FFmpeg  4.4.6
snowenc.c
Go to the documentation of this file.
1 /*
2  * Copyright (C) 2004 Michael Niedermayer <michaelni@gmx.at>
3  *
4  * This file is part of FFmpeg.
5  *
6  * FFmpeg is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU Lesser General Public
8  * License as published by the Free Software Foundation; either
9  * version 2.1 of the License, or (at your option) any later version.
10  *
11  * FFmpeg is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14  * Lesser General Public License for more details.
15  *
16  * You should have received a copy of the GNU Lesser General Public
17  * License along with FFmpeg; if not, write to the Free Software
18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19  */
20 
21 #include "libavutil/intmath.h"
22 #include "libavutil/libm.h"
23 #include "libavutil/log.h"
24 #include "libavutil/opt.h"
25 #include "libavutil/pixdesc.h"
26 #include "avcodec.h"
27 #include "internal.h"
28 #include "packet_internal.h"
29 #include "snow_dwt.h"
30 #include "snow.h"
31 
32 #include "rangecoder.h"
33 #include "mathops.h"
34 
35 #include "mpegvideo.h"
36 #include "h263.h"
37 
39 {
40  SnowContext *s = avctx->priv_data;
41  int plane_index, ret;
42  int i;
43 
44 #if FF_API_PRIVATE_OPT
46  if (avctx->prediction_method)
47  s->pred = avctx->prediction_method;
49 #endif
50 
51  if(s->pred == DWT_97
52  && (avctx->flags & AV_CODEC_FLAG_QSCALE)
53  && avctx->global_quality == 0){
54  av_log(avctx, AV_LOG_ERROR, "The 9/7 wavelet is incompatible with lossless mode.\n");
55  return AVERROR(EINVAL);
56  }
57 
58  s->spatial_decomposition_type= s->pred; //FIXME add decorrelator type r transform_type
59 
60  s->mv_scale = (avctx->flags & AV_CODEC_FLAG_QPEL) ? 2 : 4;
61  s->block_max_depth= (avctx->flags & AV_CODEC_FLAG_4MV ) ? 1 : 0;
62 
63  for(plane_index=0; plane_index<3; plane_index++){
64  s->plane[plane_index].diag_mc= 1;
65  s->plane[plane_index].htaps= 6;
66  s->plane[plane_index].hcoeff[0]= 40;
67  s->plane[plane_index].hcoeff[1]= -10;
68  s->plane[plane_index].hcoeff[2]= 2;
69  s->plane[plane_index].fast_mc= 1;
70  }
71 
72  if ((ret = ff_snow_common_init(avctx)) < 0) {
73  return ret;
74  }
75  ff_mpegvideoencdsp_init(&s->mpvencdsp, avctx);
76 
78 
79  s->version=0;
80 
81  s->m.avctx = avctx;
82  s->m.bit_rate= avctx->bit_rate;
83  s->m.lmin = avctx->mb_lmin;
84  s->m.lmax = avctx->mb_lmax;
85  s->m.mb_num = (avctx->width * avctx->height + 255) / 256; // For ratecontrol
86 
87  s->m.me.temp =
88  s->m.me.scratchpad= av_mallocz_array((avctx->width+64), 2*16*2*sizeof(uint8_t));
89  s->m.me.map = av_mallocz(ME_MAP_SIZE*sizeof(uint32_t));
90  s->m.me.score_map = av_mallocz(ME_MAP_SIZE*sizeof(uint32_t));
91  s->m.sc.obmc_scratchpad= av_mallocz(MB_SIZE*MB_SIZE*12*sizeof(uint32_t));
92  if (!s->m.me.scratchpad || !s->m.me.map || !s->m.me.score_map || !s->m.sc.obmc_scratchpad)
93  return AVERROR(ENOMEM);
94 
95  ff_h263_encode_init(&s->m); //mv_penalty
96 
97  s->max_ref_frames = av_clip(avctx->refs, 1, MAX_REF_FRAMES);
98 
99  if(avctx->flags&AV_CODEC_FLAG_PASS1){
100  if(!avctx->stats_out)
101  avctx->stats_out = av_mallocz(256);
102 
103  if (!avctx->stats_out)
104  return AVERROR(ENOMEM);
105  }
106  if((avctx->flags&AV_CODEC_FLAG_PASS2) || !(avctx->flags&AV_CODEC_FLAG_QSCALE)){
107  ret = ff_rate_control_init(&s->m);
108  if(ret < 0)
109  return ret;
110  }
111  s->pass1_rc= !(avctx->flags & (AV_CODEC_FLAG_QSCALE|AV_CODEC_FLAG_PASS2));
112 
113  switch(avctx->pix_fmt){
114  case AV_PIX_FMT_YUV444P:
115 // case AV_PIX_FMT_YUV422P:
116  case AV_PIX_FMT_YUV420P:
117 // case AV_PIX_FMT_YUV411P:
118  case AV_PIX_FMT_YUV410P:
119  s->nb_planes = 3;
120  s->colorspace_type= 0;
121  break;
122  case AV_PIX_FMT_GRAY8:
123  s->nb_planes = 1;
124  s->colorspace_type = 1;
125  break;
126 /* case AV_PIX_FMT_RGB32:
127  s->colorspace= 1;
128  break;*/
129  default:
130  av_log(avctx, AV_LOG_ERROR, "pixel format not supported\n");
131  return AVERROR_PATCHWELCOME;
132  }
133 
134  ret = av_pix_fmt_get_chroma_sub_sample(avctx->pix_fmt, &s->chroma_h_shift,
135  &s->chroma_v_shift);
136  if (ret) {
137  av_log(avctx, AV_LOG_ERROR, "pixel format invalid or unknown\n");
138  return ret;
139  }
140 
141  ff_set_cmp(&s->mecc, s->mecc.me_cmp, s->avctx->me_cmp);
142  ff_set_cmp(&s->mecc, s->mecc.me_sub_cmp, s->avctx->me_sub_cmp);
143 
144  s->input_picture = av_frame_alloc();
145  if (!s->input_picture)
146  return AVERROR(ENOMEM);
147 
148  if ((ret = ff_snow_get_buffer(s, s->input_picture)) < 0)
149  return ret;
150 
151  if(s->motion_est == FF_ME_ITER){
152  int size= s->b_width * s->b_height << 2*s->block_max_depth;
153  for(i=0; i<s->max_ref_frames; i++){
154  s->ref_mvs[i]= av_mallocz_array(size, sizeof(int16_t[2]));
155  s->ref_scores[i]= av_mallocz_array(size, sizeof(uint32_t));
156  if (!s->ref_mvs[i] || !s->ref_scores[i])
157  return AVERROR(ENOMEM);
158  }
159  }
160 
161  return 0;
162 }
163 
164 //near copy & paste from dsputil, FIXME
165 static int pix_sum(uint8_t * pix, int line_size, int w, int h)
166 {
167  int s, i, j;
168 
169  s = 0;
170  for (i = 0; i < h; i++) {
171  for (j = 0; j < w; j++) {
172  s += pix[0];
173  pix ++;
174  }
175  pix += line_size - w;
176  }
177  return s;
178 }
179 
180 //near copy & paste from dsputil, FIXME
181 static int pix_norm1(uint8_t * pix, int line_size, int w)
182 {
183  int s, i, j;
184  const uint32_t *sq = ff_square_tab + 256;
185 
186  s = 0;
187  for (i = 0; i < w; i++) {
188  for (j = 0; j < w; j ++) {
189  s += sq[pix[0]];
190  pix ++;
191  }
192  pix += line_size - w;
193  }
194  return s;
195 }
196 
197 static inline int get_penalty_factor(int lambda, int lambda2, int type){
198  switch(type&0xFF){
199  default:
200  case FF_CMP_SAD:
201  return lambda>>FF_LAMBDA_SHIFT;
202  case FF_CMP_DCT:
203  return (3*lambda)>>(FF_LAMBDA_SHIFT+1);
204  case FF_CMP_W53:
205  return (4*lambda)>>(FF_LAMBDA_SHIFT);
206  case FF_CMP_W97:
207  return (2*lambda)>>(FF_LAMBDA_SHIFT);
208  case FF_CMP_SATD:
209  case FF_CMP_DCT264:
210  return (2*lambda)>>FF_LAMBDA_SHIFT;
211  case FF_CMP_RD:
212  case FF_CMP_PSNR:
213  case FF_CMP_SSE:
214  case FF_CMP_NSSE:
215  return lambda2>>FF_LAMBDA_SHIFT;
216  case FF_CMP_BIT:
217  return 1;
218  }
219 }
220 
221 //FIXME copy&paste
222 #define P_LEFT P[1]
223 #define P_TOP P[2]
224 #define P_TOPRIGHT P[3]
225 #define P_MEDIAN P[4]
226 #define P_MV1 P[9]
227 #define FLAG_QPEL 1 //must be 1
228 
229 static int encode_q_branch(SnowContext *s, int level, int x, int y){
230  uint8_t p_buffer[1024];
231  uint8_t i_buffer[1024];
232  uint8_t p_state[sizeof(s->block_state)];
233  uint8_t i_state[sizeof(s->block_state)];
234  RangeCoder pc, ic;
235  uint8_t *pbbak= s->c.bytestream;
236  uint8_t *pbbak_start= s->c.bytestream_start;
237  int score, score2, iscore, i_len, p_len, block_s, sum, base_bits;
238  const int w= s->b_width << s->block_max_depth;
239  const int h= s->b_height << s->block_max_depth;
240  const int rem_depth= s->block_max_depth - level;
241  const int index= (x + y*w) << rem_depth;
242  const int block_w= 1<<(LOG2_MB_SIZE - level);
243  int trx= (x+1)<<rem_depth;
244  int try= (y+1)<<rem_depth;
245  const BlockNode *left = x ? &s->block[index-1] : &null_block;
246  const BlockNode *top = y ? &s->block[index-w] : &null_block;
247  const BlockNode *right = trx<w ? &s->block[index+1] : &null_block;
248  const BlockNode *bottom= try<h ? &s->block[index+w] : &null_block;
249  const BlockNode *tl = y && x ? &s->block[index-w-1] : left;
250  const BlockNode *tr = y && trx<w && ((x&1)==0 || level==0) ? &s->block[index-w+(1<<rem_depth)] : tl; //FIXME use lt
251  int pl = left->color[0];
252  int pcb= left->color[1];
253  int pcr= left->color[2];
254  int pmx, pmy;
255  int mx=0, my=0;
256  int l,cr,cb;
257  const int stride= s->current_picture->linesize[0];
258  const int uvstride= s->current_picture->linesize[1];
259  uint8_t *current_data[3]= { s->input_picture->data[0] + (x + y* stride)*block_w,
260  s->input_picture->data[1] + ((x*block_w)>>s->chroma_h_shift) + ((y*uvstride*block_w)>>s->chroma_v_shift),
261  s->input_picture->data[2] + ((x*block_w)>>s->chroma_h_shift) + ((y*uvstride*block_w)>>s->chroma_v_shift)};
262  int P[10][2];
263  int16_t last_mv[3][2];
264  int qpel= !!(s->avctx->flags & AV_CODEC_FLAG_QPEL); //unused
265  const int shift= 1+qpel;
266  MotionEstContext *c= &s->m.me;
267  int ref_context= av_log2(2*left->ref) + av_log2(2*top->ref);
268  int mx_context= av_log2(2*FFABS(left->mx - top->mx));
269  int my_context= av_log2(2*FFABS(left->my - top->my));
270  int s_context= 2*left->level + 2*top->level + tl->level + tr->level;
271  int ref, best_ref, ref_score, ref_mx, ref_my;
272  int range = MAX_MV >> (1 + qpel);
273 
274  av_assert0(sizeof(s->block_state) >= 256);
275  if(s->keyframe){
276  set_blocks(s, level, x, y, pl, pcb, pcr, 0, 0, 0, BLOCK_INTRA);
277  return 0;
278  }
279 
280 // clip predictors / edge ?
