libavcodec/ffv1.c

   1 /*
   2  * FFV1 codec for libavcodec
   3  *
   4  * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
   5  *
   6  * This file is part of FFmpeg.
   7  *
   8  * FFmpeg is free software; you can redistribute it and/or
   9  * modify it under the terms of the GNU Lesser General Public
  10  * License as published by the Free Software Foundation; either
  11  * version 2.1 of the License, or (at your option) any later version.
  12  *
  13  * FFmpeg is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  16  * Lesser General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU Lesser General Public
  19  * License along with FFmpeg; if not, write to the Free Software
  20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  21  */
  22
  23 /**
  24  * @file ffv1.c
  25  * FF Video Codec 1 (an experimental lossless codec)
  26  */
  27
  28 #include "avcodec.h"
  29 #include "bitstream.h"
  30 #include "dsputil.h"
  31 #include "rangecoder.h"
  32 #include "golomb.h"
  33
  34 #define MAX_PLANES 4
  35 #define CONTEXT_SIZE 32
  36
  37 static const int8_t quant3[256]={
  38  0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  39  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  40  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  41  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  42  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  43  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  44  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  45  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
  46 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  47 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  48 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  49 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  50 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  51 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  52 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  53 -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1, 0,
  54 };
  55 static const int8_t quant5[256]={
  56  0, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  57  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  58  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  59  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  60  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  61  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  62  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  63  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  64 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
  65 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
  66 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
  67 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
  68 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
  69 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
  70 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
  71 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-1,-1,-1,
  72 };
  73 static const int8_t quant7[256]={
  74  0, 1, 1, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  75  2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
  76  2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3,
  77  3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
  78  3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
  79  3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
  80  3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
  81  3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3,
  82 -3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,
  83 -3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,
  84 -3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,
  85 -3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,
  86 -3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-3,
  87 -3,-3,-3,-3,-3,-3,-3,-3,-3,-2,-2,-2,-2,-2,-2,-2,
  88 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
  89 -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-1,-1,
  90 };
  91 static const int8_t quant9[256]={
  92  0, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3,
  93  3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
  94  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
  95  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
  96  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
  97  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
