motion_est.c
Go to the documentation of this file.
1 /*
2  * Motion estimation
3  * Copyright (c) 2000,2001 Fabrice Bellard
4  * Copyright (c) 2002-2004 Michael Niedermayer
5  *
6  * new motion estimation (X1/EPZS) by Michael Niedermayer <michaelni@gmx.at>
7  *
8  * This file is part of Libav.
9  *
10  * Libav is free software; you can redistribute it and/or
11  * modify it under the terms of the GNU Lesser General Public
12  * License as published by the Free Software Foundation; either
13  * version 2.1 of the License, or (at your option) any later version.
14  *
15  * Libav is distributed in the hope that it will be useful,
16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18  * Lesser General Public License for more details.
19  *
20  * You should have received a copy of the GNU Lesser General Public
21  * License along with Libav; if not, write to the Free Software
22  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23  */
24 
30 #include <stdlib.h>
31 #include <stdio.h>
32 #include <limits.h>
33 #include "libavutil/intmath.h"
34 #include "avcodec.h"
35 #include "dsputil.h"
36 #include "mathops.h"
37 #include "mpegvideo.h"
38 
39 #undef NDEBUG
40 #include <assert.h>
41 
42 #define SQ(a) ((a)*(a))
43 
44 #define P_LEFT P[1]
45 #define P_TOP P[2]
46 #define P_TOPRIGHT P[3]
47 #define P_MEDIAN P[4]
48 #define P_MV1 P[9]
49 
50 static inline int sad_hpel_motion_search(MpegEncContext * s,
51  int *mx_ptr, int *my_ptr, int dmin,
52  int src_index, int ref_index,
53  int size, int h);
54 
55 static inline unsigned update_map_generation(MotionEstContext *c)
56 {
57  c->map_generation+= 1<<(ME_MAP_MV_BITS*2);
58  if(c->map_generation==0){
59  c->map_generation= 1<<(ME_MAP_MV_BITS*2);
60  memset(c->map, 0, sizeof(uint32_t)*ME_MAP_SIZE);
61  }
62  return c->map_generation;
63 }
64 
65 /* shape adaptive search stuff */
66 typedef struct Minima{
67  int height;
68  int x, y;
69  int checked;
70 }Minima;
71 
72 static int minima_cmp(const void *a, const void *b){
73  const Minima *da = (const Minima *) a;
74  const Minima *db = (const Minima *) b;
75 
76  return da->height - db->height;
77 }
78 
79 #define FLAG_QPEL 1 //must be 1
80 #define FLAG_CHROMA 2
81 #define FLAG_DIRECT 4
82 
83 static inline void init_ref(MotionEstContext *c, uint8_t *src[3], uint8_t *ref[3], uint8_t *ref2[3], int x, int y, int ref_index){
84  const int offset[3]= {
85  y*c-> stride + x,
86  ((y*c->uvstride + x)>>1),
87  ((y*c->uvstride + x)>>1),
88  };
89  int i;
90  for(i=0; i<3; i++){
91  c->src[0][i]= src [i] + offset[i];
92  c->ref[0][i]= ref [i] + offset[i];
93  }
94  if(ref_index){
95  for(i=0; i<3; i++){
96  c->ref[ref_index][i]= ref2[i] + offset[i];
97  }
98  }
99 }
100 
101 static int get_flags(MotionEstContext *c, int direct, int chroma){
102  return ((c->avctx->flags&CODEC_FLAG_QPEL) ? FLAG_QPEL : 0)
103  + (direct ? FLAG_DIRECT : 0)
104  + (chroma ? FLAG_CHROMA : 0);
105 }
106 
107 static av_always_inline int cmp_direct_inline(MpegEncContext *s, const int x, const int y, const int subx, const int suby,
108  const int size, const int h, int ref_index, int src_index,
109  me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, int qpel){
110  MotionEstContext * const c= &s->me;
111  const int stride= c->stride;
112  const int hx= subx + (x<<(1+qpel));
113  const int hy= suby + (y<<(1+qpel));
114  uint8_t * const * const ref= c->ref[ref_index];
115  uint8_t * const * const src= c->src[src_index];
116  int d;
117  //FIXME check chroma 4mv, (no crashes ...)