281 
282  P_LEFT[0]= left->mx;
283  P_LEFT[1]= left->my;
284  P_TOP [0]= top->mx;
285  P_TOP [1]= top->my;
286  P_TOPRIGHT[0]= tr->mx;
287  P_TOPRIGHT[1]= tr->my;
288 
289  last_mv[0][0]= s->block[index].mx;
290  last_mv[0][1]= s->block[index].my;
291  last_mv[1][0]= right->mx;
292  last_mv[1][1]= right->my;
293  last_mv[2][0]= bottom->mx;
294  last_mv[2][1]= bottom->my;
295 
296  s->m.mb_stride=2;
297  s->m.mb_x=
298  s->m.mb_y= 0;
299  c->skip= 0;
300 
301  av_assert1(c-> stride == stride);
302  av_assert1(c->uvstride == uvstride);
303 
304  c->penalty_factor = get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_cmp);
305  c->sub_penalty_factor= get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_sub_cmp);
306  c->mb_penalty_factor = get_penalty_factor(s->lambda, s->lambda2, c->avctx->mb_cmp);
307  c->current_mv_penalty= c->mv_penalty[s->m.f_code=1] + MAX_DMV;
308 
309  c->xmin = - x*block_w - 16+3;
310  c->ymin = - y*block_w - 16+3;
311  c->xmax = - (x+1)*block_w + (w<<(LOG2_MB_SIZE - s->block_max_depth)) + 16-3;
312  c->ymax = - (y+1)*block_w + (h<<(LOG2_MB_SIZE - s->block_max_depth)) + 16-3;
313 
314  c->xmin = FFMAX(c->xmin,-range);
315  c->xmax = FFMIN(c->xmax, range);
316  c->ymin = FFMAX(c->ymin,-range);
317  c->ymax = FFMIN(c->ymax, range);
318 
319  if(P_LEFT[0] > (c->xmax<<shift)) P_LEFT[0] = (c->xmax<<shift);
320  if(P_LEFT[1] > (c->ymax<<shift)) P_LEFT[1] = (c->ymax<<shift);
321  if(P_TOP[0] > (c->xmax<<shift)) P_TOP[0] = (c->xmax<<shift);
322  if(P_TOP[1] > (c->ymax<<shift)) P_TOP[1] = (c->ymax<<shift);
323  if(P_TOPRIGHT[0] < (c->xmin * (1<<shift))) P_TOPRIGHT[0]= (c->xmin * (1<<shift));
324  if(P_TOPRIGHT[0] > (c->xmax<<shift)) P_TOPRIGHT[0]= (c->xmax<<shift); //due to pmx no clip
325  if(P_TOPRIGHT[1] > (c->ymax<<shift)) P_TOPRIGHT[1]= (c->ymax<<shift);
326 
327  P_MEDIAN[0]= mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
328  P_MEDIAN[1]= mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
329 
330  if (!y) {
331  c->pred_x= P_LEFT[0];
332  c->pred_y= P_LEFT[1];
333  } else {
334  c->pred_x = P_MEDIAN[0];
335  c->pred_y = P_MEDIAN[1];
336  }
337 
338  score= INT_MAX;
339  best_ref= 0;
340  for(ref=0; ref<s->ref_frames; ref++){
341  init_ref(c, current_data, s->last_picture[ref]->data, NULL, block_w*x, block_w*y, 0);
342 
343  ref_score= ff_epzs_motion_search(&s->m, &ref_mx, &ref_my, P, 0, /*ref_index*/ 0, last_mv,
344  (1<<16)>>shift, level-LOG2_MB_SIZE+4, block_w);
345 
346  av_assert2(ref_mx >= c->xmin);
347  av_assert2(ref_mx <= c->xmax);
348  av_assert2(ref_my >= c->ymin);
349  av_assert2(ref_my <= c->ymax);
350 
351  ref_score= c->sub_motion_search(&s->m, &ref_mx, &ref_my, ref_score, 0, 0, level-LOG2_MB_SIZE+4, block_w);
352  ref_score= ff_get_mb_score(&s->m, ref_mx, ref_my, 0, 0, level-LOG2_MB_SIZE+4, block_w, 0);
353  ref_score+= 2*av_log2(2*ref)*c->penalty_factor;
354  if(s->ref_mvs[ref]){
355  s->ref_mvs[ref][index][0]= ref_mx;
356  s->ref_mvs[ref][index][1]= ref_my;
357  s->ref_scores[ref][index]= ref_score;
358  }
359  if(score > ref_score){
360  score= ref_score;
361  best_ref= ref;
362  mx= ref_mx;
363  my= ref_my;
364  }
365  }
366  //FIXME if mb_cmp != SSE then intra cannot be compared currently and mb_penalty vs. lambda2
367 
368  // subpel search
369  base_bits= get_rac_count(&s->c) - 8*(s->c.bytestream - s->c.bytestream_start);
370  pc= s->c;
371  pc.bytestream_start=
372  pc.bytestream= p_buffer; //FIXME end/start? and at the other stoo
373  memcpy(p_state, s->block_state, sizeof(s->block_state));
374 
375  if(level!=s->block_max_depth)
376  put_rac(&pc, &p_state[4 + s_context], 1);
377  put_rac(&pc, &p_state[1 + left->type + top->type], 0);
378  if(s->ref_frames > 1)
379  put_symbol(&pc, &p_state[128 + 1024 + 32*ref_context], best_ref, 0);
380  pred_mv(s, &pmx, &pmy, best_ref, left, top, tr);
381  put_symbol(&pc, &p_state[128 + 32*(mx_context + 16*!!best_ref)], mx - pmx, 1);
382  put_symbol(&pc, &p_state[128 + 32*(my_context + 16*!!best_ref)], my - pmy, 1);
383  p_len= pc.bytestream - pc.bytestream_start;
384  score += (s->lambda2*(get_rac_count(&pc)-base_bits))>>FF_LAMBDA_SHIFT;
385 
386  block_s= block_w*block_w;
387  sum = pix_sum(current_data[0], stride, block_w, block_w);
388  l= (sum + block_s/2)/block_s;
389  iscore = pix_norm1(current_data[0], stride, block_w) - 2*l*sum + l*l*block_s;
390 
391  if (s->nb_planes > 2) {
392  block_s= block_w*block_w>>(s->chroma_h_shift + s->chroma_v_shift);
393  sum = pix_sum(current_data[1], uvstride, block_w>>s->chroma_h_shift, block_w>>s->chroma_v_shift);
394  cb= (sum + block_s/2)/block_s;
395  // iscore += pix_norm1(&current_mb[1][0], uvstride, block_w>>1) - 2*cb*sum + cb*cb*block_s;
396  sum = pix_sum(current_data[2], uvstride, block_w>>s->chroma_h_shift, block_w>>s->chroma_v_shift);
397  cr= (sum + block_s/2)/block_s;
398  // iscore += pix_norm1(&current_mb[2][0], uvstride, block_w>>1) - 2*cr*sum + cr*cr*block_s;
399  }else
400  cb = cr = 0;
401 
402  ic= s->c;
403  ic.bytestream_start=
404  ic.bytestream= i_buffer; //FIXME end/start? and at the other stoo
405  memcpy(i_state, s->block_state, sizeof(s->block_state));
406  if(level!=s->block_max_depth)
407  put_rac(&ic, &i_state[4 + s_context], 1);
408  put_rac(&ic, &i_state[1 + left->type + top->type], 1);
409  put_symbol(&ic, &i_state[32], l-pl , 1);
410  if (s->nb_planes > 2) {
411  put_symbol(&ic, &i_state[64], cb-pcb, 1);
412  put_symbol(&ic, &i_state[96], cr-pcr, 1);
413  }
414  i_len= ic.bytestream - ic.bytestream_start;
415  iscore += (s->lambda2*(get_rac_count(&ic)-base_bits))>>FF_LAMBDA_SHIFT;
416 
417  av_assert1(iscore < 255*255*256 + s->lambda2*10);
418  av_assert1(iscore >= 0);
419  av_assert1(l>=0 && l<=255);
420  av_assert1(pl>=0 && pl<=255);
421 
422  if(level==0){
423  int varc= iscore >> 8;
424  int vard= score >> 8;
425  if (vard <= 64 || vard < varc)
426  c->scene_change_score+= ff_sqrt(vard) - ff_sqrt(varc);
427  else
428  c->scene_change_score+= s->m.qscale;
429  }
430 
431  if(level!=s->block_max_depth){
432  put_rac(&s->c, &s->block_state[4 + s_context], 0);
433  score2 = encode_q_branch(s, level+1, 2*x+0, 2*y+0);
434  score2+= encode_q_branch(s, level+1, 2*x+1, 2*y+0);
435  score2+= encode_q_branch(s, level+1, 2*x+0, 2*y+1);
436  score2+= encode_q_branch(s, level+1, 2*x+1, 2*y+1);
437  score2+= s->lambda2>>FF_LAMBDA_SHIFT; //FIXME exact split overhead
438 
439  if(score2 < score && score2 < iscore)
440  return score2;
441  }
442 
443  if(iscore < score){
444  pred_mv(s, &pmx, &pmy, 0, left, top, tr);
445  memcpy(pbbak, i_buffer, i_len);
446  s->c= ic;
447  s->c.bytestream_start= pbbak_start;
448  s->c.bytestream= pbbak + i_len;
449  set_blocks(s, level, x, y, l, cb, cr, pmx, pmy, 0, BLOCK_INTRA);
450  memcpy(s->block_state, i_state, sizeof(s->block_state));
451  return iscore;
452  }else{
453  memcpy(pbbak, p_buffer, p_len);
454  s->c= pc;
455  s->c.bytestream_start= pbbak_start;
456  s->c.bytestream= pbbak + p_len;
457  set_blocks(s, level, x, y, pl, pcb, pcr, mx, my, best_ref, 0);
458  memcpy(s->block_state, p_state, sizeof(s->block_state));
459  return score;
460  }
461 }
462 
463 static void encode_q_branch2(SnowContext *s, int level, int x, int y){
464  const int w= s->b_width << s->block_max_depth;
465  const int rem_depth= s->block_max_depth - level;
466  const int index= (x + y*w) << rem_depth;
467  int trx= (x+1)<<rem_depth;
468  BlockNode *b= &s->block[index];
469  const BlockNode *left = x ? &s->block[index-1] : &null_block;
470  const BlockNode *top = y ? &s->block[index-w] : &null_block;
471  const BlockNode *tl = y && x ? &s->block[index-w-1] : left;
472  const BlockNode *tr = y && trx<w && ((x&1)==0 || level==0) ? &s->block[index-w+(1<<rem_depth)] : tl; //FIXME use lt
473  int pl = left->color[0];
474  int pcb= left->color[1];
475  int pcr= left->color[2];
476  int pmx, pmy;
477  int ref_context= av_log2(2*left->ref) + av_log2(2*top->ref);
478  int mx_context= av_log2(2*FFABS(left->mx - top->mx)) + 16*!!b->ref;
479  int my_context= av_log2(2*FFABS(left->my - top->my)) + 16*!!b->ref;
480  int s_context= 2*left->level + 2*top->level + tl->level + tr->level;
481 
482  if(s->keyframe){
483  set_blocks(s, level, x, y, pl, pcb, pcr, 0, 0, 0, BLOCK_INTRA);
484  return;
485  }
486 
487  if(level!=s->block_max_depth){
488  if(same_block(b,b+1) && same_block(b,b+w) && same_block(b,b+w+1)){
489  put_rac(&s->c, &s->block_state[4 + s_context], 1);
490  }else{
491  put_rac(&s->c, &s->block_state[4 + s_context], 0);
492  encode_q_branch2(s, level+1, 2*x+0, 2*y+0);
493  encode_q_branch2(s, level+1, 2*x+1, 2*y+0);
494  encode_q_branch2(s, level+1, 2*x+0, 2*y+1);
495  encode_q_branch2(s, level+1, 2*x+1, 2*y+1);
496  return;
497  }
498  }
499  if(b->type & BLOCK_INTRA){
500  pred_mv(s, &pmx, &pmy, 0, left, top, tr);
501  put_rac(&s->c, &s->block_state[1 + (left->type&1) + (top->type&1)], 1);
502  put_symbol(&s->c, &s->block_state[32], b->color[0]-pl , 1);
503  if (s->nb_planes > 2) {
504  put_symbol(&s->c, &s->block_state[64], b->color[1]-pcb, 1);
505  put_symbol(&s->c, &s->block_state[96], b->color[2]-pcr, 1);
506  }
507  set_blocks(s, level, x, y, b->color[0], b->color[1], b->color[2], pmx, pmy, 0, BLOCK_INTRA);
508  }else{
509  pred_mv(s, &pmx, &pmy, b->ref, left, top, tr);
510  put_rac(&s->c, &s->block_state[1 + (left->type&1) + (top->type&1)], 0);
511  if(s->ref_frames > 1)
512  put_symbol(&s->c, &s->block_state[128 + 1024 + 32*ref_context], b->ref, 0);
513  put_symbol(&s->c, &s->block_state[128 + 32*mx_context], b->mx - pmx, 1);
514  put_symbol(&s->c, &s->block_state[128 + 32*my_context], b->my - pmy, 1);
515  set_blocks(s, level, x, y, pl, pcb, pcr, b->mx, b->my, b->ref, 0);
516  }
517 }
518 
519 static int get_dc(SnowContext *s, int mb_x, int mb_y, int plane_index){
520  int i, x2, y2;
521  Plane *p= &s->plane[plane_index];
522  const int block_size = MB_SIZE >> s->block_max_depth;
523  const int block_w = plane_index ? block_size>>s->chroma_h_shift : block_size;
524  const int block_h = plane_index ? block_size>>s->chroma_v_shift : block_size;
525  const uint8_t *obmc = plane_index ? ff_obmc_tab[s->block_max_depth+s->chroma_h_shift] : ff_obmc_tab[s->block_max_depth];
526  const int obmc_stride= plane_index ? (2*block_size)>>s->chroma_h_shift : 2*block_size;
527  const int ref_stride= s->current_picture->linesize[plane_index];
528  uint8_t *src= s-> input_picture->data[plane_index];
529  IDWTELEM *dst= (IDWTELEM*)s->m.sc.obmc_scratchpad + plane_index*block_size*block_size*4; //FIXME change to unsigned
530  const int b_stride = s->b_width << s->block_max_depth;
531  const int w= p->width;
532  const int h= p->height;
533  int index= mb_x + mb_y*b_stride;
534  BlockNode *b= &s->block[index];
535  BlockNode backup= *b;
536  int ab=0;
537  int aa=0;
538 
539  av_assert2(s->chroma_h_shift == s->chroma_v_shift); //obmc stuff above
540 
541  b->type|= BLOCK_INTRA;
542  b->color[plane_index]= 0;
543  memset(dst, 0, obmc_stride*obmc_stride*sizeof(IDWTELEM));
544 
545  for(i=0; i<4; i++){
546  int mb_x2= mb_x + (i &1) - 1;
547  int mb_y2= mb_y + (i>>1) - 1;
548  int x= block_w*mb_x2 + block_w/2;
549  int y= block_h*mb_y2 + block_h/2;
550 
551  add_yblock(s, 0, NULL, dst + (i&1)*block_w + (i>>1)*obmc_stride*block_h, NULL, obmc,
552  x, y, block_w, block_h, w, h, obmc_stride, ref_stride, obmc_stride, mb_x2, mb_y2, 0, 0, plane_index);
553 
554  for(y2= FFMAX(y, 0); y2<FFMIN(h, y+block_h); y2++){
555  for(x2= FFMAX(x, 0); x2<FFMIN(w, x+block_w); x2++){
556  int index= x2-(block_w*mb_x - block_w/2) + (y2-(block_h*mb_y - block_h/2))*obmc_stride;
557  int obmc_v= obmc[index];