  98  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
  99  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
 100 -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
 101 -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
 102 -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
 103 -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
 104 -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
 105 -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
 106 -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-3,-3,-3,-3,
 107 -3,-3,-3,-3,-3,-3,-3,-3,-3,-3,-2,-2,-2,-2,-1,-1,
 108 };
 109 static const int8_t quant11[256]={
 110  0, 1, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4,
 111  4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4,
 112  4, 4, 4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
 113  5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
 114  5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
 115  5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
 116  5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
 117  5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
 118 -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
 119 -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
 120 -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
 121 -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
 122 -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
 123 -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-4,-4,
 124 -4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,-4,
 125 -4,-4,-4,-4,-4,-3,-3,-3,-3,-3,-3,-3,-2,-2,-2,-1,
 126 };
 127 static const int8_t quant13[256]={
 128  0, 1, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4,
 129  4, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
 130  5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
 131  5, 5, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
 132  6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
 133  6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
 134  6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
 135  6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6, 6,
 136 -6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,
 137 -6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,
 138 -6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,
 139 -6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,
 140 -6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-6,-5,
 141 -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
 142 -5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,-5,
 143 -4,-4,-4,-4,-4,-4,-4,-4,-4,-3,-3,-3,-3,-2,-2,-1,
 144 };
 145
 146 static const uint8_t log2_run[32]={
 147  0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3,
 148  4, 4, 5, 5, 6, 6, 7, 7,
 149  8, 9,10,11,12,13,14,15,
 150 };
 151
 152 typedef struct VlcState{
 153     int16_t drift;
 154     uint16_t error_sum;
 155     int8_t bias;
 156     uint8_t count;
 157 } VlcState;
 158
 159 typedef struct PlaneContext{
 160     int context_count;
 161     uint8_t (*state)[CONTEXT_SIZE];
 162     VlcState *vlc_state;
 163     uint8_t interlace_bit_state[2];
 164 } PlaneContext;
 165
 166 typedef struct FFV1Context{
 167     AVCodecContext *avctx;
 168     RangeCoder c;
 169     GetBitContext gb;
 170     PutBitContext pb;
 171     int version;
 172     int width, height;
 173     int chroma_h_shift, chroma_v_shift;
 174     int flags;
 175     int picture_number;
 176     AVFrame picture;
 177     int plane_count;
 178     int ac;                              ///< 1-> CABAC 0-> golomb rice
 179     PlaneContext plane[MAX_PLANES];
 180     int16_t quant_table[5][256];
 181     int run_index;
 182     int colorspace;
 183
 184     DSPContext dsp;
 185 }FFV1Context;
 186
 187 static av_always_inline int fold(int diff, int bits){
 188     if(bits==8)
 189         diff= (int8_t)diff;
 190     else{
 191         diff+= 1<<(bits-1);
 192         diff&=(1<<bits)-1;
 193         diff-= 1<<(bits-1);
 194     }
 195
 196     return diff;
 197 }
 198
 199 static inline int predict(int_fast16_t *src, int_fast16_t *last){
 200     const int LT= last[-1];
 201     const int  T= last[ 0];
 202     const int L =  src[-1];
 203
 204     return mid_pred(L, L + T - LT, T);
 205 }
 206
 207 static inline int get_context(FFV1Context *f, int_fast16_t *src, int_fast16_t *last, int_fast16_t *last2){
 208     const int LT= last[-1];
 209     const int  T= last[ 0];