118  assert(x >= c->xmin && hx <= c->xmax<<(qpel+1) && y >= c->ymin && hy <= c->ymax<<(qpel+1));
119  if(x >= c->xmin && hx <= c->xmax<<(qpel+1) && y >= c->ymin && hy <= c->ymax<<(qpel+1)){
120  const int time_pp= s->pp_time;
121  const int time_pb= s->pb_time;
122  const int mask= 2*qpel+1;
123  if(s->mv_type==MV_TYPE_8X8){
124  int i;
125  for(i=0; i<4; i++){
126  int fx = c->direct_basis_mv[i][0] + hx;
127  int fy = c->direct_basis_mv[i][1] + hy;
128  int bx = hx ? fx - c->co_located_mv[i][0] : c->co_located_mv[i][0]*(time_pb - time_pp)/time_pp + ((i &1)<<(qpel+4));
129  int by = hy ? fy - c->co_located_mv[i][1] : c->co_located_mv[i][1]*(time_pb - time_pp)/time_pp + ((i>>1)<<(qpel+4));
130  int fxy= (fx&mask) + ((fy&mask)<<(qpel+1));
131  int bxy= (bx&mask) + ((by&mask)<<(qpel+1));
132 
133  uint8_t *dst= c->temp + 8*(i&1) + 8*stride*(i>>1);
134  if(qpel){
135  c->qpel_put[1][fxy](dst, ref[0] + (fx>>2) + (fy>>2)*stride, stride);
136  c->qpel_avg[1][bxy](dst, ref[8] + (bx>>2) + (by>>2)*stride, stride);
137  }else{
138  c->hpel_put[1][fxy](dst, ref[0] + (fx>>1) + (fy>>1)*stride, stride, 8);
139  c->hpel_avg[1][bxy](dst, ref[8] + (bx>>1) + (by>>1)*stride, stride, 8);
140  }
141  }
142  }else{
143  int fx = c->direct_basis_mv[0][0] + hx;
144  int fy = c->direct_basis_mv[0][1] + hy;
145  int bx = hx ? fx - c->co_located_mv[0][0] : (c->co_located_mv[0][0]*(time_pb - time_pp)/time_pp);
146  int by = hy ? fy - c->co_located_mv[0][1] : (c->co_located_mv[0][1]*(time_pb - time_pp)/time_pp);
147  int fxy= (fx&mask) + ((fy&mask)<<(qpel+1));
148  int bxy= (bx&mask) + ((by&mask)<<(qpel+1));
149 
150  if(qpel){
151  c->qpel_put[1][fxy](c->temp , ref[0] + (fx>>2) + (fy>>2)*stride , stride);
152  c->qpel_put[1][fxy](c->temp + 8 , ref[0] + (fx>>2) + (fy>>2)*stride + 8 , stride);
153  c->qpel_put[1][fxy](c->temp + 8*stride, ref[0] + (fx>>2) + (fy>>2)*stride + 8*stride, stride);
154  c->qpel_put[1][fxy](c->temp + 8 + 8*stride, ref[0] + (fx>>2) + (fy>>2)*stride + 8 + 8*stride, stride);
155  c->qpel_avg[1][bxy](c->temp , ref[8] + (bx>>2) + (by>>2)*stride , stride);
156  c->qpel_avg[1][bxy](c->temp + 8 , ref[8] + (bx>>2) + (by>>2)*stride + 8 , stride);
157  c->qpel_avg[1][bxy](c->temp + 8*stride, ref[8] + (bx>>2) + (by>>2)*stride + 8*stride, stride);
158  c->qpel_avg[1][bxy](c->temp + 8 + 8*stride, ref[8] + (bx>>2) + (by>>2)*stride + 8 + 8*stride, stride);
159  }else{
160  assert((fx>>1) + 16*s->mb_x >= -16);
161  assert((fy>>1) + 16*s->mb_y >= -16);
162  assert((fx>>1) + 16*s->mb_x <= s->width);
163  assert((fy>>1) + 16*s->mb_y <= s->height);
164  assert((bx>>1) + 16*s->mb_x >= -16);
165  assert((by>>1) + 16*s->mb_y >= -16);
166  assert((bx>>1) + 16*s->mb_x <= s->width);
167  assert((by>>1) + 16*s->mb_y <= s->height);
168 
169  c->hpel_put[0][fxy](c->temp, ref[0] + (fx>>1) + (fy>>1)*stride, stride, 16);
170  c->hpel_avg[0][bxy](c->temp, ref[8] + (bx>>1) + (by>>1)*stride, stride, 16);
171  }
172  }
173  d = cmp_func(s, c->temp, src[0], stride, 16);
174  }else
175  d= 256*256*256*32;
176  return d;
177 }
178 
179 static av_always_inline int cmp_inline(MpegEncContext *s, const int x, const int y, const int subx, const int suby,
180  const int size, const int h, int ref_index, int src_index,
181  me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, int qpel, int chroma){
182  MotionEstContext * const c= &s->me;
183  const int stride= c->stride;
184  const int uvstride= c->uvstride;
185  const int dxy= subx + (suby<<(1+qpel)); //FIXME log2_subpel?