558  int d;
559  if(y<0) obmc_v += obmc[index + block_h*obmc_stride];
560  if(x<0) obmc_v += obmc[index + block_w];
561  if(y+block_h>h) obmc_v += obmc[index - block_h*obmc_stride];
562  if(x+block_w>w) obmc_v += obmc[index - block_w];
563  //FIXME precalculate this or simplify it somehow else
564 
565  d = -dst[index] + (1<<(FRAC_BITS-1));
566  dst[index] = d;
567  ab += (src[x2 + y2*ref_stride] - (d>>FRAC_BITS)) * obmc_v;
568  aa += obmc_v * obmc_v; //FIXME precalculate this
569  }
570  }
571  }
572  *b= backup;
573 
574  return av_clip_uint8( ROUNDED_DIV(ab<<LOG2_OBMC_MAX, aa) ); //FIXME we should not need clipping
575 }
576 
577 static inline int get_block_bits(SnowContext *s, int x, int y, int w){
578  const int b_stride = s->b_width << s->block_max_depth;
579  const int b_height = s->b_height<< s->block_max_depth;
580  int index= x + y*b_stride;
581  const BlockNode *b = &s->block[index];
582  const BlockNode *left = x ? &s->block[index-1] : &null_block;
583  const BlockNode *top = y ? &s->block[index-b_stride] : &null_block;
584  const BlockNode *tl = y && x ? &s->block[index-b_stride-1] : left;
585  const BlockNode *tr = y && x+w<b_stride ? &s->block[index-b_stride+w] : tl;
586  int dmx, dmy;
587 // int mx_context= av_log2(2*FFABS(left->mx - top->mx));
588 // int my_context= av_log2(2*FFABS(left->my - top->my));
589 
590  if(x<0 || x>=b_stride || y>=b_height)
591  return 0;
592 /*
593 1 0 0
594 01X 1-2 1
595 001XX 3-6 2-3
596 0001XXX 7-14 4-7
597 00001XXXX 15-30 8-15
598 */
599 //FIXME try accurate rate
600 //FIXME intra and inter predictors if surrounding blocks are not the same type
601  if(b->type & BLOCK_INTRA){
602  return 3+2*( av_log2(2*FFABS(left->color[0] - b->color[0]))
603  + av_log2(2*FFABS(left->color[1] - b->color[1]))
604  + av_log2(2*FFABS(left->color[2] - b->color[2])));
605  }else{
606  pred_mv(s, &dmx, &dmy, b->ref, left, top, tr);
607  dmx-= b->mx;
608  dmy-= b->my;
609  return 2*(1 + av_log2(2*FFABS(dmx)) //FIXME kill the 2* can be merged in lambda
610  + av_log2(2*FFABS(dmy))
611  + av_log2(2*b->ref));
612  }
613 }
614 
615 static int get_block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index, uint8_t (*obmc_edged)[MB_SIZE * 2]){
616  Plane *p= &s->plane[plane_index];
617  const int block_size = MB_SIZE >> s->block_max_depth;
618  const int block_w = plane_index ? block_size>>s->chroma_h_shift : block_size;
619  const int block_h = plane_index ? block_size>>s->chroma_v_shift : block_size;
620  const int obmc_stride= plane_index ? (2*block_size)>>s->chroma_h_shift : 2*block_size;
621  const int ref_stride= s->current_picture->linesize[plane_index];
622  uint8_t *dst= s->current_picture->data[plane_index];
623  uint8_t *src= s-> input_picture->data[plane_index];
624  IDWTELEM *pred= (IDWTELEM*)s->m.sc.obmc_scratchpad + plane_index*block_size*block_size*4;
625  uint8_t *cur = s->scratchbuf;
626  uint8_t *tmp = s->emu_edge_buffer;
627  const int b_stride = s->b_width << s->block_max_depth;
628  const int b_height = s->b_height<< s->block_max_depth;
629  const int w= p->width;
630  const int h= p->height;
631  int distortion;
632  int rate= 0;
633  const int penalty_factor= get_penalty_factor(s->lambda, s->lambda2, s->avctx->me_cmp);
634  int sx= block_w*mb_x - block_w/2;
635  int sy= block_h*mb_y - block_h/2;
636  int x0= FFMAX(0,-sx);
637  int y0= FFMAX(0,-sy);
638  int x1= FFMIN(block_w*2, w-sx);
639  int y1= FFMIN(block_h*2, h-sy);
640  int i,x,y;
641 
642  av_assert2(s->chroma_h_shift == s->chroma_v_shift); //obmc and square assumtions below chckinhg only block_w
643 
644  ff_snow_pred_block(s, cur, tmp, ref_stride, sx, sy, block_w*2, block_h*2, &s->block[mb_x + mb_y*b_stride], plane_index, w, h);
645 
646  for(y=y0; y<y1; y++){
647  const uint8_t *obmc1= obmc_edged[y];
648  const IDWTELEM *pred1 = pred + y*obmc_stride;
649  uint8_t *cur1 = cur + y*ref_stride;
650  uint8_t *dst1 = dst + sx + (sy+y)*ref_stride;
651  for(x=x0; x<x1; x++){
652 #if FRAC_BITS >= LOG2_OBMC_MAX
653  int v = (cur1[x] * obmc1[x]) << (FRAC_BITS - LOG2_OBMC_MAX);
654 #else
655  int v = (cur1[x] * obmc1[x] + (1<<(LOG2_OBMC_MAX - FRAC_BITS-1))) >> (LOG2_OBMC_MAX - FRAC_BITS);
656 #endif
657  v = (v + pred1[x]) >> FRAC_BITS;
658  if(v&(~255)) v= ~(v>>31);
659  dst1[x] = v;
660  }
661  }
662 
663  /* copy the regions where obmc[] = (uint8_t)256 */
664  if(LOG2_OBMC_MAX == 8
665  && (mb_x == 0 || mb_x == b_stride-1)
666  && (mb_y == 0 || mb_y == b_height-1)){
667  if(mb_x == 0)
668  x1 = block_w;
669  else
670  x0 = block_w;
671  if(mb_y == 0)
672  y1 = block_h;
673  else
674  y0 = block_h;
675  for(y=y0; y<y1; y++)
676  memcpy(dst + sx+x0 + (sy+y)*ref_stride, cur + x0 + y*ref_stride, x1-x0);
677  }
678 
679  if(block_w==16){
680  /* FIXME rearrange dsputil to fit 32x32 cmp functions */
681  /* FIXME check alignment of the cmp wavelet vs the encoding wavelet */
682  /* FIXME cmps overlap but do not cover the wavelet's whole support.
683  * So improving the score of one block is not strictly guaranteed
684  * to improve the score of the whole frame, thus iterative motion
685  * estimation does not always converge. */
686  if(s->avctx->me_cmp == FF_CMP_W97)
687  distortion = ff_w97_32_c(&s->m, src + sx + sy*ref_stride, dst + sx + sy*ref_stride, ref_stride, 32);
688  else if(s->avctx->me_cmp == FF_CMP_W53)
689  distortion = ff_w53_32_c(&s->m, src + sx + sy*ref_stride, dst + sx + sy*ref_stride, ref_stride, 32);
690  else{
691  distortion = 0;
692  for(i=0; i<4; i++){
693  int off = sx+16*(i&1) + (sy+16*(i>>1))*ref_stride;
694  distortion += s->mecc.me_cmp[0](&s->m, src + off, dst + off, ref_stride, 16);
695  }
696  }
697  }else{
698  av_assert2(block_w==8);
699  distortion = s->mecc.me_cmp[0](&s->m, src + sx + sy*ref_stride, dst + sx + sy*ref_stride, ref_stride, block_w*2);
700  }
701 
702  if(plane_index==0){
703  for(i=0; i<4; i++){
704 /* ..RRr
705  * .RXx.
706  * rxx..
707  */
708  rate += get_block_bits(s, mb_x + (i&1) - (i>>1), mb_y + (i>>1), 1);
709  }
710  if(mb_x == b_stride-2)
711  rate += get_block_bits(s, mb_x + 1, mb_y + 1, 1);
712  }
713  return distortion + rate*penalty_factor;
714 }
715 
716 static int get_4block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index){
717  int i, y2;
718  Plane *p= &s->plane[plane_index];
719  const int block_size = MB_SIZE >> s->block_max_depth;
720  const int block_w = plane_index ? block_size>>s->chroma_h_shift : block_size;
721  const int block_h = plane_index ? block_size>>s->chroma_v_shift : block_size;
722  const uint8_t *obmc = plane_index ? ff_obmc_tab[s->block_max_depth+s->chroma_h_shift] : ff_obmc_tab[s->block_max_depth];
723  const int obmc_stride= plane_index ? (2*block_size)>>s->chroma_h_shift : 2*block_size;
724  const int ref_stride= s->current_picture->linesize[plane_index];
725  uint8_t *dst= s->current_picture->data[plane_index];
726  uint8_t *src= s-> input_picture->data[plane_index];
727  //FIXME zero_dst is const but add_yblock changes dst if add is 0 (this is never the case for dst=zero_dst
728  // const has only been removed from zero_dst to suppress a warning
729  static IDWTELEM zero_dst[4096]; //FIXME
730  const int b_stride = s->b_width << s->block_max_depth;
731  const int w= p->width;
732  const int h= p->height;
733  int distortion= 0;
734  int rate= 0;
735  const int penalty_factor= get_penalty_factor(s->lambda, s->lambda2, s->avctx->me_cmp);
736 
737  av_assert2(s->chroma_h_shift == s->chroma_v_shift); //obmc and square assumtions below
738 
739  for(i=0; i<9; i++){
740  int mb_x2= mb_x + (i%3) - 1;
741  int mb_y2= mb_y + (i/3) - 1;
742  int x= block_w*mb_x2 + block_w/2;
743  int y= block_h*mb_y2 + block_h/2;
744 
745  add_yblock(s, 0, NULL, zero_dst, dst, obmc,
746  x, y, block_w, block_h, w, h, /*dst_stride*/0, ref_stride, obmc_stride, mb_x2, mb_y2, 1, 1, plane_index);
747 
748  //FIXME find a cleaner/simpler way to skip the outside stuff
749  for(y2= y; y2<0; y2++)
750  memcpy(dst + x + y2*ref_stride, src + x + y2*ref_stride, block_w);
751  for(y2= h; y2<y+block_h; y2++)
752  memcpy(dst + x + y2*ref_stride, src + x + y2*ref_stride, block_w);
753  if(x<0){
754  for(y2= y; y2<y+block_h; y2++)
755  memcpy(dst + x + y2*ref_stride, src + x + y2*ref_stride, -x);
756  }
757  if(x+block_w > w){
758  for(y2= y; y2<y+block_h; y2++)
759  memcpy(dst + w + y2*ref_stride, src + w + y2*ref_stride, x+block_w - w);
760  }
761 
762  av_assert1(block_w== 8 || block_w==16);
763  distortion += s->mecc.me_cmp[block_w==8](&s->m, src + x + y*ref_stride, dst + x + y*ref_stride, ref_stride, block_h);
764  }
765 
766  if(plane_index==0){
767  BlockNode *b= &s->block[mb_x+mb_y*b_stride];
768  int merged= same_block(b,b+1) && same_block(b,b+b_stride) && same_block(b,b+b_stride+1);
769 
770 /* ..RRRr
771  * .RXXx.
772  * .RXXx.
773  * rxxx.
774  */
775  if(merged)
776  rate = get_block_bits(s, mb_x, mb_y, 2);
777  for(i=merged?4:0; i<9; i++){
778  static const int dxy[9][2] = {{0,0},{1,0},{0,1},{1,1},{2,0},{2,1},{-1,2},{0,2},{1,2}};
779  rate += get_block_bits(s, mb_x + dxy[i][0], mb_y + dxy[i][1], 1);
780  }
781  }
782  return distortion + rate*penalty_factor;
783 }
784 
785 static int encode_subband_c0run(SnowContext *s, SubBand *b, const IDWTELEM *src, const IDWTELEM *parent, int stride, int orientation){
786  const int w= b->width;
787  const int h= b->height;
788  int x, y;
789 
790  if(1){
791  int run=0;
792  int *runs = s->run_buffer;
793  int run_index=0;
794  int max_index;
795 
796  for(y=0; y<h; y++){
797  for(x=0; x<w; x++){
798  int v, p=0;
799  int /*ll=0, */l=0, lt=0, t=0, rt=0;
800  v= src[x + y*stride];
801 
802  if(y){
803  t= src[x + (y-1)*stride];
804  if(x){
805  lt= src[x - 1 + (y-1)*stride];
806  }
807  if(x + 1 < w){
808  rt= src[x + 1 + (y-1)*stride];
809  }
810  }
811  if(x){
812  l= src[x - 1 + y*stride];
813  /*if(x > 1){
814  if(orientation==1) ll= src[y + (x-2)*stride];
815  else ll= src[x - 2 + y*stride];
816  }*/
817  }
818  if(parent){
819  int px= x>>1;
820  int py= y>>1;
821  if(px<b->parent->width && py<b->parent->height)
822  p= parent[px + py*2*stride];
823  }
824  if(!(/*ll|*/l|lt|t|rt|p)){
825  if(v){
826  runs[run_index++]= run;
827  run=0;
828  }else{
829  run++;
830  }
831  }
832  }
833  }
834  max_index= run_index;
835  runs[run_index++]= run;
836  run_index=0;
837  run= runs[run_index++];
838 
839  put_symbol2(&s->c, b->state[30], max_index, 0);
840  if(run_index <= max_index)
841  put_symbol2(&s->c, b->state[1], run, 3);
842 
843  for(y=0; y<h; y++){
844  if(s->c.bytestream_end - s->c.bytestream < w*40){
845  av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
846  return AVERROR(ENOMEM);
847  }
848  for(x=0; x<w; x++){
849  int v, p=0;
850  int /*ll=0, */l=0, lt=0, t=0, rt=0;
851  v= src[x + y*stride];
852 
853  if(y){
854  t= src[x + (y-1)*stride];
855  if(x){
856  lt= src[x - 1 + (y-1)*stride];
857  }
858  if(x + 1 < w){
859  rt= src[x + 1 + (y-1)*stride];
860  }
861  }
862  if(x){
863  l= src[x - 1 + y*stride];
864  /*if(x > 1){
865  if(orientation==1) ll= src[y + (x-2)*stride];
866  else ll= src[x - 2 + y*stride];
867  }*/
868  }
869  if(parent){
870  int px= x>>1;
871  int py= y>>1;
872  if(px<b->parent->width && py<b->parent->height)
873  p= parent[px + py*2*stride];
874  }
875  if(/*ll|*/l|lt|t|rt|p){
876  int context= av_log2(/*FFABS(ll) + */3*FFABS(l) + FFABS(lt) + 2*FFABS(t) + FFABS(rt) + FFABS(p));
877 
878  put_rac(&s->c, &b->state[0][context], !!v);
879  }else{
880  if(!run){
881  run= runs[run_index++];
882 
883  if(run_index <= max_index)
884  put_symbol2(&s->c, b->state[1], run, 3);
885  av_assert2(v);
886  }else{
887  run--;