 210     const int RT= last[ 1];
 211     const int L =  src[-1];
 212
 213     if(f->quant_table[3][127]){
 214         const int TT= last2[0];
 215         const int LL=  src[-2];
 216         return f->quant_table[0][(L-LT) & 0xFF] + f->quant_table[1][(LT-T) & 0xFF] + f->quant_table[2][(T-RT) & 0xFF]
 217               +f->quant_table[3][(LL-L) & 0xFF] + f->quant_table[4][(TT-T) & 0xFF];
 218     }else
 219         return f->quant_table[0][(L-LT) & 0xFF] + f->quant_table[1][(LT-T) & 0xFF] + f->quant_table[2][(T-RT) & 0xFF];
 220 }
 221
 222 static inline void put_symbol(RangeCoder *c, uint8_t *state, int v, int is_signed){
 223     int i;
 224
 225     if(v){
 226         const int a= FFABS(v);
 227         const int e= av_log2(a);
 228         put_rac(c, state+0, 0);
 229
 230         assert(e<=9);
 231
 232         for(i=0; i<e; i++){
 233             put_rac(c, state+1+i, 1);  //1..10
 234         }
 235         put_rac(c, state+1+i, 0);
 236
 237         for(i=e-1; i>=0; i--){
 238             put_rac(c, state+22+i, (a>>i)&1); //22..31
 239         }
 240
 241         if(is_signed)
 242             put_rac(c, state+11 + e, v < 0); //11..21
 243     }else{
 244         put_rac(c, state+0, 1);
 245     }
 246 }
 247
 248 static inline int get_symbol(RangeCoder *c, uint8_t *state, int is_signed){
 249     if(get_rac(c, state+0))
 250         return 0;
 251     else{
 252         int i, e, a;
 253         e= 0;
 254         while(get_rac(c, state+1 + e)){ //1..10
 255             e++;
 256         }
 257         assert(e<=9);
 258
 259         a= 1;
 260         for(i=e-1; i>=0; i--){
 261             a += a + get_rac(c, state+22 + i); //22..31
 262         }
 263
 264         if(is_signed && get_rac(c, state+11 + e)) //11..21
 265             return -a;
 266         else
 267             return a;
 268     }
 269 }
 270
 271 static inline void update_vlc_state(VlcState * const state, const int v){
 272     int drift= state->drift;
 273     int count= state->count;
 274     state->error_sum += FFABS(v);
 275     drift += v;
 276
 277     if(count == 128){ //FIXME variable
 278         count >>= 1;
 279         drift >>= 1;
 280         state->error_sum >>= 1;
 281     }
 282     count++;
 283
 284     if(drift <= -count){
 285         if(state->bias > -128) state->bias--;
 286
 287         drift += count;
 288         if(drift <= -count)
 289             drift= -count + 1;
 290     }else if(drift > 0){
 291         if(state->bias <  127) state->bias++;
 292
 293         drift -= count;
 294         if(drift > 0)
 295             drift= 0;
 296     }
 297
 298     state->drift= drift;
 299     state->count= count;
 300 }
 301
 302 static inline void put_vlc_symbol(PutBitContext *pb, VlcState * const state, int v, int bits){
 303     int i, k, code;
 304 //printf("final: %d ", v);
 305     v = fold(v - state->bias, bits);
 306
 307     i= state->count;
 308     k=0;
 309     while(i < state->error_sum){ //FIXME optimize
 310         k++;
 311         i += i;
 312     }
 313
 314     assert(k<=8);
 315
 316 #if 0 // JPEG LS
 317     if(k==0 && 2*state->drift <= - state->count) code= v ^ (-1);
 318     else                                         code= v;
 319 #else
 320      code= v ^ ((2*state->drift + state->count)>>31);
 321 #endif
 322
 323 //printf("v:%d/%d bias:%d error:%d drift:%d count:%d k:%d\n", v, code, state->bias, state->error_sum, state->drift, state->count, k);
 324     set_sr_golomb(pb, code, k, 12, bits);
 325
 326     update_vlc_state(state, v);
 327 }
 328
 329 static inline int get_vlc_symbol(GetBitContext *gb, VlcState * const state, int bits){
 330     int k, i, v, ret;
 331
 332     i= state->count;
 333     k=0;
 334     while(i < state->error_sum){ //FIXME optimize
 335         k++;
 336         i += i;
 337     }
 338
 339     assert(k<=8);
 340
 341     v= get_sr_golomb(gb, k, 12, bits);
 342 //printf("v:%d bias:%d error:%d drift:%d count:%d k:%d", v, state->bias, state->error_sum, state->drift, state->count, k);
 343
 344 #if 0 // JPEG LS
 345     if(k==0 && 2*state->drift <= - state->count) v ^= (-1);
 346 #else
 347      v ^= ((2*state->drift + state->count)>>31);
 348 #endif
 349
 350     ret= fold(v + state->bias, bits);
 351
 352     update_vlc_state(state, v);
 353 //printf("final: %d\n", ret);
 354     return ret;
 355 }
 356
 357 #ifdef CONFIG_ENCODERS
 358 static inline int encode_line(FFV1Context *s, int w, int_fast16_t *sample[2], int plane_index, int bits){