186  const int hx= subx + (x<<(1+qpel));
187  const int hy= suby + (y<<(1+qpel));
188  uint8_t * const * const ref= c->ref[ref_index];
189  uint8_t * const * const src= c->src[src_index];
190  int d;
191  //FIXME check chroma 4mv, (no crashes ...)
192  int uvdxy; /* no, it might not be used uninitialized */
193  if(dxy){
194  if(qpel){
195  c->qpel_put[size][dxy](c->temp, ref[0] + x + y*stride, stride); //FIXME prototype (add h)
196  if(chroma){
197  int cx= hx/2;
198  int cy= hy/2;
199  cx= (cx>>1)|(cx&1);
200  cy= (cy>>1)|(cy&1);
201  uvdxy= (cx&1) + 2*(cy&1);
202  //FIXME x/y wrong, but mpeg4 qpel is sick anyway, we should drop as much of it as possible in favor for h264
203  }
204  }else{
205  c->hpel_put[size][dxy](c->temp, ref[0] + x + y*stride, stride, h);
206  if(chroma)
207  uvdxy= dxy | (x&1) | (2*(y&1));
208  }
209  d = cmp_func(s, c->temp, src[0], stride, h);
210  }else{
211  d = cmp_func(s, src[0], ref[0] + x + y*stride, stride, h);
212  if(chroma)
213  uvdxy= (x&1) + 2*(y&1);
214  }
215  if(chroma){
216  uint8_t * const uvtemp= c->temp + 16*stride;
217  c->hpel_put[size+1][uvdxy](uvtemp , ref[1] + (x>>1) + (y>>1)*uvstride, uvstride, h>>1);
218  c->hpel_put[size+1][uvdxy](uvtemp+8, ref[2] + (x>>1) + (y>>1)*uvstride, uvstride, h>>1);
219  d += chroma_cmp_func(s, uvtemp , src[1], uvstride, h>>1);
220  d += chroma_cmp_func(s, uvtemp+8, src[2], uvstride, h>>1);
221  }
222  return d;
223 }
224 
225 static int cmp_simple(MpegEncContext *s, const int x, const int y,
226  int ref_index, int src_index,
227  me_cmp_func cmp_func, me_cmp_func chroma_cmp_func){
228  return cmp_inline(s,x,y,0,0,0,16,ref_index,src_index, cmp_func, chroma_cmp_func, 0, 0);
229 }
230 
231 static int cmp_fpel_internal(MpegEncContext *s, const int x, const int y,
232  const int size, const int h, int ref_index, int src_index,
233  me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, const int flags){
234  if(flags&FLAG_DIRECT){
235  return cmp_direct_inline(s,x,y,0,0,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, flags&FLAG_QPEL);
236  }else{
237  return cmp_inline(s,x,y,0,0,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, 0, flags&FLAG_CHROMA);
238  }
239 }
240 
241 static int cmp_internal(MpegEncContext *s, const int x, const int y, const int subx, const int suby,
242  const int size, const int h, int ref_index, int src_index,
243  me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, const int flags){
244  if(flags&FLAG_DIRECT){
245  return cmp_direct_inline(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, flags&FLAG_QPEL);
246  }else{
247  return cmp_inline(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, flags&FLAG_QPEL, flags&FLAG_CHROMA);
248  }
249 }
250 
254 static av_always_inline int cmp(MpegEncContext *s, const int x, const int y, const int subx, const int suby,
255  const int size, const int h, int ref_index, int src_index,
256  me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, const int flags){
259  && flags==0 && h==16 && size==0 && subx==0 && suby==0){
260  return cmp_simple(s,x,y,ref_index,src_index, cmp_func, chroma_cmp_func);
261  }else if(av_builtin_constant_p(subx) && av_builtin_constant_p(suby)
262  && subx==0 && suby==0){
263  return cmp_fpel_internal(s,x,y,size,h,ref_index,src_index, cmp_func, chroma_cmp_func,flags);
264  }else{