888  av_assert2(!v);
889  }
890  }
891  if(v){
892  int context= av_log2(/*FFABS(ll) + */3*FFABS(l) + FFABS(lt) + 2*FFABS(t) + FFABS(rt) + FFABS(p));
893  int l2= 2*FFABS(l) + (l<0);
894  int t2= 2*FFABS(t) + (t<0);
895 
896  put_symbol2(&s->c, b->state[context + 2], FFABS(v)-1, context-4);
897  put_rac(&s->c, &b->state[0][16 + 1 + 3 + ff_quant3bA[l2&0xFF] + 3*ff_quant3bA[t2&0xFF]], v<0);
898  }
899  }
900  }
901  }
902  return 0;
903 }
904 
905 static int encode_subband(SnowContext *s, SubBand *b, const IDWTELEM *src, const IDWTELEM *parent, int stride, int orientation){
906 // encode_subband_qtree(s, b, src, parent, stride, orientation);
907 // encode_subband_z0run(s, b, src, parent, stride, orientation);
908  return encode_subband_c0run(s, b, src, parent, stride, orientation);
909 // encode_subband_dzr(s, b, src, parent, stride, orientation);
910 }
911 
912 static av_always_inline int check_block(SnowContext *s, int mb_x, int mb_y, int p[3], int intra, uint8_t (*obmc_edged)[MB_SIZE * 2], int *best_rd){
913  const int b_stride= s->b_width << s->block_max_depth;
914  BlockNode *block= &s->block[mb_x + mb_y * b_stride];
915  BlockNode backup= *block;
916  unsigned value;
917  int rd, index;
918 
919  av_assert2(mb_x>=0 && mb_y>=0);
920  av_assert2(mb_x<b_stride);
921 
922  if(intra){
923  block->color[0] = p[0];
924  block->color[1] = p[1];
925  block->color[2] = p[2];
926  block->type |= BLOCK_INTRA;
927  }else{
928  index= (p[0] + 31*p[1]) & (ME_CACHE_SIZE-1);
929  value= s->me_cache_generation + (p[0]>>10) + (p[1]<<6) + (block->ref<<12);
930  if(s->me_cache[index] == value)
931  return 0;
932  s->me_cache[index]= value;
933 
934  block->mx= p[0];
935  block->my= p[1];
936  block->type &= ~BLOCK_INTRA;
937  }
938 
939  rd= get_block_rd(s, mb_x, mb_y, 0, obmc_edged) + s->intra_penalty * !!intra;
940 
941 //FIXME chroma
942  if(rd < *best_rd){
943  *best_rd= rd;
944  return 1;
945  }else{
946  *block= backup;
947  return 0;
948  }
949 }
950 
951 /* special case for int[2] args we discard afterwards,
952  * fixes compilation problem with gcc 2.95 */
953 static av_always_inline int check_block_inter(SnowContext *s, int mb_x, int mb_y, int p0, int p1, uint8_t (*obmc_edged)[MB_SIZE * 2], int *best_rd){
954  int p[2] = {p0, p1};
955  return check_block(s, mb_x, mb_y, p, 0, obmc_edged, best_rd);
956 }
957 
958 static av_always_inline int check_4block_inter(SnowContext *s, int mb_x, int mb_y, int p0, int p1, int ref, int *best_rd){
959  const int b_stride= s->b_width << s->block_max_depth;
960  BlockNode *block= &s->block[mb_x + mb_y * b_stride];
961  BlockNode backup[4];
962  unsigned value;
963  int rd, index;
964 
965  /* We don't initialize backup[] during variable declaration, because
966  * that fails to compile on MSVC: "cannot convert from 'BlockNode' to
967  * 'int16_t'". */
968  backup[0] = block[0];
969  backup[1] = block[1];
970  backup[2] = block[b_stride];
971  backup[3] = block[b_stride + 1];
972 
973  av_assert2(mb_x>=0 && mb_y>=0);
974  av_assert2(mb_x<b_stride);
975  av_assert2(((mb_x|mb_y)&1) == 0);
976 
977  index= (p0 + 31*p1) & (ME_CACHE_SIZE-1);
978  value= s->me_cache_generation + (p0>>10) + (p1<<6) + (block->ref<<12);
979  if(s->me_cache[index] == value)
980  return 0;
981  s->me_cache[index]= value;
982 
983  block->mx= p0;
984  block->my= p1;
985  block->ref= ref;
986  block->type &= ~BLOCK_INTRA;
987  block[1]= block[b_stride]= block[b_stride+1]= *block;
988 
989  rd= get_4block_rd(s, mb_x, mb_y, 0);
990 
991 //FIXME chroma
992  if(rd < *best_rd){
993  *best_rd= rd;
994  return 1;
995  }else{
996  block[0]= backup[0];
997  block[1]= backup[1];
998  block[b_stride]= backup[2];
999  block[b_stride+1]= backup[3];
1000  return 0;
1001  }
1002 }
1003 
1005  int pass, mb_x, mb_y;
1006  const int b_width = s->b_width << s->block_max_depth;
1007  const int b_height= s->b_height << s->block_max_depth;
1008  const int b_stride= b_width;
1009  int color[3];
1010 
1011  {
1012  RangeCoder r = s->c;
1013  uint8_t state[sizeof(s->block_state)];
1014  memcpy(state, s->block_state, sizeof(s->block_state));
1015  for(mb_y= 0; mb_y<s->b_height; mb_y++)
1016  for(mb_x= 0; mb_x<s->b_width; mb_x++)
1017  encode_q_branch(s, 0, mb_x, mb_y);
1018  s->c = r;
1019  memcpy(s->block_state, state, sizeof(s->block_state));
1020  }
1021 
1022  for(pass=0; pass<25; pass++){
1023  int change= 0;
1024 
1025  for(mb_y= 0; mb_y<b_height; mb_y++){
1026  for(mb_x= 0; mb_x<b_width; mb_x++){
1027  int dia_change, i, j, ref;
1028  int best_rd= INT_MAX, ref_rd;
1029  BlockNode backup, ref_b;
1030  const int index= mb_x + mb_y * b_stride;
1031  BlockNode *block= &s->block[index];
1032  BlockNode *tb = mb_y ? &s->block[index-b_stride ] : NULL;
1033  BlockNode *lb = mb_x ? &s->block[index -1] : NULL;
1034  BlockNode *rb = mb_x+1<b_width ? &s->block[index +1] : NULL;
1035  BlockNode *bb = mb_y+1<b_height ? &s->block[index+b_stride ] : NULL;
1036  BlockNode *tlb= mb_x && mb_y ? &s->block[index-b_stride-1] : NULL;
1037  BlockNode *trb= mb_x+1<b_width && mb_y ? &s->block[index-b_stride+1] : NULL;
1038  BlockNode *blb= mb_x && mb_y+1<b_height ? &s->block[index+b_stride-1] : NULL;
1039  BlockNode *brb= mb_x+1<b_width && mb_y+1<b_height ? &s->block[index+b_stride+1] : NULL;
1040  const int b_w= (MB_SIZE >> s->block_max_depth);
1041  uint8_t obmc_edged[MB_SIZE * 2][MB_SIZE * 2];
1042 
1043  if(pass && (block->type & BLOCK_OPT))
1044  continue;
1045  block->type |= BLOCK_OPT;
1046 
1047  backup= *block;
1048 
1049  if(!s->me_cache_generation)
1050  memset(s->me_cache, 0, sizeof(s->me_cache));
1051  s->me_cache_generation += 1<<22;
1052 
1053  //FIXME precalculate
1054  {
1055  int x, y;
1056  for (y = 0; y < b_w * 2; y++)
1057  memcpy(obmc_edged[y], ff_obmc_tab[s->block_max_depth] + y * b_w * 2, b_w * 2);
1058  if(mb_x==0)
1059  for(y=0; y<b_w*2; y++)
1060  memset(obmc_edged[y], obmc_edged[y][0] + obmc_edged[y][b_w-1], b_w);
1061  if(mb_x==b_stride-1)
1062  for(y=0; y<b_w*2; y++)
1063  memset(obmc_edged[y]+b_w, obmc_edged[y][b_w] + obmc_edged[y][b_w*2-1], b_w);
1064  if(mb_y==0){
1065  for(x=0; x<b_w*2; x++)
1066  obmc_edged[0][x] += obmc_edged[b_w-1][x];
1067  for(y=1; y<b_w; y++)
1068  memcpy(obmc_edged[y], obmc_edged[0], b_w*2);
1069  }
1070  if(mb_y==b_height-1){
1071  for(x=0; x<b_w*2; x++)
1072  obmc_edged[b_w*2-1][x] += obmc_edged[b_w][x];
1073  for(y=b_w; y<b_w*2-1; y++)
1074  memcpy(obmc_edged[y], obmc_edged[b_w*2-1], b_w*2);
1075  }
1076  }
1077 
1078  //skip stuff outside the picture
1079  if(mb_x==0 || mb_y==0 || mb_x==b_width-1 || mb_y==b_height-1){
1080  uint8_t *src= s-> input_picture->data[0];
1081  uint8_t *dst= s->current_picture->data[0];
1082  const int stride= s->current_picture->linesize[0];
1083  const int block_w= MB_SIZE >> s->block_max_depth;
1084  const int block_h= MB_SIZE >> s->block_max_depth;
1085  const int sx= block_w*mb_x - block_w/2;
1086  const int sy= block_h*mb_y - block_h/2;
1087  const int w= s->plane[0].width;
1088  const int h= s->plane[0].height;
1089  int y;
1090 
1091  for(y=sy; y<0; y++)
1092  memcpy(dst + sx + y*stride, src + sx + y*stride, block_w*2);
1093  for(y=h; y<sy+block_h*2; y++)
1094  memcpy(dst + sx + y*stride, src + sx + y*stride, block_w*2);
1095  if(sx<0){
1096  for(y=sy; y<sy+block_h*2; y++)
1097  memcpy(dst + sx + y*stride, src + sx + y*stride, -sx);
1098  }
1099  if(sx+block_w*2 > w){
1100  for(y=sy; y<sy+block_h*2; y++)
1101  memcpy(dst + w + y*stride, src + w + y*stride, sx+block_w*2 - w);
1102  }
1103  }
1104 
1105  // intra(black) = neighbors' contribution to the current block
1106  for(i=0; i < s->nb_planes; i++)
1107  color[i]= get_dc(s, mb_x, mb_y, i);
1108 
1109  // get previous score (cannot be cached due to OBMC)
1110  if(pass > 0 && (block->type&BLOCK_INTRA)){
1111  int color0[3]= {block->color[0], block->color[1], block->color[2]};
1112  check_block(s, mb_x, mb_y, color0, 1, obmc_edged, &best_rd);
1113  }else
1114  check_block_inter(s, mb_x, mb_y, block->mx, block->my, obmc_edged, &best_rd);
1115 
1116  ref_b= *block;
1117  ref_rd= best_rd;
1118  for(ref=0; ref < s->ref_frames; ref++){
1119  int16_t (*mvr)[2]= &s->ref_mvs[ref][index];
1120  if(s->ref_scores[ref][index] > s->ref_scores[ref_b.ref][index]*3/2) //FIXME tune threshold
1121  continue;
1122  block->ref= ref;
1123  best_rd= INT_MAX;
1124 
1125  check_block_inter(s, mb_x, mb_y, mvr[0][0], mvr[0][1], obmc_edged, &best_rd);
1126  check_block_inter(s, mb_x, mb_y, 0, 0, obmc_edged, &best_rd);
1127  if(tb)
1128  check_block_inter(s, mb_x, mb_y, mvr[-b_stride][0], mvr[-b_stride][1], obmc_edged, &best_rd);
1129  if(lb)
1130  check_block_inter(s, mb_x, mb_y, mvr[-1][0], mvr[-1][1], obmc_edged, &best_rd);
1131  if(rb)
1132  check_block_inter(s, mb_x, mb_y, mvr[1][0], mvr[1][1], obmc_edged, &best_rd);
1133  if(bb)
1134  check_block_inter(s, mb_x, mb_y, mvr[b_stride][0], mvr[b_stride][1], obmc_edged, &best_rd);
1135 
1136  /* fullpel ME */
1137  //FIXME avoid subpel interpolation / round to nearest integer
1138  do{
1139  int newx = block->mx;
1140  int newy = block->my;
1141  int dia_size = s->iterative_dia_size ? s->iterative_dia_size : FFMAX(s->avctx->dia_size, 1);
1142  dia_change=0;
1143  for(i=0; i < dia_size; i++){
1144  for(j=0; j<i; j++){
1145  dia_change |= check_block_inter(s, mb_x, mb_y, newx+4*(i-j), newy+(4*j), obmc_edged, &best_rd);
1146  dia_change |= check_block_inter(s, mb_x, mb_y, newx-4*(i-j), newy-(4*j), obmc_edged, &best_rd);
1147  dia_change |= check_block_inter(s, mb_x, mb_y, newx-(4*j), newy+4*(i-j), obmc_edged, &best_rd);
1148  dia_change |= check_block_inter(s, mb_x, mb_y, newx+(4*j), newy-4*(i-j), obmc_edged, &best_rd);
1149  }
1150  }
1151  }while(dia_change);
1152  /* subpel ME */
1153  do{
1154  static const int square[8][2]= {{+1, 0},{-1, 0},{ 0,+1},{ 0,-1},{+1,+1},{-1,-1},{+1,-1},{-1,+1},};
1155  dia_change=0;
1156  for(i=0; i<8; i++)
1157  dia_change |= check_block_inter(s, mb_x, mb_y, block->mx+square[i][0], block->my+square[i][1], obmc_edged, &best_rd);
1158  }while(dia_change);
1159  //FIXME or try the standard 2 pass qpel or similar
1160 
1161  mvr[0][0]= block->mx;
1162  mvr[0][1]= block->my;
1163  if(ref_rd > best_rd){
1164  ref_rd= best_rd;
1165  ref_b= *block;
1166  }
1167  }
1168  best_rd= ref_rd;
1169  *block= ref_b;
1170  check_block(s, mb_x, mb_y, color, 1, obmc_edged, &best_rd);
1171  //FIXME RD style color selection
1172  if(!same_block(block, &backup)){
1173  if(tb ) tb ->type &= ~BLOCK_OPT;
1174  if(lb ) lb ->type &= ~BLOCK_OPT;
1175  if(rb ) rb ->type &= ~BLOCK_OPT;
1176  if(bb ) bb ->type &= ~BLOCK_OPT;
1177  if(tlb) tlb->type &= ~BLOCK_OPT;
1178  if(trb) trb->type &= ~BLOCK_OPT;
1179  if(blb) blb->type &= ~BLOCK_OPT;
1180  if(brb) brb->type &= ~BLOCK_OPT;
1181  change ++;
1182  }
1183  }
1184  }
1185  av_log(s->avctx, AV_LOG_DEBUG, "pass:%d changed:%d\n", pass, change);
1186  if(!change)
1187  break;
1188  }
1189 
1190  if(s->block_max_depth == 1){
1191  int change= 0;
1192  for(mb_y= 0; mb_y<b_height; mb_y+=2){
1193  for(mb_x= 0; mb_x<b_width; mb_x+=2){
1194  int i;
1195  int best_rd, init_rd;
1196  const int index= mb_x + mb_y * b_stride;
1197  BlockNode *b[4];
1198 
1199  b[0]= &s->block[index];
1200  b[1]= b[0]+1;
1201  b[2]= b[0]+b_stride;
1202  b[3]= b[2]+1;
1203  if(same_block(b[0], b[1]) &&
1204  same_block(b[0], b[2]) &&
1205  same_block(b[0], b[3]))
1206  continue;
1207 
1208  if(!s->me_cache_generation)
1209  memset(s->me_cache, 0, sizeof(s->me_cache));
1210  s->me_cache_generation += 1<<22;
1211 
1212  init_rd= best_rd= get_4block_rd(s, mb_x, mb_y, 0);
1213 
1214  //FIXME more multiref search?