 359     PlaneContext * const p= &s->plane[plane_index];
 360     RangeCoder * const c= &s->c;
 361     int x;
 362     int run_index= s->run_index;
 363     int run_count=0;
 364     int run_mode=0;
 365
 366     if(s->ac){
 367         if(c->bytestream_end - c->bytestream < w*20){
 368             av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
 369             return -1;
 370         }
 371     }else{
 372         if(s->pb.buf_end - s->pb.buf - (put_bits_count(&s->pb)>>3) < w*4){
 373             av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
 374             return -1;
 375         }
 376     }
 377
 378     for(x=0; x<w; x++){
 379         int diff, context;
 380
 381         context= get_context(s, sample[0]+x, sample[1]+x, sample[2]+x);
 382         diff= sample[0][x] - predict(sample[0]+x, sample[1]+x);
 383
 384         if(context < 0){
 385             context = -context;
 386             diff= -diff;
 387         }
 388
 389         diff= fold(diff, bits);
 390
 391         if(s->ac){
 392             put_symbol(c, p->state[context], diff, 1);
 393         }else{
 394             if(context == 0) run_mode=1;
 395
 396             if(run_mode){
 397
 398                 if(diff){
 399                     while(run_count >= 1<<log2_run[run_index]){
 400                         run_count -= 1<<log2_run[run_index];
 401                         run_index++;
 402                         put_bits(&s->pb, 1, 1);
 403                     }
 404
 405                     put_bits(&s->pb, 1 + log2_run[run_index], run_count);
 406                     if(run_index) run_index--;
 407                     run_count=0;
 408                     run_mode=0;
 409                     if(diff>0) diff--;
 410                 }else{
 411                     run_count++;
 412                 }
 413             }
 414
 415 //            printf("count:%d index:%d, mode:%d, x:%d y:%d pos:%d\n", run_count, run_index, run_mode, x, y, (int)put_bits_count(&s->pb));
 416
 417             if(run_mode == 0)
 418                 put_vlc_symbol(&s->pb, &p->vlc_state[context], diff, bits);
 419         }
 420     }
 421     if(run_mode){
 422         while(run_count >= 1<<log2_run[run_index]){
 423             run_count -= 1<<log2_run[run_index];
 424             run_index++;
 425             put_bits(&s->pb, 1, 1);
 426         }
 427
 428         if(run_count)
 429             put_bits(&s->pb, 1, 1);
 430     }
 431     s->run_index= run_index;
 432
 433     return 0;
 434 }
 435
 436 static void encode_plane(FFV1Context *s, uint8_t *src, int w, int h, int stride, int plane_index){
 437     int x,y,i;
 438     const int ring_size= s->avctx->context_model ? 3 : 2;
 439     int_fast16_t sample_buffer[ring_size][w+6], *sample[ring_size];
 440     s->run_index=0;
 441
 442     memset(sample_buffer, 0, sizeof(sample_buffer));
 443
 444     for(y=0; y<h; y++){
 445         for(i=0; i<ring_size; i++)
 446             sample[i]= sample_buffer[(h+i-y)%ring_size]+3;
 447
 448         sample[0][-1]= sample[1][0  ];
 449         sample[1][ w]= sample[1][w-1];
 450 //{START_TIMER
 451         for(x=0; x<w; x++){
 452             sample[0][x]= src[x + stride*y];
 453         }
 454         encode_line(s, w, sample, plane_index, 8);
 455 //STOP_TIMER("encode line")}
 456     }
 457 }
 458
 459 static void encode_rgb_frame(FFV1Context *s, uint32_t *src, int w, int h, int stride){
 460     int x, y, p, i;
 461     const int ring_size= s->avctx->context_model ? 3 : 2;
 462     int_fast16_t sample_buffer[3][ring_size][w+6], *sample[3][ring_size];
 463     s->run_index=0;
 464
 465     memset(sample_buffer, 0, sizeof(sample_buffer));
 466
 467     for(y=0; y<h; y++){
 468         for(i=0; i<ring_size; i++)
 469             for(p=0; p<3; p++)
 470                 sample[p][i]= sample_buffer[p][(h+i-y)%ring_size]+3;
 471
 472         for(x=0; x<w; x++){
 473             int v= src[x + stride*y];
 474             int b= v&0xFF;
 475             int g= (v>>8)&0xFF;
 476             int r= (v>>16)&0xFF;
 477
 478             b -= g;
 479             r -= g;
 480             g += (b + r)>>2;
 481             b += 0x100;
 482             r += 0x100;
 483
 484 //            assert(g>=0 && b>=0 && r>=0);
 485 //            assert(g<256 && b<512 && r<512);
 486             sample[0][0][x]= g;
 487             sample[1][0][x]= b;
 488             sample[2][0][x]= r;
 489         }
 490         for(p=0; p<3; p++){
 491             sample[p][0][-1]= sample[p][1][0  ];