265  return cmp_internal(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, flags);
266  }
267 }
268 
269 static int cmp_hpel(MpegEncContext *s, const int x, const int y, const int subx, const int suby,
270  const int size, const int h, int ref_index, int src_index,
271  me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, const int flags){
272  if(flags&FLAG_DIRECT){
273  return cmp_direct_inline(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, 0);
274  }else{
275  return cmp_inline(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, 0, flags&FLAG_CHROMA);
276  }
277 }
278 
279 static int cmp_qpel(MpegEncContext *s, const int x, const int y, const int subx, const int suby,
280  const int size, const int h, int ref_index, int src_index,
281  me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, const int flags){
282  if(flags&FLAG_DIRECT){
283  return cmp_direct_inline(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, 1);
284  }else{
285  return cmp_inline(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, 1, flags&FLAG_CHROMA);
286  }
287 }
288 
289 #include "motion_est_template.c"
290 
291 static int zero_cmp(void *s, uint8_t *a, uint8_t *b, int stride, int h){
292  return 0;
293 }
294 
295 static void zero_hpel(uint8_t *a, const uint8_t *b, int stride, int h){
296 }
297 
299  MotionEstContext * const c= &s->me;
300  int cache_size= FFMIN(ME_MAP_SIZE>>ME_MAP_SHIFT, 1<<ME_MAP_SHIFT);
301  int dia_size= FFMAX(FFABS(s->avctx->dia_size)&255, FFABS(s->avctx->pre_dia_size)&255);
302 
304  av_log(s->avctx, AV_LOG_ERROR, "ME_MAP size is too small for SAB diamond\n");
305  return -1;
306  }
307  //special case of snow is needed because snow uses its own iterative ME code
308  if(s->me_method!=ME_ZERO && s->me_method!=ME_EPZS && s->me_method!=ME_X1 && s->avctx->codec_id != CODEC_ID_SNOW){
309  av_log(s->avctx, AV_LOG_ERROR, "me_method is only allowed to be set to zero and epzs; for hex,umh,full and others see dia_size\n");
310  return -1;
311  }
312 
313  c->avctx= s->avctx;
314 
315  if(cache_size < 2*dia_size && !c->stride){
316  av_log(s->avctx, AV_LOG_INFO, "ME_MAP size may be a little small for the selected diamond size\n");
317  }
318 
320  ff_set_cmp(&s->dsp, s->dsp.me_cmp, c->avctx->me_cmp);
322  ff_set_cmp(&s->dsp, s->dsp.mb_cmp, c->avctx->mb_cmp);
323 
324  c->flags = get_flags(c, 0, c->avctx->me_cmp &FF_CMP_CHROMA);
326  c->mb_flags = get_flags(c, 0, c->avctx->mb_cmp &FF_CMP_CHROMA);
327 
328 /*FIXME s->no_rounding b_type*/
329  if(s->flags&CODEC_FLAG_QPEL){
333  else c->qpel_put= s->dsp.put_qpel_pixels_tab;
334  }else{
337  else if( c->avctx->me_sub_cmp == FF_CMP_SAD
338  && c->avctx-> me_cmp == FF_CMP_SAD
339  && c->avctx-> mb_cmp == FF_CMP_SAD)
340  c->sub_motion_search= sad_hpel_motion_search; // 2050 vs. 2450 cycles
341  else
343  }
344  c->hpel_avg= s->dsp.avg_pixels_tab;
346  else c->hpel_put= s->dsp.put_pixels_tab;
347 
348  if(s->linesize){
349  c->stride = s->linesize;
350  c->uvstride= s->uvlinesize;
351  }else{
352  c->stride = 16*s->mb_width + 32;
353  c->uvstride= 8*s->mb_width + 16;
354  }
355 
356  /* 8x8 fullpel search would need a 4x4 chroma compare, which we do
357  * not have yet, and even if we had, the motion estimation code
358  * does not expect it. */
359  if(s->codec_id != CODEC_ID_SNOW){
360  if((c->avctx->me_cmp&FF_CMP_CHROMA)/* && !s->dsp.me_cmp[2]*/){
361  s->dsp.me_cmp[2]= zero_cmp;