1215  check_4block_inter(s, mb_x, mb_y,
1216  (b[0]->mx + b[1]->mx + b[2]->mx + b[3]->mx + 2) >> 2,
1217  (b[0]->my + b[1]->my + b[2]->my + b[3]->my + 2) >> 2, 0, &best_rd);
1218 
1219  for(i=0; i<4; i++)
1220  if(!(b[i]->type&BLOCK_INTRA))
1221  check_4block_inter(s, mb_x, mb_y, b[i]->mx, b[i]->my, b[i]->ref, &best_rd);
1222 
1223  if(init_rd != best_rd)
1224  change++;
1225  }
1226  }
1227  av_log(s->avctx, AV_LOG_ERROR, "pass:4mv changed:%d\n", change*4);
1228  }
1229 }
1230 
1231 static void encode_blocks(SnowContext *s, int search){
1232  int x, y;
1233  int w= s->b_width;
1234  int h= s->b_height;
1235 
1236  if(s->motion_est == FF_ME_ITER && !s->keyframe && search)
1237  iterative_me(s);
1238 
1239  for(y=0; y<h; y++){
1240  if(s->c.bytestream_end - s->c.bytestream < w*MB_SIZE*MB_SIZE*3){ //FIXME nicer limit
1241  av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
1242  return;
1243  }
1244  for(x=0; x<w; x++){
1245  if(s->motion_est == FF_ME_ITER || !search)
1246  encode_q_branch2(s, 0, x, y);
1247  else
1248  encode_q_branch (s, 0, x, y);
1249  }
1250  }
1251 }
1252 
1253 static void quantize(SnowContext *s, SubBand *b, IDWTELEM *dst, DWTELEM *src, int stride, int bias){
1254  const int w= b->width;
1255  const int h= b->height;
1256  const int qlog= av_clip(s->qlog + b->qlog, 0, QROOT*16);
1257  const int qmul= ff_qexp[qlog&(QROOT-1)]<<((qlog>>QSHIFT) + ENCODER_EXTRA_BITS);
1258  int x,y, thres1, thres2;
1259 
1260  if(s->qlog == LOSSLESS_QLOG){
1261  for(y=0; y<h; y++)
1262  for(x=0; x<w; x++)
1263  dst[x + y*stride]= src[x + y*stride];
1264  return;
1265  }
1266 
1267  bias= bias ? 0 : (3*qmul)>>3;
1268  thres1= ((qmul - bias)>>QEXPSHIFT) - 1;
1269  thres2= 2*thres1;
1270 
1271  if(!bias){
1272  for(y=0; y<h; y++){
1273  for(x=0; x<w; x++){
1274  int i= src[x + y*stride];
1275 
1276  if((unsigned)(i+thres1) > thres2){
1277  if(i>=0){
1278  i<<= QEXPSHIFT;
1279  i/= qmul; //FIXME optimize
1280  dst[x + y*stride]= i;
1281  }else{
1282  i= -i;
1283  i<<= QEXPSHIFT;
1284  i/= qmul; //FIXME optimize
1285  dst[x + y*stride]= -i;
1286  }
1287  }else
1288  dst[x + y*stride]= 0;
1289  }
1290  }
1291  }else{
1292  for(y=0; y<h; y++){
1293  for(x=0; x<w; x++){
1294  int i= src[x + y*stride];
1295 
1296  if((unsigned)(i+thres1) > thres2){
1297  if(i>=0){
1298  i<<= QEXPSHIFT;
1299  i= (i + bias) / qmul; //FIXME optimize
1300  dst[x + y*stride]= i;
1301  }else{
1302  i= -i;
1303  i<<= QEXPSHIFT;
1304  i= (i + bias) / qmul; //FIXME optimize
1305  dst[x + y*stride]= -i;
1306  }
1307  }else
1308  dst[x + y*stride]= 0;
1309  }
1310  }
1311  }
1312 }
1313 
1315  const int w= b->width;
1316  const int h= b->height;
1317  const int qlog= av_clip(s->qlog + b->qlog, 0, QROOT*16);
1318  const int qmul= ff_qexp[qlog&(QROOT-1)]<<(qlog>>QSHIFT);
1319  const int qadd= (s->qbias*qmul)>>QBIAS_SHIFT;
1320  int x,y;
1321 
1322  if(s->qlog == LOSSLESS_QLOG) return;
1323 
1324  for(y=0; y<h; y++){
1325  for(x=0; x<w; x++){
1326  int i= src[x + y*stride];
1327  if(i<0){
1328  src[x + y*stride]= -((-i*qmul + qadd)>>(QEXPSHIFT)); //FIXME try different bias
1329  }else if(i>0){
1330  src[x + y*stride]= (( i*qmul + qadd)>>(QEXPSHIFT));
1331  }
1332  }
1333  }
1334 }
1335 
1336 static void decorrelate(SnowContext *s, SubBand *b, IDWTELEM *src, int stride, int inverse, int use_median){
1337  const int w= b->width;
1338  const int h= b->height;
1339  int x,y;
1340 
1341  for(y=h-1; y>=0; y--){
1342  for(x=w-1; x>=0; x--){
1343  int i= x + y*stride;
1344 
1345  if(x){
1346  if(use_median){
1347  if(y && x+1<w) src[i] -= mid_pred(src[i - 1], src[i - stride], src[i - stride + 1]);
1348  else src[i] -= src[i - 1];
1349  }else{
1350  if(y) src[i] -= mid_pred(src[i - 1], src[i - stride], src[i - 1] + src[i - stride] - src[i - 1 - stride]);
1351  else src[i] -= src[i - 1];
1352  }
1353  }else{
1354  if(y) src[i] -= src[i - stride];
1355  }
1356  }
1357  }
1358 }
1359 
1360 static void correlate(SnowContext *s, SubBand *b, IDWTELEM *src, int stride, int inverse, int use_median){
1361  const int w= b->width;
1362  const int h= b->height;
1363  int x,y;
1364 
1365  for(y=0; y<h; y++){
1366  for(x=0; x<w; x++){
1367  int i= x + y*stride;
1368 
1369  if(x){
1370  if(use_median){
1371  if(y && x+1<w) src[i] += mid_pred(src[i - 1], src[i - stride], src[i - stride + 1]);
1372  else src[i] += src[i - 1];
1373  }else{
1374  if(y) src[i] += mid_pred(src[i - 1], src[i - stride], src[i - 1] + src[i - stride] - src[i - 1 - stride]);
1375  else src[i] += src[i - 1];
1376  }
1377  }else{
1378  if(y) src[i] += src[i - stride];
1379  }
1380  }
1381  }
1382 }
1383 
1385  int plane_index, level, orientation;
1386 
1387  for(plane_index=0; plane_index<FFMIN(s->nb_planes, 2); plane_index++){
1388  for(level=0; level<s->spatial_decomposition_count; level++){
1389  for(orientation=level ? 1:0; orientation<4; orientation++){
1390  if(orientation==2) continue;
1391  put_symbol(&s->c, s->header_state, s->plane[plane_index].band[level][orientation].qlog, 1);
1392  }
1393  }
1394  }
1395 }
1396 
1398  int plane_index, i;
1399  uint8_t kstate[32];
1400 
1401  memset(kstate, MID_STATE, sizeof(kstate));
1402 
1403  put_rac(&s->c, kstate, s->keyframe);
1404  if(s->keyframe || s->always_reset){
1406  s->last_spatial_decomposition_type=
1407  s->last_qlog=
1408  s->last_qbias=
1409  s->last_mv_scale=
1410  s->last_block_max_depth= 0;
1411  for(plane_index=0; plane_index<2; plane_index++){
1412  Plane *p= &s->plane[plane_index];
1413  p->last_htaps=0;
1414  p->last_diag_mc=0;
1415  memset(p->last_hcoeff, 0, sizeof(p->last_hcoeff));
1416  }
1417  }
1418  if(s->keyframe){
1419  put_symbol(&s->c, s->header_state, s->version, 0);
1420  put_rac(&s->c, s->header_state, s->always_reset);
1421  put_symbol(&s->c, s->header_state, s->temporal_decomposition_type, 0);
1422  put_symbol(&s->c, s->header_state, s->temporal_decomposition_count, 0);
1423  put_symbol(&s->c, s->header_state, s->spatial_decomposition_count, 0);
1424  put_symbol(&s->c, s->header_state, s->colorspace_type, 0);
1425  if (s->nb_planes > 2) {
1426  put_symbol(&s->c, s->header_state, s->chroma_h_shift, 0);
1427  put_symbol(&s->c, s->header_state, s->chroma_v_shift, 0);
1428  }
1429  put_rac(&s->c, s->header_state, s->spatial_scalability);
1430 // put_rac(&s->c, s->header_state, s->rate_scalability);
1431  put_symbol(&s->c, s->header_state, s->max_ref_frames-1, 0);
1432 
1433  encode_qlogs(s);
1434  }
1435 
1436  if(!s->keyframe){
1437  int update_mc=0;
1438  for(plane_index=0; plane_index<FFMIN(s->nb_planes, 2); plane_index++){
1439  Plane *p= &s->plane[plane_index];
1440  update_mc |= p->last_htaps != p->htaps;
1441  update_mc |= p->last_diag_mc != p->diag_mc;
1442  update_mc |= !!memcmp(p->last_hcoeff, p->hcoeff, sizeof(p->hcoeff));
1443  }
1444  put_rac(&s->c, s->header_state, update_mc);
1445  if(update_mc){
1446  for(plane_index=0; plane_index<FFMIN(s->nb_planes, 2); plane_index++){
1447  Plane *p= &s->plane[plane_index];
1448  put_rac(&s->c, s->header_state, p->diag_mc);
1449  put_symbol(&s->c, s->header_state, p->htaps/2-1, 0);
1450  for(i= p->htaps/2; i; i--)
1451  put_symbol(&s->c, s->header_state, FFABS(p->hcoeff[i]), 0);
1452  }
1453  }
1454  if(s->last_spatial_decomposition_count != s->spatial_decomposition_count){
1455  put_rac(&s->c, s->header_state, 1);
1456  put_symbol(&s->c, s->header_state, s->spatial_decomposition_count, 0);
1457  encode_qlogs(s);
1458  }else
1459  put_rac(&s->c, s->header_state, 0);
1460  }
1461 
1462  put_symbol(&s->c, s->header_state, s->spatial_decomposition_type - s->last_spatial_decomposition_type, 1);
1463  put_symbol(&s->c, s->header_state, s->qlog - s->last_qlog , 1);
1464  put_symbol(&s->c, s->header_state, s->mv_scale - s->last_mv_scale, 1);
1465  put_symbol(&s->c, s->header_state, s->qbias - s->last_qbias , 1);
1466  put_symbol(&s->c, s->header_state, s->block_max_depth - s->last_block_max_depth, 1);
1467 
1468 }
1469 
1471  int plane_index;
1472 
1473  if(!s->keyframe){
1474  for(plane_index=0; plane_index<2; plane_index++){
1475  Plane *p= &s->plane[plane_index];
1476  p->last_diag_mc= p->diag_mc;
1477  p->last_htaps = p->htaps;
1478  memcpy(p->last_hcoeff, p->hcoeff, sizeof(p->hcoeff));
1479  }
1480  }
1481 
1482  s->last_spatial_decomposition_type = s->spatial_decomposition_type;
1483  s->last_qlog = s->qlog;
1484  s->last_qbias = s->qbias;
1485  s->last_mv_scale = s->mv_scale;
1486  s->last_block_max_depth = s->block_max_depth;
1487  s->last_spatial_decomposition_count = s->spatial_decomposition_count;
1488 }
1489 
1490 static int qscale2qlog(int qscale){
1491  return lrint(QROOT*log2(qscale / (float)FF_QP2LAMBDA))
1492  + 61*QROOT/8; ///< 64 > 60
1493 }
1494 
1496 {
1497  /* Estimate the frame's complexity as a sum of weighted dwt coefficients.
