w->quant+((w->quant+3)>>3);
yading@10:         w->divide_quant_dc_chroma = ((1<<16) + (w->quant_dc_chroma>>1)) / w->quant_dc_chroma;
yading@10:     }
yading@10:     x8_reset_vlc_tables(w);
yading@10: 
yading@10:     s->resync_mb_x=0;
yading@10:     s->resync_mb_y=0;
yading@10: 
yading@10:     for(s->mb_y=0; s->mb_y < s->mb_height*2; s->mb_y++){
yading@10:         x8_init_block_index(s);
yading@10:         mb_xy=(s->mb_y>>1)*s->mb_stride;
yading@10: 
yading@10:         for(s->mb_x=0; s->mb_x < s->mb_width*2; s->mb_x++){
yading@10:             x8_get_prediction(w);
yading@10:             if(x8_setup_spatial_predictor(w,0)) goto error;
yading@10:             if(x8_decode_intra_mb(w,0)) goto error;
yading@10: 
yading@10:             if( s->mb_x & s->mb_y & 1 ){
yading@10:                 x8_get_prediction_chroma(w);
yading@10: 
yading@10:                 /*when setting up chroma, no vlc is read,
yading@10:                 so no error condition can be reached*/
yading@10:                 x8_setup_spatial_predictor(w,1);
yading@10:                 if(x8_decode_intra_mb(w,1)) goto error;
yading@10: 
yading@10:                 x8_setup_spatial_predictor(w,2);
yading@10:                 if(x8_decode_intra_mb(w,2)) goto error;
yading@10: 
yading@10:                 s->dest[1]+= 8;
yading@10:                 s->dest[2]+= 8;
yading@10: 
yading@10:                 /*emulate MB info in the relevant tables*/
yading@10:                 s->mbskip_table [mb_xy]=0;
yading@10:                 s->mbintra_table[mb_xy]=1;
yading@10:                 s->current_picture.qscale_table[mb_xy] = w->quant;
yading@10:                 mb_xy++;
yading@10:             }
yading@10:             s->dest[0]+= 8;
yading@10:         }
yading@10:         if(s->mb_y&1){
yading@10:             ff_mpeg_draw_horiz_band(s, (s->mb_y-1)*8, 16);
yading@10:         }
yading@10:     }
yading@10: 
yading@10: error:
yading@10:     ff_er_add_slice(&s->er, s->resync_mb_x, s->resync_mb_y,
yading@10:                         (s->mb_x>>1)-1, (s->mb_y>>1)-1,
yading@10:                         ER_MB_END );
yading@10:     return 0;
yading@10: }