362  }
363  if((c->avctx->me_sub_cmp&FF_CMP_CHROMA) && !s->dsp.me_sub_cmp[2]){
364  s->dsp.me_sub_cmp[2]= zero_cmp;
365  }
366  c->hpel_put[2][0]= c->hpel_put[2][1]=
367  c->hpel_put[2][2]= c->hpel_put[2][3]= zero_hpel;
368  }
369 
370  if(s->codec_id == CODEC_ID_H261){
372  }
373 
374  return 0;
375 }
376 
377 static inline void no_motion_search(MpegEncContext * s,
378  int *mx_ptr, int *my_ptr)
379 {
380  *mx_ptr = 16 * s->mb_x;
381  *my_ptr = 16 * s->mb_y;
382 }
383 
384 #define Z_THRESHOLD 256
385 
386 #define CHECK_SAD_HALF_MV(suffix, x, y) \
387 {\
388  d= s->dsp.pix_abs[size][(x?1:0)+(y?2:0)](NULL, pix, ptr+((x)>>1), stride, h);\
389  d += (mv_penalty[pen_x + x] + mv_penalty[pen_y + y])*penalty_factor;\
390  COPY3_IF_LT(dminh, d, dx, x, dy, y)\
391 }
392 
394  int *mx_ptr, int *my_ptr, int dmin,
395  int src_index, int ref_index,
396  int size, int h)
397 {
398  MotionEstContext * const c= &s->me;
399  const int penalty_factor= c->sub_penalty_factor;
400  int mx, my, dminh;
401  uint8_t *pix, *ptr;
402  int stride= c->stride;
403  const int flags= c->sub_flags;
405 
406  assert(flags == 0);
407 
408  if(c->skip){
409 // printf("S");
410  *mx_ptr = 0;
411  *my_ptr = 0;
412  return dmin;
413  }
414 // printf("N");
415 
416  pix = c->src[src_index][0];
417 
418  mx = *mx_ptr;
419  my = *my_ptr;
420  ptr = c->ref[ref_index][0] + (my * stride) + mx;
421 
422  dminh = dmin;
423 
424  if (mx > xmin && mx < xmax &&
425  my > ymin && my < ymax) {
426  int dx=0, dy=0;
427  int d, pen_x, pen_y;
428  const int index= (my<<ME_MAP_SHIFT) + mx;
429  const int t= score_map[(index-(1<<ME_MAP_SHIFT))&(ME_MAP_SIZE-1)];
430  const int l= score_map[(index- 1 )&(ME_MAP_SIZE-1)];
431  const int r= score_map[(index+ 1 )&(ME_MAP_SIZE-1)];
432  const int b= score_map[(index+(1<<ME_MAP_SHIFT))&(ME_MAP_SIZE-1)];
433  mx<<=1;
434  my<<=1;
435 
436 
437  pen_x= pred_x + mx;
438  pen_y= pred_y + my;
439 
440  ptr-= stride;
441  if(t<=b){
442  CHECK_SAD_HALF_MV(y2 , 0, -1)
443  if(l<=r){
444  CHECK_SAD_HALF_MV(xy2, -1, -1)
445  if(t+r<=b+l){
446  CHECK_SAD_HALF_MV(xy2, +1, -1)
447  ptr+= stride;
448  }else{
449  ptr+= stride;
450  CHECK_SAD_HALF_MV(xy2, -1, +1)
451  }
452  CHECK_SAD_HALF_MV(x2 , -1, 0)
453  }else{
454  CHECK_SAD_HALF_MV(xy2, +1, -1)
455  if(t+l<=b+r){
456  CHECK_SAD_HALF_MV(xy2, -1, -1)
457  ptr+= stride;
458  }else{
459  ptr+= stride;
460  CHECK_SAD_HALF_MV(xy2, +1, +1)
461  }
462  CHECK_SAD_HALF_MV(x2 , +1, 0)
463  }
464  }else{
465  if(l<=r){
466  if(t+l<=b+r){
467  CHECK_SAD_HALF_MV(xy2, -1, -1)
468  ptr+= stride;
469  }else{
470  ptr+= stride;
471  CHECK_SAD_HALF_MV(xy2, +1, +1)
472  }
473  CHECK_SAD_HALF_MV(x2 , -1, 0)
474  CHECK_SAD_HALF_MV(xy2, -1, +1)
475  }else{
476  if(t+r<=b+l){
477  CHECK_SAD_HALF_MV(xy2, +1, -1)
478  ptr+= stride;
479  }else{
480  ptr+= stride;
481  CHECK_SAD_HALF_MV(xy2, -1, +1)
482  }
483  CHECK_SAD_HALF_MV(x2 , +1, 0)
484  CHECK_SAD_HALF_MV(xy2, +1, +1)
485  }
486  CHECK_SAD_HALF_MV(y2 , 0, +1)
487  }
488  mx+=dx;
489  my+=dy;
490 
491  }else{
492  mx<<=1;
493  my<<=1;
494  }
495 
496  *mx_ptr = mx;
497  *my_ptr = my;
498  return dminh;
499 }
500 
501 static inline void set_p_mv_tables(MpegEncContext * s, int mx, int my, int mv4)
502 {
503  const int xy= s->mb_x + s->mb_y*s->mb_stride;
504 
505  s->p_mv_table[xy][0] = mx;
506  s->p_mv_table[xy][1] = my;
507 