 492             sample[p][1][ w]= sample[p][1][w-1];
 493             encode_line(s, w, sample[p], FFMIN(p, 1), 9);
 494         }
 495     }
 496 }
 497
 498 static void write_quant_table(RangeCoder *c, int16_t *quant_table){
 499     int last=0;
 500     int i;
 501     uint8_t state[CONTEXT_SIZE];
 502     memset(state, 128, sizeof(state));
 503
 504     for(i=1; i<128 ; i++){
 505         if(quant_table[i] != quant_table[i-1]){
 506             put_symbol(c, state, i-last-1, 0);
 507             last= i;
 508         }
 509     }
 510     put_symbol(c, state, i-last-1, 0);
 511 }
 512
 513 static void write_header(FFV1Context *f){
 514     uint8_t state[CONTEXT_SIZE];
 515     int i;
 516     RangeCoder * const c= &f->c;
 517
 518     memset(state, 128, sizeof(state));
 519
 520     put_symbol(c, state, f->version, 0);
 521     put_symbol(c, state, f->avctx->coder_type, 0);
 522     put_symbol(c, state, f->colorspace, 0); //YUV cs type
 523     put_rac(c, state, 1); //chroma planes
 524         put_symbol(c, state, f->chroma_h_shift, 0);
 525         put_symbol(c, state, f->chroma_v_shift, 0);
 526     put_rac(c, state, 0); //no transparency plane
 527
 528     for(i=0; i<5; i++)
 529         write_quant_table(c, f->quant_table[i]);
 530 }
 531 #endif /* CONFIG_ENCODERS */
 532
 533 static int common_init(AVCodecContext *avctx){
 534     FFV1Context *s = avctx->priv_data;
 535     int width, height;
 536
 537     s->avctx= avctx;
 538     s->flags= avctx->flags;
 539
 540     dsputil_init(&s->dsp, avctx);
 541
 542     width= s->width= avctx->width;
 543     height= s->height= avctx->height;
 544
 545     assert(width && height);
 546
 547     return 0;
 548 }
 549
 550 #ifdef CONFIG_ENCODERS
 551 static int encode_init(AVCodecContext *avctx)
 552 {
 553     FFV1Context *s = avctx->priv_data;
 554     int i;
 555
 556     common_init(avctx);
 557
 558     s->version=0;
 559     s->ac= avctx->coder_type;
 560
 561     s->plane_count=2;
 562     for(i=0; i<256; i++){
 563         s->quant_table[0][i]=           quant11[i];
 564         s->quant_table[1][i]=        11*quant11[i];
 565         if(avctx->context_model==0){
 566             s->quant_table[2][i]=     11*11*quant11[i];
 567             s->quant_table[3][i]=
 568             s->quant_table[4][i]=0;
 569         }else{
 570             s->quant_table[2][i]=     11*11*quant5 [i];
 571             s->quant_table[3][i]=   5*11*11*quant5 [i];
 572             s->quant_table[4][i]= 5*5*11*11*quant5 [i];
 573         }
 574     }
 575
 576     for(i=0; i<s->plane_count; i++){
 577         PlaneContext * const p= &s->plane[i];
 578
 579         if(avctx->context_model==0){
 580             p->context_count= (11*11*11+1)/2;
 581         }else{
 582             p->context_count= (11*11*5*5*5+1)/2;
 583         }
 584
 585         if(s->ac){
 586             if(!p->state) p->state= av_malloc(CONTEXT_SIZE*p->context_count*sizeof(uint8_t));
 587         }else{
 588             if(!p->vlc_state) p->vlc_state= av_malloc(p->context_count*sizeof(VlcState));
 589         }
 590     }
 591
 592     avctx->coded_frame= &s->picture;
 593     switch(avctx->pix_fmt){
 594     case PIX_FMT_YUV444P:
 595     case PIX_FMT_YUV422P:
 596     case PIX_FMT_YUV420P:
 597     case PIX_FMT_YUV411P:
 598     case PIX_FMT_YUV410P:
 599         s->colorspace= 0;
 600         break;
 601     case PIX_FMT_RGB32:
 602         s->colorspace= 1;
 603         break;
 604     default:
 605         av_log(avctx, AV_LOG_ERROR, "format not supported\n");
 606         return -1;
 607     }
 608     avcodec_get_chroma_sub_sample(avctx->pix_fmt, &s->chroma_h_shift, &s->chroma_v_shift);
 609
 610     s->picture_number=0;
 611
 612     return 0;
 613 }
 614 #endif /* CONFIG_ENCODERS */
 615
 616
 617 static void clear_state(FFV1Context *f){
 618     int i, j;
 619
 620     for(i=0; i<f->plane_count; i++){
 621         PlaneContext *p= &f->plane[i];
 622
 623         p->interlace_bit_state[0]= 128;
 624         p->interlace_bit_state[1]= 128;
 625
 626         for(j=0; j<p->context_count; j++){
 627             if(f->ac){
 628                 memset(p->state[j], 128, sizeof(uint8_t)*CONTEXT_SIZE);
 629             }else{
 630                 p->vlc_state[j].drift= 0;
 631                 p->vlc_state[j].error_sum= 4; //FFMAX((RANGE + 32)/64, 2);
 632                 p->vlc_state[j].bias= 0;
 633                 p->vlc_state[j].count= 1;
 634             }
 635         }
 636     }
 637 }
 638
 639 #ifdef CONFIG_ENCODERS