1498  * FIXME we know exact mv bits at this point,
1499  * but ratecontrol isn't set up to include them. */
1500  uint32_t coef_sum= 0;
1501  int level, orientation, delta_qlog;
1502 
1503  for(level=0; level<s->spatial_decomposition_count; level++){
1504  for(orientation=level ? 1 : 0; orientation<4; orientation++){
1505  SubBand *b= &s->plane[0].band[level][orientation];
1506  IDWTELEM *buf= b->ibuf;
1507  const int w= b->width;
1508  const int h= b->height;
1509  const int stride= b->stride;
1510  const int qlog= av_clip(2*QROOT + b->qlog, 0, QROOT*16);
1511  const int qmul= ff_qexp[qlog&(QROOT-1)]<<(qlog>>QSHIFT);
1512  const int qdiv= (1<<16)/qmul;
1513  int x, y;
1514  //FIXME this is ugly
1515  for(y=0; y<h; y++)
1516  for(x=0; x<w; x++)
1517  buf[x+y*stride]= b->buf[x+y*stride];
1518  if(orientation==0)
1519  decorrelate(s, b, buf, stride, 1, 0);
1520  for(y=0; y<h; y++)
1521  for(x=0; x<w; x++)
1522  coef_sum+= abs(buf[x+y*stride]) * qdiv >> 16;
1523  }
1524  }
1525 
1526  /* ugly, ratecontrol just takes a sqrt again */
1527  av_assert0(coef_sum < INT_MAX);
1528  coef_sum = (uint64_t)coef_sum * coef_sum >> 16;
1529 
1530  if(pict->pict_type == AV_PICTURE_TYPE_I){
1531  s->m.current_picture.mb_var_sum= coef_sum;
1532  s->m.current_picture.mc_mb_var_sum= 0;
1533  }else{
1534  s->m.current_picture.mc_mb_var_sum= coef_sum;
1535  s->m.current_picture.mb_var_sum= 0;
1536  }
1537 
1538  pict->quality= ff_rate_estimate_qscale(&s->m, 1);
1539  if (pict->quality < 0)
1540  return INT_MIN;
1541  s->lambda= pict->quality * 3/2;
1542  delta_qlog= qscale2qlog(pict->quality) - s->qlog;
1543  s->qlog+= delta_qlog;
1544  return delta_qlog;
1545 }
1546 
1548  int width = p->width;
1549  int height= p->height;
1550  int level, orientation, x, y;
1551 
1552  for(level=0; level<s->spatial_decomposition_count; level++){
1553  int64_t error=0;
1554  for(orientation=level ? 1 : 0; orientation<4; orientation++){
1555  SubBand *b= &p->band[level][orientation];
1556  IDWTELEM *ibuf= b->ibuf;
1557 
1558  memset(s->spatial_idwt_buffer, 0, sizeof(*s->spatial_idwt_buffer)*width*height);
1559  ibuf[b->width/2 + b->height/2*b->stride]= 256*16;
1560  ff_spatial_idwt(s->spatial_idwt_buffer, s->temp_idwt_buffer, width, height, width, s->spatial_decomposition_type, s->spatial_decomposition_count);
1561  for(y=0; y<height; y++){
1562  for(x=0; x<width; x++){
1563  int64_t d= s->spatial_idwt_buffer[x + y*width]*16;
1564  error += d*d;
1565  }
1566  }
1567  if (orientation == 2)
1568  error /= 2;
1569  b->qlog= (int)(QROOT * log2(352256.0/sqrt(error)) + 0.5);
1570  if (orientation != 1)
1571  error = 0;
1572  }
1573  p->band[level][1].qlog = p->band[level][2].qlog;
1574  }
1575 }
1576 
1578  const AVFrame *pict, int *got_packet)
1579 {
1580  SnowContext *s = avctx->priv_data;
1581  RangeCoder * const c= &s->c;
1582  AVFrame *pic;
1583  const int width= s->avctx->width;
1584  const int height= s->avctx->height;
1585  int level, orientation, plane_index, i, y, ret;
1586  uint8_t rc_header_bak[sizeof(s->header_state)];
1587  uint8_t rc_block_bak[sizeof(s->block_state)];
1588 
1589  if ((ret = ff_alloc_packet2(avctx, pkt, s->b_width*s->b_height*MB_SIZE*MB_SIZE*3 + AV_INPUT_BUFFER_MIN_SIZE, 0)) < 0)
1590  return ret;
1591 
1593  ff_build_rac_states(c, (1LL<<32)/20, 256-8);
1594 
1595  for(i=0; i < s->nb_planes; i++){
1596  int hshift= i ? s->chroma_h_shift : 0;
1597  int vshift= i ? s->chroma_v_shift : 0;
1598  for(y=0; y<AV_CEIL_RSHIFT(height, vshift); y++)
1599  memcpy(&s->input_picture->data[i][y * s->input_picture->linesize[i]],
1600  &pict->data[i][y * pict->linesize[i]],
1601  AV_CEIL_RSHIFT(width, hshift));
1602  s->mpvencdsp.draw_edges(s->input_picture->data[i], s->input_picture->linesize[i],
1603  AV_CEIL_RSHIFT(width, hshift), AV_CEIL_RSHIFT(height, vshift),
1604  EDGE_WIDTH >> hshift, EDGE_WIDTH >> vshift,
1605  EDGE_TOP | EDGE_BOTTOM);
1606 
1607  }
1608  emms_c();
1609  pic = s->input_picture;
1610  pic->pict_type = pict->pict_type;
1611  pic->quality = pict->quality;
1612 
1613  s->m.picture_number= avctx->frame_number;
1614  if(avctx->flags&AV_CODEC_FLAG_PASS2){
1615  s->m.pict_type = pic->pict_type = s->m.rc_context.entry[avctx->frame_number].new_pict_type;
1616  s->keyframe = pic->pict_type == AV_PICTURE_TYPE_I;
1617  if(!(avctx->flags&AV_CODEC_FLAG_QSCALE)) {
1618  pic->quality = ff_rate_estimate_qscale(&s->m, 0);
1619  if (pic->quality < 0)
1620  return -1;
1621  }
1622  }else{
1623  s->keyframe= avctx->gop_size==0 || avctx->frame_number % avctx->gop_size == 0;
1624  s->m.pict_type = pic->pict_type = s->keyframe ? AV_PICTURE_TYPE_I : AV_PICTURE_TYPE_P;
1625  }
1626 
1627  if(s->pass1_rc && avctx->frame_number == 0)
1628  pic->quality = 2*FF_QP2LAMBDA;
1629  if (pic->quality) {
1630  s->qlog = qscale2qlog(pic->quality);
1631  s->lambda = pic->quality * 3/2;
1632  }
1633  if (s->qlog < 0 || (!pic->quality && (avctx->flags & AV_CODEC_FLAG_QSCALE))) {
1634  s->qlog= LOSSLESS_QLOG;
1635  s->lambda = 0;
1636  }//else keep previous frame's qlog until after motion estimation
1637 
1638 #if FF_API_CODED_FRAME
1640  av_frame_unref(avctx->coded_frame);
1642 #endif
1643 
1644  if (s->current_picture->data[0]) {
1645  int w = s->avctx->width;
1646  int h = s->avctx->height;
1647 
1648 #if FF_API_CODED_FRAME
1649  ret = av_frame_make_writable(s->current_picture);
1650  if (ret < 0)
1651  return ret;
1652 #endif
1653 
1654  s->mpvencdsp.draw_edges(s->current_picture->data[0],
1655  s->current_picture->linesize[0], w , h ,
1657  if (s->current_picture->data[2]) {
1658  s->mpvencdsp.draw_edges(s->current_picture->data[1],
1659  s->current_picture->linesize[1], w>>s->chroma_h_shift, h>>s->chroma_v_shift,
1660  EDGE_WIDTH>>s->chroma_h_shift, EDGE_WIDTH>>s->chroma_v_shift, EDGE_TOP | EDGE_BOTTOM);
1661  s->mpvencdsp.draw_edges(s->current_picture->data[2],
1662  s->current_picture->linesize[2], w>>s->chroma_h_shift, h>>s->chroma_v_shift,
1663  EDGE_WIDTH>>s->chroma_h_shift, EDGE_WIDTH>>s->chroma_v_shift, EDGE_TOP | EDGE_BOTTOM);
1664  }
1665  emms_c();
1666  }
1667 
1669 #if FF_API_CODED_FRAME
1671  ret = av_frame_ref(avctx->coded_frame, s->current_picture);
1673 #endif
1674  if (ret < 0)
1675  return ret;
1676 
1677  s->m.current_picture_ptr= &s->m.current_picture;
1678  s->m.current_picture.f = s->current_picture;
1679  s->m.current_picture.f->pts = pict->pts;
1680  if(pic->pict_type == AV_PICTURE_TYPE_P){
1681  int block_width = (width +15)>>4;
1682  int block_height= (height+15)>>4;
1683  int stride= s->current_picture->linesize[0];
1684 
1685  av_assert0(s->current_picture->data[0]);
1686  av_assert0(s->last_picture[0]->data[0]);
1687 
1688  s->m.avctx= s->avctx;
1689  s->m. last_picture.f = s->last_picture[0];
1690  s->m. new_picture.f = s->input_picture;
1691  s->m. last_picture_ptr= &s->m. last_picture;
1692  s->m.linesize = stride;
1693  s->m.uvlinesize= s->current_picture->linesize[1];
1694  s->m.width = width;
1695  s->m.height= height;
1696  s->m.mb_width = block_width;
1697  s->m.mb_height= block_height;
1698  s->m.mb_stride= s->m.mb_width+1;
1699  s->m.b8_stride= 2*s->m.mb_width+1;
1700  s->m.f_code=1;
1701  s->m.pict_type = pic->pict_type;
1702  s->m.motion_est= s->motion_est;
1703  s->m.me.scene_change_score=0;
1704  s->m.me.dia_size = avctx->dia_size;
1705  s->m.quarter_sample= (s->avctx->flags & AV_CODEC_FLAG_QPEL)!=0;
1706  s->m.out_format= FMT_H263;
1707  s->m.unrestricted_mv= 1;
1708 
1709  s->m.lambda = s->lambda;
1710  s->m.qscale= (s->m.lambda*139 + FF_LAMBDA_SCALE*64) >> (FF_LAMBDA_SHIFT + 7);
1711  s->lambda2= s->m.lambda2= (s->m.lambda*s->m.lambda + FF_LAMBDA_SCALE/2) >> FF_LAMBDA_SHIFT;
1712 
1713  s->m.mecc= s->mecc; //move
1714  s->m.qdsp= s->qdsp; //move
1715  s->m.hdsp = s->hdsp;
1716  ff_init_me(&s->m);
1717  s->hdsp = s->m.hdsp;
1718  s->mecc= s->m.mecc;
1719  }
1720 
1721  if(s->pass1_rc){
1722  memcpy(rc_header_bak, s->header_state, sizeof(s->header_state));
1723  memcpy(rc_block_bak, s->block_state, sizeof(s->block_state));
1724  }
1725 
1726 redo_frame:
1727 
1728  s->spatial_decomposition_count= 5;
1729 
1730  while( !(width >>(s->chroma_h_shift + s->spatial_decomposition_count))
1731  || !(height>>(s->chroma_v_shift + s->spatial_decomposition_count)))
1732  s->spatial_decomposition_count--;
1733 
1734  if (s->spatial_decomposition_count <= 0) {
1735  av_log(avctx, AV_LOG_ERROR, "Resolution too low\n");
1736  return AVERROR(EINVAL);
1737  }
1738 
1739  s->m.pict_type = pic->pict_type;
1740  s->qbias = pic->pict_type == AV_PICTURE_TYPE_P ? 2 : 0;
1741 
1743 
1744  if(s->last_spatial_decomposition_count != s->spatial_decomposition_count){
1745  for(plane_index=0; plane_index < s->nb_planes; plane_index++){
1746  calculate_visual_weight(s, &s->plane[plane_index]);
1747  }
1748  }
1749 
1750  encode_header(s);
1751  s->m.misc_bits = 8*(s->c.bytestream - s->c.bytestream_start);
1752  encode_blocks(s, 1);
1753  s->m.mv_bits = 8*(s->c.bytestream - s->c.bytestream_start) - s->m.misc_bits;
1754 
1755  for(plane_index=0; plane_index < s->nb_planes; plane_index++){
1756  Plane *p= &s->plane[plane_index];
1757  int w= p->width;
1758  int h= p->height;
1759  int x, y;
1760 // int bits= put_bits_count(&s->c.pb);
1761 
1762  if (!s->memc_only) {
1763  //FIXME optimize
1764  if(pict->data[plane_index]) //FIXME gray hack
1765  for(y=0; y<h; y++){
1766  for(x=0; x<w; x++){
1767  s->spatial_idwt_buffer[y*w + x]= pict->data[plane_index][y*pict->linesize[plane_index] + x]<<FRAC_BITS;
1768  }
1769  }
1770  predict_plane(s, s->spatial_idwt_buffer, plane_index, 0);
1771 
1772 #if FF_API_PRIVATE_OPT
1774  if(s->avctx->scenechange_threshold)
1775  s->scenechange_threshold = s->avctx->scenechange_threshold;
1777 #endif
1778 
1779  if( plane_index==0
1780  && pic->pict_type == AV_PICTURE_TYPE_P
1781  && !(avctx->flags&AV_CODEC_FLAG_PASS2)
1782  && s->m.me.scene_change_score > s->scenechange_threshold){
1784  ff_build_rac_states(c, (1LL<<32)/20, 256-8);
1786  s->keyframe=1;
1787  s->current_picture->key_frame=1;
1788  goto redo_frame;
1789  }
1790 
1791  if(s->qlog == LOSSLESS_QLOG){
1792  for(y=0; y<h; y++){
1793  for(x=0; x<w; x++){
1794  s->spatial_dwt_buffer[y*w + x]= (s->spatial_idwt_buffer[y*w + x] + (1<<(FRAC_BITS-1))-1)>>FRAC_BITS;
1795  }
1796  }
1797  }else{
1798  for(y=0; y<h; y++){
1799  for(x=0; x<w; x++){
1800  s->spatial_dwt_buffer[y*w + x]= s->spatial_idwt_buffer[y*w + x] * (1 << ENCODER_EXTRA_BITS);
1801  }
1802  }
1803  }
1804 
1805  ff_spatial_dwt(s->spatial_dwt_buffer, s->temp_dwt_buffer, w, h, w, s->spatial_decomposition_type, s->spatial_decomposition_count);
1806 
1807  if(s->pass1_rc && plane_index==0){
1808  int delta_qlog = ratecontrol_1pass(s, pic);
1809  if (delta_qlog <= INT_MIN)
1810  return -1;
1811  if(delta_qlog){
1812  //reordering qlog in the bitstream would eliminate this reset
1814  memcpy(s->header_state, rc_header_bak, sizeof(s->header_state));
1815  memcpy(s->block_state, rc_block_bak, sizeof(s->block_state));
1816  encode_header(s);
1817  encode_blocks(s, 0);
1818  }
1819  }
1820 
1821  for(level=0; level<s->spatial_decomposition_count; level++){
1822  for(orientation=level ? 1 : 0; orientation<4; orientation++){
1823  SubBand *b= &p->band[level][orientation];
1824 
1825  quantize(s, b, b->ibuf, b->buf, b->stride, s->qbias);
1826  if(orientation==0)
1827  decorrelate(s, b, b->ibuf, b->stride, pic->pict_type == AV_PICTURE_TYPE_P, 0);
1828  if (!s->no_bitstream)
1829  encode_subband(s, b, b->ibuf, b->parent ? b->parent->ibuf : NULL, b->stride, orientation);
1830  av_assert0(b->parent==NULL || b->parent->stride == b->stride*2);
1831  if(orientation==0)
1832  correlate(s, b, b->ibuf, b->stride, 1, 0);
1833  }
1834  }
1835 
1836  for(level=0; level<s->spatial_decomposition_count; level++){
1837  for(orientation=level ? 1 : 0; orientation<4; orientation++){
1838  SubBand *b= &p->band[level][orientation];
1839 
1840  dequantize(s, b, b->ibuf, b->stride);
1841  }
1842  }
1843 
1844  ff_spatial_idwt(s->spatial_idwt_buffer, s->temp_idwt_buffer, w, h, w, s->spatial_decomposition_type, s->spatial_decomposition_count);
1845  if(s->qlog == LOSSLESS_QLOG){
1846  for(y=0; y<h; y++){
1847  for(x=0; x<w; x++){
1848  s->spatial_idwt_buffer[y*w + x]<<=FRAC_BITS;
1849  }
1850  }
1851  }
1852  predict_plane(s, s->spatial_idwt_buffer, plane_index, 1);
1853  }else{
1854  //ME/MC only
1855  if(pic->pict_type == AV_PICTURE_TYPE_I){
1856  for(y=0; y<h; y++){
1857  for(x=0; x<w; x++){
1858  s->current_picture->data[plane_index][y*s->current_picture->linesize[plane_index] + x]=
1859  pict->data[plane_index][y*pict->linesize[plane_index] + x];
1860  }
1861  }
1862  }else{
1863  memset(s->spatial_idwt_buffer, 0, sizeof(IDWTELEM)*w*h);
1864  predict_plane(s, s->spatial_idwt_buffer, plane_index, 1);
1865  }
1866  }
1867  if(s->avctx->flags&AV_CODEC_FLAG_PSNR){
1868  int64_t error= 0;
1869 