508  /* has already been set to the 4 MV if 4MV is done */
509  if(mv4){
510  int mot_xy= s->block_index[0];
511 
512  s->current_picture.f.motion_val[0][mot_xy ][0] = mx;
513  s->current_picture.f.motion_val[0][mot_xy ][1] = my;
514  s->current_picture.f.motion_val[0][mot_xy + 1][0] = mx;
515  s->current_picture.f.motion_val[0][mot_xy + 1][1] = my;
516 
517  mot_xy += s->b8_stride;
518  s->current_picture.f.motion_val[0][mot_xy ][0] = mx;
519  s->current_picture.f.motion_val[0][mot_xy ][1] = my;
520  s->current_picture.f.motion_val[0][mot_xy + 1][0] = mx;
521  s->current_picture.f.motion_val[0][mot_xy + 1][1] = my;
522  }
523 }
524 
528 static inline void get_limits(MpegEncContext *s, int x, int y)
529 {
530  MotionEstContext * const c= &s->me;
531  int range= c->avctx->me_range >> (1 + !!(c->flags&FLAG_QPEL));
532 /*
533  if(c->avctx->me_range) c->range= c->avctx->me_range >> 1;
534  else c->range= 16;
535 */
536  if (s->unrestricted_mv) {
537  c->xmin = - x - 16;
538  c->ymin = - y - 16;
539  c->xmax = - x + s->mb_width *16;
540  c->ymax = - y + s->mb_height*16;
541  } else if (s->out_format == FMT_H261){
542  // Search range of H261 is different from other codec standards
543  c->xmin = (x > 15) ? - 15 : 0;
544  c->ymin = (y > 15) ? - 15 : 0;
545  c->xmax = (x < s->mb_width * 16 - 16) ? 15 : 0;
546  c->ymax = (y < s->mb_height * 16 - 16) ? 15 : 0;
547  } else {
548  c->xmin = - x;
549  c->ymin = - y;
550  c->xmax = - x + s->mb_width *16 - 16;
551  c->ymax = - y + s->mb_height*16 - 16;
552  }
553  if(range){
554  c->xmin = FFMAX(c->xmin,-range);
555  c->xmax = FFMIN(c->xmax, range);
556  c->ymin = FFMAX(c->ymin,-range);
557  c->ymax = FFMIN(c->ymax, range);
558  }
559 }
560 
561 static inline void init_mv4_ref(MotionEstContext *c){
562  const int stride= c->stride;
563 
564  c->ref[1][0] = c->ref[0][0] + 8;
565  c->ref[2][0] = c->ref[0][0] + 8*stride;
566  c->ref[3][0] = c->ref[2][0] + 8;
567  c->src[1][0] = c->src[0][0] + 8;
568  c->src[2][0] = c->src[0][0] + 8*stride;
569  c->src[3][0] = c->src[2][0] + 8;
570 }
571 
572 static inline int h263_mv4_search(MpegEncContext *s, int mx, int my, int shift)
573 {
574  MotionEstContext * const c= &s->me;
575  const int size= 1;
576  const int h=8;
577  int block;
578  int P[10][2];
579  int dmin_sum=0, mx4_sum=0, my4_sum=0;
580  int same=1;
581  const int stride= c->stride;
582  uint8_t *mv_penalty= c->current_mv_penalty;
583 
584  init_mv4_ref(c);
585 
586  for(block=0; block<4; block++){
587  int mx4, my4;
588  int pred_x4, pred_y4;
589  int dmin4;
590  static const int off[4]= {2, 1, 1, -1};
591  const int mot_stride = s->b8_stride;
592  const int mot_xy = s->block_index[block];
593 
594  P_LEFT[0] = s->current_picture.f.motion_val[0][mot_xy - 1][0];
595  P_LEFT[1] = s->current_picture.f.motion_val[0][mot_xy - 1][1];
596 
597  if(P_LEFT[0] > (c->xmax<<shift)) P_LEFT[0] = (c->xmax<<shift);
598 
599  /* special case for first line */
600  if (s->first_slice_line && block<2) {
601  c->pred_x= pred_x4= P_LEFT[0];
602  c->pred_y= pred_y4= P_LEFT[1];
603  } else {
604  P_TOP[0] = s->current_picture.f.motion_val[0][mot_xy - mot_stride ][0];
605  P_TOP[1] = s->current_picture.f.motion_val[0][mot_xy - mot_stride ][1];
606  P_TOPRIGHT[0] = s->current_picture.f.motion_val[0][mot_xy - mot_stride + off[block]][0];
607  P_TOPRIGHT[1] = s->current_picture.f.motion_val[0][mot_xy - mot_stride + off[block]][1];