 640 static int encode_frame(AVCodecContext *avctx, unsigned char *buf, int buf_size, void *data){
 641     FFV1Context *f = avctx->priv_data;
 642     RangeCoder * const c= &f->c;
 643     AVFrame *pict = data;
 644     const int width= f->width;
 645     const int height= f->height;
 646     AVFrame * const p= &f->picture;
 647     int used_count= 0;
 648     uint8_t keystate=128;
 649
 650     ff_init_range_encoder(c, buf, buf_size);
 651 //    ff_init_cabac_states(c, ff_h264_lps_range, ff_h264_mps_state, ff_h264_lps_state, 64);
 652     ff_build_rac_states(c, 0.05*(1LL<<32), 256-8);
 653
 654     *p = *pict;
 655     p->pict_type= FF_I_TYPE;
 656
 657     if(avctx->gop_size==0 || f->picture_number % avctx->gop_size == 0){
 658         put_rac(c, &keystate, 1);
 659         p->key_frame= 1;
 660         write_header(f);
 661         clear_state(f);
 662     }else{
 663         put_rac(c, &keystate, 0);
 664         p->key_frame= 0;
 665     }
 666
 667     if(!f->ac){
 668         used_count += ff_rac_terminate(c);
 669 //printf("pos=%d\n", used_count);
 670         init_put_bits(&f->pb, buf + used_count, buf_size - used_count);
 671     }
 672
 673     if(f->colorspace==0){
 674         const int chroma_width = -((-width )>>f->chroma_h_shift);
 675         const int chroma_height= -((-height)>>f->chroma_v_shift);
 676
 677         encode_plane(f, p->data[0], width, height, p->linesize[0], 0);
 678
 679         encode_plane(f, p->data[1], chroma_width, chroma_height, p->linesize[1], 1);
 680         encode_plane(f, p->data[2], chroma_width, chroma_height, p->linesize[2], 1);
 681     }else{
 682         encode_rgb_frame(f, (uint32_t*)(p->data[0]), width, height, p->linesize[0]/4);
 683     }
 684     emms_c();
 685
 686     f->picture_number++;
 687
 688     if(f->ac){
 689         return ff_rac_terminate(c);
 690     }else{
 691         flush_put_bits(&f->pb); //nicer padding FIXME
 692         return used_count + (put_bits_count(&f->pb)+7)/8;
 693     }
 694 }
 695 #endif /* CONFIG_ENCODERS */
 696
 697 static int common_end(AVCodecContext *avctx){
 698     FFV1Context *s = avctx->priv_data;
 699     int i;
 700
 701     for(i=0; i<s->plane_count; i++){
 702         PlaneContext *p= &s->plane[i];
 703
 704         av_freep(&p->state);
 705         av_freep(&p->vlc_state);
 706     }
 707
 708     return 0;
 709 }
 710
 711 static inline void decode_line(FFV1Context *s, int w, int_fast16_t *sample[2], int plane_index, int bits){
 712     PlaneContext * const p= &s->plane[plane_index];
 713     RangeCoder * const c= &s->c;
 714     int x;
 715     int run_count=0;
 716     int run_mode=0;
 717     int run_index= s->run_index;
 718
 719     for(x=0; x<w; x++){
 720         int diff, context, sign;
 721
 722         context= get_context(s, sample[1] + x, sample[0] + x, sample[1] + x);
 723         if(context < 0){
 724             context= -context;
 725             sign=1;
 726         }else
 727             sign=0;
 728
 729
 730         if(s->ac){
 731             diff= get_symbol(c, p->state[context], 1);
 732         }else{
 733             if(context == 0 && run_mode==0) run_mode=1;
 734
 735             if(run_mode){
 736                 if(run_count==0 && run_mode==1){
 737                     if(get_bits1(&s->gb)){
 738                         run_count = 1<<log2_run[run_index];
 739                         if(x + run_count <= w) run_index++;
 740                     }else{
 741                         if(log2_run[run_index]) run_count = get_bits(&s->gb, log2_run[run_index]);
 742                         else run_count=0;
 743                         if(run_index) run_index--;
 744                         run_mode=2;
 745                     }
 746                 }
 747                 run_count--;
 748                 if(run_count < 0){
 749                     run_mode=0;
 750                     run_count=0;
 751                     diff= get_vlc_symbol(&s->gb, &p->vlc_state[context], bits);
 752                     if(diff>=0) diff++;
 753                 }else
 754                     diff=0;
 755             }else
 756                 diff= get_vlc_symbol(&s->gb, &p->vlc_state[context], bits);
 757
 758 //            printf("count:%d index:%d, mode:%d, x:%d y:%d pos:%d\n", run_count, run_index, run_mode, x, y, get_bits_count(&s->gb));
 759         }
 760
 761         if(sign) diff= -diff;
 762
 763         sample[1][x]= (predict(sample[1] + x, sample[0] + x) + diff) & ((1<<bits)-1);
 764     }