1870  if(pict->data[plane_index]) //FIXME gray hack
1871  for(y=0; y<h; y++){
1872  for(x=0; x<w; x++){
1873  int d= s->current_picture->data[plane_index][y*s->current_picture->linesize[plane_index] + x] - pict->data[plane_index][y*pict->linesize[plane_index] + x];
1874  error += d*d;
1875  }
1876  }
1877  s->avctx->error[plane_index] += error;
1878  s->encoding_error[plane_index] = error;
1879  }
1880 
1881  }
1882  emms_c();
1883 
1885 
1886  ff_snow_release_buffer(avctx);
1887 
1888  s->current_picture->coded_picture_number = avctx->frame_number;
1889  s->current_picture->pict_type = pic->pict_type;
1890  s->current_picture->quality = pic->quality;
1891  s->m.frame_bits = 8*(s->c.bytestream - s->c.bytestream_start);
1892  s->m.p_tex_bits = s->m.frame_bits - s->m.misc_bits - s->m.mv_bits;
1893  s->m.current_picture.f->display_picture_number =
1894  s->m.current_picture.f->coded_picture_number = avctx->frame_number;
1895  s->m.current_picture.f->quality = pic->quality;
1896  s->m.total_bits += 8*(s->c.bytestream - s->c.bytestream_start);
1897  if(s->pass1_rc)
1898  if (ff_rate_estimate_qscale(&s->m, 0) < 0)
1899  return -1;
1900  if(avctx->flags&AV_CODEC_FLAG_PASS1)
1901  ff_write_pass1_stats(&s->m);
1902  s->m.last_pict_type = s->m.pict_type;
1903 #if FF_API_STAT_BITS
1905  avctx->frame_bits = s->m.frame_bits;
1906  avctx->mv_bits = s->m.mv_bits;
1907  avctx->misc_bits = s->m.misc_bits;
1908  avctx->p_tex_bits = s->m.p_tex_bits;
1910 #endif
1911 
1912  emms_c();
1913 
1914  ff_side_data_set_encoder_stats(pkt, s->current_picture->quality,
1915  s->encoding_error,
1916  (s->avctx->flags&AV_CODEC_FLAG_PSNR) ? 4 : 0,
1917  s->current_picture->pict_type);
1918 
1919 #if FF_API_ERROR_FRAME
1921  memcpy(s->current_picture->error, s->encoding_error, sizeof(s->encoding_error));
1923 #endif
1924 
1925  pkt->size = ff_rac_terminate(c, 0);
1926  if (s->current_picture->key_frame)
1928  *got_packet = 1;
1929 
1930  return 0;
1931 }
1932 
1934 {
1935  SnowContext *s = avctx->priv_data;
1936 
1939  av_frame_free(&s->input_picture);
1940  av_freep(&avctx->stats_out);
1941 
1942  return 0;
1943 }
1944 
1945 #define OFFSET(x) offsetof(SnowContext, x)
1946 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
1947 static const AVOption options[] = {
1948  {"motion_est", "motion estimation algorithm", OFFSET(motion_est), AV_OPT_TYPE_INT, {.i64 = FF_ME_EPZS }, FF_ME_ZERO, FF_ME_ITER, VE, "motion_est" },
1949  { "zero", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_ZERO }, 0, 0, VE, "motion_est" },
1950  { "epzs", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_EPZS }, 0, 0, VE, "motion_est" },
1951  { "xone", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_XONE }, 0, 0, VE, "motion_est" },
1952  { "iter", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = FF_ME_ITER }, 0, 0, VE, "motion_est" },
1953  { "memc_only", "Only do ME/MC (I frames -> ref, P frame -> ME+MC).", OFFSET(memc_only), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
1954  { "no_bitstream", "Skip final bitstream writeout.", OFFSET(no_bitstream), AV_OPT_TYPE_BOOL, { .i64 = 0 }, 0, 1, VE },
1955  { "intra_penalty", "Penalty for intra blocks in block decission", OFFSET(intra_penalty), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, VE },
1956  { "iterative_dia_size", "Dia size for the iterative ME", OFFSET(iterative_dia_size), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, VE },
1957  { "sc_threshold", "Scene change threshold", OFFSET(scenechange_threshold), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, VE },
1958  { "pred", "Spatial decomposition type", OFFSET(pred), AV_OPT_TYPE_INT, { .i64 = 0 }, DWT_97, DWT_53, VE, "pred" },
1959  { "dwt97", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = 0 }, INT_MIN, INT_MAX, VE, "pred" },
1960  { "dwt53", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = 1 }, INT_MIN, INT_MAX, VE, "pred" },
1961  { "rc_eq", "Set rate control equation. When computing the expression, besides the standard functions "
1962  "defined in the section 'Expression Evaluation', the following functions are available: "
1963  "bits2qp(bits), qp2bits(qp). Also the following constants are available: iTex pTex tex mv "
1964  "fCode iCount mcVar var isI isP isB avgQP qComp avgIITex avgPITex avgPPTex avgBPTex avgTex.",
1965  OFFSET(m.rc_eq), AV_OPT_TYPE_STRING, { .str = NULL }, 0, 0, VE },
1966  { NULL },
1967 };
1968 
1969 static const AVClass snowenc_class = {
1970  .class_name = "snow encoder",
1971  .item_name = av_default_item_name,
1972  .option = options,
1973  .version = LIBAVUTIL_VERSION_INT,
1974 };
1975 
1977  .name = "snow",
1978  .long_name = NULL_IF_CONFIG_SMALL("Snow"),
1979  .type = AVMEDIA_TYPE_VIDEO,
1980  .id = AV_CODEC_ID_SNOW,
1981  .priv_data_size = sizeof(SnowContext),
1982  .init = encode_init,
1983  .encode2 = encode_frame,
1984  .close = encode_end,
1985  .pix_fmts = (const enum AVPixelFormat[]){
1989  },
1990  .priv_class = &snowenc_class,
1991  .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE |
1993 };
static uint32_t inverse(uint32_t v)
find multiplicative inverse modulo 2 ^ 32
Definition: asfcrypt.c:35
#define av_always_inline
Definition: attributes.h:45
#define av_cold
Definition: attributes.h:88
uint8_t
#define av_assert2(cond)
assert() equivalent, that does lie in speed critical code.
Definition: avassert.h:64
#define av_assert1(cond)
assert() equivalent, that does not lie in speed critical code.
Definition: avassert.h:53
#define av_assert0(cond)
assert() equivalent, that is always enabled.
Definition: avassert.h:37
Libavcodec external API header.
#define FF_CMP_DCT264
Definition: avcodec.h:955
#define FF_CMP_W53
Definition: avcodec.h:952
#define FF_CMP_SSE
Definition: avcodec.h:942
#define FF_CMP_DCT
Definition: avcodec.h:944
#define FF_CMP_W97
Definition: avcodec.h:953
#define FF_CMP_BIT
Definition: avcodec.h:946
#define FF_CMP_SATD
Definition: avcodec.h:943
#define FF_CMP_NSSE
Definition: avcodec.h:951
#define FF_CMP_SAD
Definition: avcodec.h:941
#define FF_CMP_PSNR
Definition: avcodec.h:945
#define FF_CMP_RD
Definition: avcodec.h:947
int ff_side_data_set_encoder_stats(AVPacket *pkt, int quality, int64_t *error, int error_count, int pict_type)
Definition: avpacket.c:820
static av_cold int init(AVCodecContext *avctx)
Definition: avrndec.c:31
#define s(width, name)
Definition: cbs_vp9.c:257
static struct @321 state
#define MB_SIZE
Definition: cinepakenc.c:54
#define FFMIN(a, b)
Definition: common.h:105
#define AV_CEIL_RSHIFT(a, b)
Definition: common.h:58
#define av_clip
Definition: common.h:122
#define ROUNDED_DIV(a, b)
Definition: common.h:56
#define FFMAX(a, b)
Definition: common.h:103
#define av_clip_uint8
Definition: common.h:128
#define FFABS(a)
Absolute value, Note, INT_MIN / INT64_MIN result in undefined behavior as they are not representable ...
Definition: common.h:72
#define NULL
Definition: coverity.c:32
long long int64_t
Definition: coverity.c:34
#define abs(x)
Definition: cuda_runtime.h:35
int DWTELEM
Definition: dirac_dwt.h:26
short IDWTELEM
Definition: dirac_dwt.h:27
int ff_alloc_packet2(AVCodecContext *avctx, AVPacket *avpkt, int64_t size, int64_t min_size)
Check AVPacket size and/or allocate data.
Definition: encode.c:33
double value
Definition: eval.c:98
int
static av_noinline void put_symbol(RangeCoder *c, uint8_t *state, int v, int is_signed)
Definition: ffv1enc.c:232
#define put_rac(C, S, B)
#define FRAC_BITS
@ AV_OPT_TYPE_CONST
Definition: opt.h:234
@ AV_OPT_TYPE_INT
Definition: opt.h:225
@ AV_OPT_TYPE_BOOL
Definition: opt.h:242
@ AV_OPT_TYPE_STRING
Definition: opt.h:229
#define AV_CODEC_FLAG_QPEL
Use qpel MC.
Definition: avcodec.h:287
#define AV_CODEC_FLAG_PASS2
Use internal 2pass ratecontrol in second pass mode.
Definition: avcodec.h:300
#define AV_CODEC_FLAG_QSCALE
Use fixed qscale.
Definition: avcodec.h:275
#define AV_CODEC_FLAG_PASS1
Use internal 2pass ratecontrol in first pass mode.
Definition: avcodec.h:296
#define AV_CODEC_FLAG_PSNR
error[?] variables will be set during encoding.
Definition: avcodec.h:312
#define AV_CODEC_FLAG_4MV
4 MV per MB allowed / advanced prediction for H.263.
Definition: avcodec.h:279
@ AV_CODEC_ID_SNOW
Definition: codec_id.h:262
#define AV_INPUT_BUFFER_MIN_SIZE
minimum encoding buffer size Used to avoid some checks during header writing.
Definition: avcodec.h:222
#define AV_PKT_FLAG_KEY
The packet contains a keyframe.
Definition: packet.h:410
#define FF_LAMBDA_SCALE
Definition: avutil.h:226
#define FF_LAMBDA_SHIFT
Definition: avutil.h:225
#define FF_QP2LAMBDA
factor to convert from H.263 QP to lambda
Definition: avutil.h:227
#define AVERROR_PATCHWELCOME
Not yet implemented in FFmpeg, patches welcome.
Definition: error.h:62
#define AVERROR(e)
Definition: error.h:43
void av_frame_unref(AVFrame *frame)
Unreference all the buffers referenced by frame and reset the frame fields.
Definition: frame.c:553
int av_frame_ref(AVFrame *dst, const AVFrame *src)
Set up a new reference to the data described by the source frame.
Definition: frame.c:443
void av_frame_free(AVFrame **frame)
Free the frame and any dynamically allocated objects in it, e.g.
Definition: frame.c:203
AVFrame * av_frame_alloc(void)
Allocate an AVFrame and set its fields to default values.
Definition: frame.c:190
int av_frame_make_writable(AVFrame *frame)
Ensure that the frame data is writable, avoiding data copy if possible.
Definition: frame.c:611
#define AV_LOG_DEBUG
Stuff which is only useful for libav* developers.
Definition: log.h:215
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
Definition: log.h:194
const char * av_default_item_name(void *ptr)
Return the context name.
Definition: log.c:235
void * av_mallocz(size_t size)
Allocate a memory block with alignment suitable for all memory accesses (including vectors if availab...
Definition: mem.c:237
void * av_mallocz_array(size_t nmemb, size_t size)
Allocate a memory block for an array with av_mallocz().
Definition: mem.c:190
@ AVMEDIA_TYPE_VIDEO
Definition: avutil.h:201
@ AV_PICTURE_TYPE_I
Intra.
Definition: avutil.h:274
@ AV_PICTURE_TYPE_P
Predicted.
Definition: avutil.h:275
#define LIBAVUTIL_VERSION_INT
Definition: version.h:85
int index
Definition: gxfenc.c:89
void ff_h263_encode_init(MpegEncContext *s)
Definition: ituh263enc.c:757
cl_device_type type
int i
Definition: input.c:407
#define av_log2
Definition: intmath.h:83
static void pred_mv(DiracBlock *block, int stride, int x, int y, int ref)
Definition: diracdec.c:1390
#define FF_CODEC_CAP_INIT_THREADSAFE
The codec does not modify any global variables in the init function, allowing to call the init functi...
Definition: internal.h:41
#define FF_CODEC_CAP_INIT_CLEANUP
The codec allows calling the close function for deallocation even if the init function returned a fai...
Definition: internal.h:49
common internal API header
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification.
Definition: internal.h:117
#define FF_DISABLE_DEPRECATION_WARNINGS
Definition: internal.h:83
#define FF_ENABLE_DEPRECATION_WARNINGS
Definition: internal.h:84
#define emms_c()
Definition: internal.h:54
static enum AVPixelFormat pix_fmts[]
Definition: libkvazaar.c:309
Replacements for frequently missing libm functions.
#define log2(x)
Definition: libm.h:404
uint8_t w
Definition: llviddspenc.c:39
int stride
Definition: mace.c:144
#define mid_pred
Definition: mathops.h:97
#define ff_sqrt
Definition: mathops.h:206
const uint32_t ff_square_tab[512]
Definition: me_cmp.c:34
void ff_set_cmp(MECmpContext *c, me_cmp_func *cmp, int type)
Definition: me_cmp.c:475
int ff_init_me(MpegEncContext *s)
Definition: motion_est.c:306
static void init_ref(MotionEstContext *c, uint8_t *src[3], uint8_t *ref[3], uint8_t *ref2[3], int x, int y, int ref_index)
Definition: motion_est.c:83
#define MAX_MV
Definition: motion_est.h:35
#define ME_MAP_SIZE
Definition: motion_est.h:38
int ff_epzs_motion_search(struct MpegEncContext *s, int *mx_ptr, int *my_ptr, int P[10][2], int src_index, int ref_index, int16_t(*last_mv)[2], int ref_mv_scale, int size, int h)
#define FF_ME_EPZS
Definition: motion_est.h:41
#define FF_ME_XONE
Definition: motion_est.h:42
int ff_get_mb_score(struct MpegEncContext *s, int mx, int my, int src_index, int ref_index, int size, int h, int add_rate)
#define FF_ME_ZERO
Definition: motion_est.h:40
#define MAX_DMV
Definition: motion_est.h:37
#define P
#define EDGE_WIDTH
Definition: mpegpicture.h:33
@ FMT_H263
Definition: mpegutils.h:126
mpegvideo header.
av_cold void ff_mpegvideoencdsp_init(MpegvideoEncDSPContext *c, AVCodecContext *avctx)
#define EDGE_BOTTOM
#define EDGE_TOP
AVOptions.
int av_pix_fmt_get_chroma_sub_sample(enum AVPixelFormat pix_fmt, int *h_shift, int *v_shift)
Utility function to access log2_chroma_w log2_chroma_h from the pixel format AVPixFmtDescriptor.
Definition: pixdesc.c:2601
AVPixelFormat
Pixel format.