608  if(P_TOP[1] > (c->ymax<<shift)) P_TOP[1] = (c->ymax<<shift);
609  if(P_TOPRIGHT[0] < (c->xmin<<shift)) P_TOPRIGHT[0]= (c->xmin<<shift);
610  if(P_TOPRIGHT[0] > (c->xmax<<shift)) P_TOPRIGHT[0]= (c->xmax<<shift);
611  if(P_TOPRIGHT[1] > (c->ymax<<shift)) P_TOPRIGHT[1]= (c->ymax<<shift);
612 
613  P_MEDIAN[0]= mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
614  P_MEDIAN[1]= mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
615 
616  c->pred_x= pred_x4 = P_MEDIAN[0];
617  c->pred_y= pred_y4 = P_MEDIAN[1];
618  }
619  P_MV1[0]= mx;
620  P_MV1[1]= my;
621 
622  dmin4 = epzs_motion_search4(s, &mx4, &my4, P, block, block, s->p_mv_table, (1<<16)>>shift);
623 
624  dmin4= c->sub_motion_search(s, &mx4, &my4, dmin4, block, block, size, h);
625 
626  if(s->dsp.me_sub_cmp[0] != s->dsp.mb_cmp[0]){
627  int dxy;
628  const int offset= ((block&1) + (block>>1)*stride)*8;
629  uint8_t *dest_y = c->scratchpad + offset;
630  if(s->quarter_sample){
631  uint8_t *ref= c->ref[block][0] + (mx4>>2) + (my4>>2)*stride;
632  dxy = ((my4 & 3) << 2) | (mx4 & 3);
633 
634  if(s->no_rounding)
635  s->dsp.put_no_rnd_qpel_pixels_tab[1][dxy](dest_y , ref , stride);
636  else
637  s->dsp.put_qpel_pixels_tab [1][dxy](dest_y , ref , stride);
638  }else{
639  uint8_t *ref= c->ref[block][0] + (mx4>>1) + (my4>>1)*stride;
640  dxy = ((my4 & 1) << 1) | (mx4 & 1);
641 
642  if(s->no_rounding)
643  s->dsp.put_no_rnd_pixels_tab[1][dxy](dest_y , ref , stride, h);
644  else
645  s->dsp.put_pixels_tab [1][dxy](dest_y , ref , stride, h);
646  }
647  dmin_sum+= (mv_penalty[mx4-pred_x4] + mv_penalty[my4-pred_y4])*c->mb_penalty_factor;
648  }else
649  dmin_sum+= dmin4;
650 
651  if(s->quarter_sample){
652  mx4_sum+= mx4/2;
653  my4_sum+= my4/2;
654  }else{
655  mx4_sum+= mx4;
656  my4_sum+= my4;
657  }
658 
659  s->current_picture.f.motion_val[0][s->block_index[block]][0] = mx4;
660  s->current_picture.f.motion_val[0][s->block_index[block]][1] = my4;
661 
662  if(mx4 != mx || my4 != my) same=0;
663  }
664 
665  if(same)
666  return INT_MAX;
667 
668  if(s->dsp.me_sub_cmp[0] != s->dsp.mb_cmp[0]){
669  dmin_sum += s->dsp.mb_cmp[0](s, s->new_picture.f.data[0] + s->mb_x*16 + s->mb_y*16*stride, c->scratchpad, stride, 16);
670  }
671 
672  if(c->avctx->mb_cmp&FF_CMP_CHROMA){
673  int dxy;
674  int mx, my;
675  int offset;
676 
677  mx= ff_h263_round_chroma(mx4_sum);
678  my= ff_h263_round_chroma(my4_sum);
679  dxy = ((my & 1) << 1) | (mx & 1);
680 
681  offset= (s->mb_x*8 + (mx>>1)) + (s->mb_y*8 + (my>>1))*s->uvlinesize;
682 
683  if(s->no_rounding){
684  s->dsp.put_no_rnd_pixels_tab[1][dxy](c->scratchpad , s->last_picture.f.data[1] + offset, s->uvlinesize, 8);
685  s->dsp.put_no_rnd_pixels_tab[1][dxy](c->scratchpad + 8, s->last_picture.f.data[2] + offset, s->uvlinesize, 8);
686  }else{
687  s->dsp.put_pixels_tab [1][dxy](c->scratchpad , s->last_picture.f.data[1] + offset, s->uvlinesize, 8);
688  s->dsp.put_pixels_tab [1][dxy](c->scratchpad + 8, s->last_picture.f.data[2] + offset, s->uvlinesize, 8);
689  }
690 
691  dmin_sum += s->dsp.mb_cmp[1](s, s->new_picture.f.data[1] + s->mb_x*8 + s->mb_y*8*s->uvlinesize, c->scratchpad , s->uvlinesize, 8);
692  dmin_sum += s->dsp.mb_cmp[1](s, s->new_picture.f.data[2] + s->mb_x*8 + s->mb_y*8*s->uvlinesize, c->scratchpad+8, s->uvlinesize, 8);
693  }
694 
695  c->pred_x= mx;