 765     s->run_index= run_index;
 766 }
 767
 768 static void decode_plane(FFV1Context *s, uint8_t *src, int w, int h, int stride, int plane_index){
 769     int x, y;
 770     int_fast16_t sample_buffer[2][w+6];
 771     int_fast16_t *sample[2]= {sample_buffer[0]+3, sample_buffer[1]+3};
 772
 773     s->run_index=0;
 774
 775     memset(sample_buffer, 0, sizeof(sample_buffer));
 776
 777     for(y=0; y<h; y++){
 778         int_fast16_t *temp= sample[0]; //FIXME try a normal buffer
 779
 780         sample[0]= sample[1];
 781         sample[1]= temp;
 782
 783         sample[1][-1]= sample[0][0  ];
 784         sample[0][ w]= sample[0][w-1];
 785
 786 //{START_TIMER
 787         decode_line(s, w, sample, plane_index, 8);
 788         for(x=0; x<w; x++){
 789             src[x + stride*y]= sample[1][x];
 790         }
 791 //STOP_TIMER("decode-line")}
 792     }
 793 }
 794
 795 static void decode_rgb_frame(FFV1Context *s, uint32_t *src, int w, int h, int stride){
 796     int x, y, p;
 797     int_fast16_t sample_buffer[3][2][w+6];
 798     int_fast16_t *sample[3][2]= {
 799         {sample_buffer[0][0]+3, sample_buffer[0][1]+3},
 800         {sample_buffer[1][0]+3, sample_buffer[1][1]+3},
 801         {sample_buffer[2][0]+3, sample_buffer[2][1]+3}};
 802
 803     s->run_index=0;
 804
 805     memset(sample_buffer, 0, sizeof(sample_buffer));
 806
 807     for(y=0; y<h; y++){
 808         for(p=0; p<3; p++){
 809             int_fast16_t *temp= sample[p][0]; //FIXME try a normal buffer
 810
 811             sample[p][0]= sample[p][1];
 812             sample[p][1]= temp;
 813
 814             sample[p][1][-1]= sample[p][0][0  ];
 815             sample[p][0][ w]= sample[p][0][w-1];
 816             decode_line(s, w, sample[p], FFMIN(p, 1), 9);
 817         }
 818         for(x=0; x<w; x++){
 819             int g= sample[0][1][x];
 820             int b= sample[1][1][x];
 821             int r= sample[2][1][x];
 822
 823 //            assert(g>=0 && b>=0 && r>=0);
 824 //            assert(g<256 && b<512 && r<512);
 825
 826             b -= 0x100;
 827             r -= 0x100;
 828             g -= (b + r)>>2;
 829             b += g;
 830             r += g;
 831
 832             src[x + stride*y]= b + (g<<8) + (r<<16);
 833         }
 834     }
 835 }
 836
 837 static int read_quant_table(RangeCoder *c, int16_t *quant_table, int scale){
 838     int v;
 839     int i=0;
 840     uint8_t state[CONTEXT_SIZE];
 841
 842     memset(state, 128, sizeof(state));
 843
 844     for(v=0; i<128 ; v++){
 845         int len= get_symbol(c, state, 0) + 1;
 846
 847         if(len + i > 128) return -1;
 848
 849         while(len--){
 850             quant_table[i] = scale*v;
 851             i++;
 852 //printf("%2d ",v);
 853 //if(i%16==0) printf("\n");
 854         }
 855     }
 856
 857     for(i=1; i<128; i++){
 858         quant_table[256-i]= -quant_table[i];
 859     }
 860     quant_table[128]= -quant_table[127];
 861
 862     return 2*v - 1;
 863 }
 864
 865 static int read_header(FFV1Context *f){
 866     uint8_t state[CONTEXT_SIZE];
 867     int i, context_count;
 868     RangeCoder * const c= &f->c;
 869
 870     memset(state, 128, sizeof(state));
 871
 872     f->version= get_symbol(c, state, 0);
 873     f->ac= f->avctx->coder_type= get_symbol(c, state, 0);
 874     f->colorspace= get_symbol(c, state, 0); //YUV cs type
 875     get_rac(c, state); //no chroma = false
 876     f->chroma_h_shift= get_symbol(c, state, 0);
 877     f->chroma_v_shift= get_symbol(c, state, 0);
 878     get_rac(c, state); //transparency plane
 879     f->plane_count= 2;
 880
 881     if(f->colorspace==0){
 882         switch(16*f->chroma_h_shift + f->chroma_v_shift){
 883         case 0x00: f->avctx->pix_fmt= PIX_FMT_YUV444P; break;
 884         case 0x10: f->avctx->pix_fmt= PIX_FMT_YUV422P; break;
 885         case 0x11: f->avctx->pix_fmt= PIX_FMT_YUV420P; break;
 886         case 0x20: f->avctx->pix_fmt= PIX_FMT_YUV411P; break;
 887         case 0x22: f->avctx->pix_fmt= PIX_FMT_YUV410P; break;
 888         default:
 889             av_log(f->avctx, AV_LOG_ERROR, "format not supported\n");
 890             return -1;
 891         }
 892     }else if(f->colorspace==1){
 893         if(f->chroma_h_shift || f->chroma_v_shift){
 894             av_log(f->avctx, AV_LOG_ERROR, "chroma subsampling not supported in this colorspace\n");
 895             return -1;
 896         }
 897         f->avctx->pix_fmt= PIX_FMT_RGB32;
 898     }else{