Definition: pixfmt.h:64
@ AV_PIX_FMT_NONE
Definition: pixfmt.h:65
@ AV_PIX_FMT_YUV420P
planar YUV 4:2:0, 12bpp, (1 Cr & Cb sample per 2x2 Y samples)
Definition: pixfmt.h:66
@ AV_PIX_FMT_GRAY8
Y , 8bpp.
Definition: pixfmt.h:74
@ AV_PIX_FMT_YUV410P
planar YUV 4:1:0, 9bpp, (1 Cr & Cb sample per 4x4 Y samples)
Definition: pixfmt.h:72
@ AV_PIX_FMT_YUV444P
planar YUV 4:4:4, 24bpp, (1 Cr & Cb sample per 1x1 Y samples)
Definition: pixfmt.h:71
int ff_rac_terminate(RangeCoder *c, int version)
Terminates the range coder.
Definition: rangecoder.c:109
void ff_build_rac_states(RangeCoder *c, int factor, int max_p)
Definition: rangecoder.c:68
av_cold void ff_init_range_encoder(RangeCoder *c, uint8_t *buf, int buf_size)
Definition: rangecoder.c:42
Range coder.
static int get_rac_count(RangeCoder *c)
Definition: rangecoder.h:85
av_cold int ff_rate_control_init(MpegEncContext *s)
Definition: ratecontrol.c:472
av_cold void ff_rate_control_uninit(MpegEncContext *s)
Definition: ratecontrol.c:672
float ff_rate_estimate_qscale(MpegEncContext *s, int dry_run)
Definition: ratecontrol.c:868
void ff_write_pass1_stats(MpegEncContext *s)
Definition: ratecontrol.c:38
#define tb
Definition: regdef.h:68
#define t2
Definition: regdef.h:30
static int square(int x)
Definition: roqvideoenc.c:193
static const float pred[4]
Definition: siprdata.h:259
av_cold int ff_snow_common_init(AVCodecContext *avctx)
Definition: snow.c:439
void ff_snow_pred_block(SnowContext *s, uint8_t *dst, uint8_t *tmp, ptrdiff_t stride, int sx, int sy, int b_w, int b_h, const BlockNode *block, int plane_index, int w, int h)
Definition: snow.c:328
int ff_snow_get_buffer(SnowContext *s, AVFrame *frame)
Definition: snow.c:70
void ff_snow_release_buffer(AVCodecContext *avctx)
Definition: snow.c:646
void ff_snow_reset_contexts(SnowContext *s)
Definition: snow.c:97
int ff_snow_common_init_after_header(AVCodecContext *avctx)
Definition: snow.c:521
int ff_snow_frame_start(SnowContext *s)
Definition: snow.c:661
av_cold void ff_snow_common_end(SnowContext *s)
Definition: snow.c:699
int ff_snow_alloc_blocks(SnowContext *s)
Definition: snow.c:111
#define BLOCK_OPT
Block needs no checks in this round of iterative motion estiation.
Definition: snow.h:59
#define LOG2_MB_SIZE
Definition: snow.h:73
#define QROOT
Definition: snow.h:44
static av_always_inline void predict_plane(SnowContext *s, IDWTELEM *buf, int plane_index, int add)
Definition: snow.h:456
#define MID_STATE
Definition: snow.h:40
#define QSHIFT
Definition: snow.h:43
#define FF_ME_ITER
Definition: snow.h:38
#define QEXPSHIFT
Definition: snow.h:508
static void set_blocks(SnowContext *s, int level, int x, int y, int l, int cb, int cr, int mx, int my, int ref, int type)
Definition: snow.h:463
const uint8_t *const ff_obmc_tab[4]
Definition: snowdata.h:123
static void put_symbol2(RangeCoder *c, uint8_t *state, int v, int log2)
Definition: snow.h:564
uint8_t ff_qexp[QROOT]
Definition: snowdata.h:128
#define ENCODER_EXTRA_BITS
Definition: snow.h:75
#define BLOCK_INTRA
Intra block, inter otherwise.
Definition: snow.h:58
#define LOSSLESS_QLOG
Definition: snow.h:45
#define QBIAS_SHIFT
Definition: snow.h:164
#define MAX_REF_FRAMES
Definition: snow.h:47
static av_always_inline int same_block(BlockNode *a, BlockNode *b)
Definition: snow.h:271
#define LOG2_OBMC_MAX
Definition: snow.h:49
const int8_t ff_quant3bA[256]
Definition: snowdata.h:104
static av_always_inline void add_yblock(SnowContext *s, int sliced, slice_buffer *sb, IDWTELEM *dst, uint8_t *dst8, const uint8_t *obmc, int src_x, int src_y, int b_w, int b_h, int w, int h, int dst_stride, int src_stride, int obmc_stride, int b_x, int b_y, int add, int offset_dst, int plane_index)
Definition: snow.h:281
#define ME_CACHE_SIZE
Definition: snow.h:172
static const BlockNode null_block
Definition: snow.h:64
void ff_spatial_idwt(IDWTELEM *buffer, IDWTELEM *temp, int width, int height, int stride, int type, int decomposition_count)
Definition: snow_dwt.c:731
int ff_w53_32_c(struct MpegEncContext *v, uint8_t *pix1, uint8_t *pix2, ptrdiff_t line_size, int h)
Definition: snow_dwt.c:832
void ff_spatial_dwt(DWTELEM *buffer, DWTELEM *temp, int width, int height, int stride, int type, int decomposition_count)
Definition: snow_dwt.c:319
int ff_w97_32_c(struct MpegEncContext *v, uint8_t *pix1, uint8_t *pix2, ptrdiff_t line_size, int h)
Definition: snow_dwt.c:837
#define DWT_97
Definition: snow_dwt.h:68
#define DWT_53
Definition: snow_dwt.h:69
static int get_4block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index)
Definition: snowenc.c:716
static av_always_inline int check_4block_inter(SnowContext *s, int mb_x, int mb_y, int p0, int p1, int ref, int *best_rd)
Definition: snowenc.c:958
static int qscale2qlog(int qscale)
Definition: snowenc.c:1490
static const AVClass snowenc_class
Definition: snowenc.c:1969
static int get_penalty_factor(int lambda, int lambda2, int type)
Definition: snowenc.c:197
static const AVOption options[]
Definition: snowenc.c:1947
#define VE
Definition: snowenc.c:1946
static int get_block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index, uint8_t(*obmc_edged)[MB_SIZE *2])
Definition: snowenc.c:615
static int encode_subband_c0run(SnowContext *s, SubBand *b, const IDWTELEM *src, const IDWTELEM *parent, int stride, int orientation)
Definition: snowenc.c:785
static void encode_q_branch2(SnowContext *s, int level, int x, int y)
Definition: snowenc.c:463
#define P_TOP
Definition: snowenc.c:223
static void calculate_visual_weight(SnowContext *s, Plane *p)
Definition: snowenc.c:1547
static void correlate(SnowContext *s, SubBand *b, IDWTELEM *src, int stride, int inverse, int use_median)
Definition: snowenc.c:1360
static int get_block_bits(SnowContext *s, int x, int y, int w)
Definition: snowenc.c:577
#define P_MEDIAN
Definition: snowenc.c:225
static av_cold int encode_init(AVCodecContext *avctx)
Definition: snowenc.c:38
static av_always_inline int check_block(SnowContext *s, int mb_x, int mb_y, int p[3], int intra, uint8_t(*obmc_edged)[MB_SIZE *2], int *best_rd)
Definition: snowenc.c:912
static int pix_sum(uint8_t *pix, int line_size, int w, int h)
Definition: snowenc.c:165
static int ratecontrol_1pass(SnowContext *s, AVFrame *pict)
Definition: snowenc.c:1495
static int get_dc(SnowContext *s, int mb_x, int mb_y, int plane_index)
Definition: snowenc.c:519
static int encode_frame(AVCodecContext *avctx, AVPacket *pkt, const AVFrame *pict, int *got_packet)
Definition: snowenc.c:1577
static void encode_qlogs(SnowContext *s)
Definition: snowenc.c:1384
static av_always_inline int check_block_inter(SnowContext *s, int mb_x, int mb_y, int p0, int p1, uint8_t(*obmc_edged)[MB_SIZE *2], int *best_rd)
Definition: snowenc.c:953
#define P_LEFT
Definition: snowenc.c:222
static int encode_q_branch(SnowContext *s, int level, int x, int y)
Definition: snowenc.c:229
static void decorrelate(SnowContext *s, SubBand *b, IDWTELEM *src, int stride, int inverse, int use_median)
Definition: snowenc.c:1336
static av_cold int encode_end(AVCodecContext *avctx)
Definition: snowenc.c:1933
static void encode_header(SnowContext *s)
Definition: snowenc.c:1397
AVCodec ff_snow_encoder
Definition: snowenc.c:1976
#define OFFSET(x)
Definition: snowenc.c:1945
static void dequantize(SnowContext *s, SubBand *b, IDWTELEM *src, int stride)
Definition: snowenc.c:1314
static void quantize(SnowContext *s, SubBand *b, IDWTELEM *dst, DWTELEM *src, int stride, int bias)
Definition: snowenc.c:1253
static void iterative_me(SnowContext *s)
Definition: snowenc.c:1004
static void update_last_header_values(SnowContext *s)
Definition: snowenc.c:1470
static int pix_norm1(uint8_t *pix, int line_size, int w)
Definition: snowenc.c:181
#define P_TOPRIGHT
Definition: snowenc.c:224
static void encode_blocks(SnowContext *s, int search)
Definition: snowenc.c:1231
static int encode_subband(SnowContext *s, SubBand *b, const IDWTELEM *src, const IDWTELEM *parent, int stride, int orientation)
Definition: snowenc.c:905
static int shift(int a, int b)
Definition: sonic.c:82
Describe the class of an AVClass context structure.
Definition: log.h:67
const char * class_name
The name of the class; usually it is the same name as the context structure type to which the AVClass...
Definition: log.h:72
main external API structure.
Definition: avcodec.h:536
enum AVPixelFormat pix_fmt
Pixel format, see AV_PIX_FMT_xxx.
Definition: avcodec.h:746
int width
picture width / height.
Definition: avcodec.h:709
char * stats_out
pass1 encoding statistics output buffer
Definition: avcodec.h:1561
int global_quality
Global quality for codecs which cannot change it per frame.
Definition: avcodec.h:602
attribute_deprecated int frame_bits
Definition: avcodec.h:1553
attribute_deprecated int mv_bits
Definition: avcodec.h:1535
int mb_lmax
maximum MB Lagrange multiplier
Definition: avcodec.h:1090
int dia_size
ME diamond size & shape.
Definition: avcodec.h:964
attribute_deprecated AVFrame * coded_frame
the picture in the bitstream
Definition: avcodec.h:1768
attribute_deprecated int misc_bits
Definition: avcodec.h:1549
int64_t bit_rate
the average bitrate
Definition: avcodec.h:586
int gop_size
the number of pictures in a group of pictures, or 0 for intra_only
Definition: avcodec.h:731
int frame_number
Frame counter, set by libavcodec.
Definition: avcodec.h:1227
int refs
number of reference frames
Definition: avcodec.h:1124
int mb_lmin
minimum MB Lagrange multiplier
Definition: avcodec.h:1083
int flags
AV_CODEC_FLAG_*.
Definition: avcodec.h:616
attribute_deprecated int prediction_method
Definition: avcodec.h:895
attribute_deprecated int p_tex_bits
Definition: avcodec.h:1541
void * priv_data
Definition: avcodec.h:563
AVCodec.
Definition: codec.h:197
const char * name
Name of the codec implementation.
Definition: codec.h:204
This structure describes decoded (raw) audio or video data.
Definition: frame.h:318
int64_t pts
Presentation timestamp in time_base units (time when frame should be shown to user).
Definition: frame.h:411
uint8_t * data[AV_NUM_DATA_POINTERS]
pointer to the picture/channel planes.
Definition: frame.h:332
int quality
quality (between 1 (good) and FF_LAMBDA_MAX (bad))
Definition: frame.h:441
int linesize[AV_NUM_DATA_POINTERS]
For video, size in bytes of each picture line.
Definition: frame.h:349
enum AVPictureType pict_type
Picture type of the frame.
Definition: frame.h:401
AVOption.
Definition: opt.h:248
This structure stores compressed data.
Definition: packet.h:346
int flags
A combination of AV_PKT_FLAG values.
Definition: packet.h:375
int size
Definition: packet.h:370
uint8_t * data
Definition: packet.h:369
Definition: snow.h:51
uint8_t ref
Reference frame index.
Definition: snow.h:54
int16_t mx
Motion vector component X, see mv_scale.
Definition: snow.h:52
uint8_t color[3]
Color for intra.
Definition: snow.h:55
int16_t my
Motion vector component Y, see mv_scale.
Definition: snow.h:53
uint8_t type
Bitfield of BLOCK_*.
Definition: snow.h:56
uint8_t level
Definition: snow.h:61
Motion estimation context.
Definition: motion_est.h:47
Definition: cfhd.h:120
int last_htaps
Definition: snow.h:109
SubBand band[DWT_LEVELS_3D][4]
Definition: cfhd.h:133
int htaps
Definition: snow.h:104
int last_diag_mc
Definition: snow.h:111
int width
Definition: cfhd.h:121
int8_t last_hcoeff[HTAPS_MAX/2]
Definition: snow.h:110
int height
Definition: cfhd.h:122
int8_t hcoeff[HTAPS_MAX/2]
Definition: snow.h:105
int diag_mc
Definition: snow.h:106
Definition: cfhd.h:111
int qlog
log(qscale)/log[2^(1/6)]
Definition: snow.h:88
uint8_t run
Definition: svq3.c:205
uint8_t level
Definition: svq3.c:206
#define lrint
Definition: tablegen.h:53
#define av_freep(p)
#define av_log(a,...)
static void error(const char *err)
static uint8_t tmp[11]
Definition: aes_ctr.c:27
#define src
Definition: vp8dsp.c:255
static int16_t block[64]
Definition: dct.c:116
static int ref[MAX_W *MAX_W]
Definition: jpeg2000dwt.c:107
AVPacket * pkt
Definition: movenc.c:59
enum AVPictureType last_picture
Definition: movenc.c:69
#define height
#define width
int size
#define pass
Definition: tx_template.c:347
const char * b
Definition: vf_curves.c:118
const char * r
Definition: vf_curves.c:116
static float search(FOCContext *foc, int pass, int maxpass, int xmin, int xmax, int ymin, int ymax, int *best_x, int *best_y, float best_score)
Definition: vf_find_rect.c:155
static double cr(void *priv, double x, double y)
Definition: vf_geq.c:216
static double cb(void *priv, double x, double y)
Definition: vf_geq.c:215
static double c[64]