 899         av_log(f->avctx, AV_LOG_ERROR, "colorspace not supported\n");
 900         return -1;
 901     }
 902
 903 //printf("%d %d %d\n", f->chroma_h_shift, f->chroma_v_shift,f->avctx->pix_fmt);
 904
 905     context_count=1;
 906     for(i=0; i<5; i++){
 907         context_count*= read_quant_table(c, f->quant_table[i], context_count);
 908         if(context_count < 0 || context_count > 32768){
 909             av_log(f->avctx, AV_LOG_ERROR, "read_quant_table error\n");
 910             return -1;
 911         }
 912     }
 913     context_count= (context_count+1)/2;
 914
 915     for(i=0; i<f->plane_count; i++){
 916         PlaneContext * const p= &f->plane[i];
 917
 918         p->context_count= context_count;
 919
 920         if(f->ac){
 921             if(!p->state) p->state= av_malloc(CONTEXT_SIZE*p->context_count*sizeof(uint8_t));
 922         }else{
 923             if(!p->vlc_state) p->vlc_state= av_malloc(p->context_count*sizeof(VlcState));
 924         }
 925     }
 926
 927     return 0;
 928 }
 929
 930 static int decode_init(AVCodecContext *avctx)
 931 {
 932 //    FFV1Context *s = avctx->priv_data;
 933
 934     common_init(avctx);
 935
 936     return 0;
 937 }
 938
 939 static int decode_frame(AVCodecContext *avctx, void *data, int *data_size, uint8_t *buf, int buf_size){
 940     FFV1Context *f = avctx->priv_data;
 941     RangeCoder * const c= &f->c;
 942     const int width= f->width;
 943     const int height= f->height;
 944     AVFrame * const p= &f->picture;
 945     int bytes_read;
 946     uint8_t keystate= 128;
 947
 948     AVFrame *picture = data;
 949
 950     ff_init_range_decoder(c, buf, buf_size);
 951     ff_build_rac_states(c, 0.05*(1LL<<32), 256-8);
 952
 953
 954     p->pict_type= FF_I_TYPE; //FIXME I vs. P
 955     if(get_rac(c, &keystate)){
 956         p->key_frame= 1;
 957         if(read_header(f) < 0)
 958             return -1;
 959         clear_state(f);
 960     }else{
 961         p->key_frame= 0;
 962     }
 963     if(!f->plane[0].state && !f->plane[0].vlc_state)
 964         return -1;
 965
 966     p->reference= 0;
 967     if(avctx->get_buffer(avctx, p) < 0){
 968         av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
 969         return -1;
 970     }
 971
 972     if(avctx->debug&FF_DEBUG_PICT_INFO)
 973         av_log(avctx, AV_LOG_ERROR, "keyframe:%d coder:%d\n", p->key_frame, f->ac);
 974
 975     if(!f->ac){
 976         bytes_read = c->bytestream - c->bytestream_start - 1;
 977         if(bytes_read ==0) av_log(avctx, AV_LOG_ERROR, "error at end of AC stream\n"); //FIXME
 978 //printf("pos=%d\n", bytes_read);
 979         init_get_bits(&f->gb, buf + bytes_read, buf_size - bytes_read);
 980     } else {
 981         bytes_read = 0; /* avoid warning */
 982     }
 983
 984     if(f->colorspace==0){
 985         const int chroma_width = -((-width )>>f->chroma_h_shift);
 986         const int chroma_height= -((-height)>>f->chroma_v_shift);
 987         decode_plane(f, p->data[0], width, height, p->linesize[0], 0);
 988
 989         decode_plane(f, p->data[1], chroma_width, chroma_height, p->linesize[1], 1);
 990         decode_plane(f, p->data[2], chroma_width, chroma_height, p->linesize[2], 1);
 991     }else{
 992         decode_rgb_frame(f, (uint32_t*)p->data[0], width, height, p->linesize[0]/4);
 993     }
 994
 995     emms_c();
 996
 997     f->picture_number++;
 998
 999     *picture= *p;
1000
1001     avctx->release_buffer(avctx, p); //FIXME
1002
1003     *data_size = sizeof(AVFrame);
1004
1005     if(f->ac){
1006         bytes_read= c->bytestream - c->bytestream_start - 1;
1007         if(bytes_read ==0) av_log(f->avctx, AV_LOG_ERROR, "error at end of frame\n");
1008     }else{
1009         bytes_read+= (get_bits_count(&f->gb)+7)/8;
1010     }
1011
1012     return bytes_read;
1013 }
1014
1015 AVCodec ffv1_decoder = {
1016     "ffv1",
1017     CODEC_TYPE_VIDEO,
1018     CODEC_ID_FFV1,
1019     sizeof(FFV1Context),
1020     decode_init,
1021     NULL,
1022     common_end,
1023     decode_frame,
1024     CODEC_CAP_DR1 /*| CODEC_CAP_DRAW_HORIZ_BAND*/,
1025     NULL
1026 };
1027
1028 #ifdef CONFIG_ENCODERS
1029 AVCodec ffv1_encoder = {
1030     "ffv1",
1031     CODEC_TYPE_VIDEO,
1032     CODEC_ID_FFV1,
1033     sizeof(FFV1Context),
1034     encode_init,
1035     encode_frame,
1036     common_end,
1037     .pix_fmts= (enum PixelFormat[]){PIX_FMT_YUV420P, PIX_FMT_YUV444P, PIX_FMT_YUV422P, PIX_FMT_YUV411P, PIX_FMT_YUV410P, PIX_FMT_RGB32, -1},
1038 };
1039 #endif