696  c->pred_y= my;
697 
698  switch(c->avctx->mb_cmp&0xFF){
699  /*case FF_CMP_SSE:
700  return dmin_sum+ 32*s->qscale*s->qscale;*/
701  case FF_CMP_RD:
702  return dmin_sum;
703  default:
704  return dmin_sum+ 11*c->mb_penalty_factor;
705  }
706 }
707 
708 static inline void init_interlaced_ref(MpegEncContext *s, int ref_index){
709  MotionEstContext * const c= &s->me;
710 
711  c->ref[1+ref_index][0] = c->ref[0+ref_index][0] + s->linesize;
712  c->src[1][0] = c->src[0][0] + s->linesize;
713  if(c->flags & FLAG_CHROMA){
714  c->ref[1+ref_index][1] = c->ref[0+ref_index][1] + s->uvlinesize;
715  c->ref[1+ref_index][2] = c->ref[0+ref_index][2] + s->uvlinesize;
716  c->src[1][1] = c->src[0][1] + s->uvlinesize;
717  c->src[1][2] = c->src[0][2] + s->uvlinesize;
718  }
719 }
720 
721 static int interlaced_search(MpegEncContext *s, int ref_index,
722  int16_t (*mv_tables[2][2])[2], uint8_t *field_select_tables[2], int mx, int my, int user_field_select)
723 {
724  MotionEstContext * const c= &s->me;
725  const int size=0;
726  const int h=8;
727  int block;
728  int P[10][2];
729  uint8_t * const mv_penalty= c->current_mv_penalty;
730  int same=1;
731  const int stride= 2*s->linesize;
732  int dmin_sum= 0;
733  const int mot_stride= s->mb_stride;
734  const int xy= s->mb_x + s->mb_y*mot_stride;
735 
736  c->ymin>>=1;
737  c->ymax>>=1;
738  c->stride<<=1;
739  c->uvstride<<=1;
740  init_interlaced_ref(s, ref_index);
741 
742  for(block=0; block<2; block++){
743  int field_select;
744  int best_dmin= INT_MAX;
745  int best_field= -1;
746 
747  for(field_select=0; field_select<2; field_select++){
748  int dmin, mx_i, my_i;
749  int16_t (*mv_table)[2]= mv_tables[block][field_select];
750 
751  if(user_field_select){
752  assert(field_select==0 || field_select==1);
753  assert(field_select_tables[block][xy]==0 || field_select_tables[block][xy]==1);
754  if(field_select_tables[block][xy] != field_select)
755  continue;
756  }
757 
758  P_LEFT[0] = mv_table[xy - 1][0];
759  P_LEFT[1] = mv_table[xy - 1][1];
760  if(P_LEFT[0] > (c->xmax<<1)) P_LEFT[0] = (c->xmax<<1);
761 
762  c->pred_x= P_LEFT[0];
763  c->pred_y= P_LEFT[1];
764 
765  if(!s->first_slice_line){
766  P_TOP[0] = mv_table[xy - mot_stride][0];
767  P_TOP[1] = mv_table[xy - mot_stride][1];
768  P_TOPRIGHT[0] = mv_table[xy - mot_stride + 1][0];
769  P_TOPRIGHT[1] = mv_table[xy - mot_stride + 1][1];
770  if(P_TOP[1] > (c->ymax<<1)) P_TOP[1] = (c->ymax<<1);
771  if(P_TOPRIGHT[0] < (c->xmin<<1)) P_TOPRIGHT[0]= (c->xmin<<1);
772  if(P_TOPRIGHT[0] > (c->xmax<<1)) P_TOPRIGHT[0]= (c->xmax<<1);
773  if(P_TOPRIGHT[1] > (c->ymax<<1)) P_TOPRIGHT[1]= (c->ymax<<1);
774 
775  P_MEDIAN[0]= mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
776  P_MEDIAN[1]= mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
777  }
778  P_MV1[0]= mx; //FIXME not correct if block != field_select
779  P_MV1[1]= my / 2;
780 
781  dmin = epzs_motion_search2(s, &mx_i, &my_i, P, block, field_select+ref_index, mv_table, (1<<16)>>1);
782 
783  dmin= c->sub_motion_search(s, &mx_i, &my_i, dmin, block, field_select+ref_index, size, h);
784 
785  mv_table[xy][0]= mx_i;
786  mv_table[xy][1]= my_i;
787 
788  if(s->dsp.me_sub_cmp[0] != s->dsp.mb_cmp[0]){
789  int dxy;
790 
791  //FIXME chroma ME
792  uint8_t *ref= c->ref[field_select+ref_index][0] + (mx_i>>1) + (my_i>>1)*stride;
793  dxy = ((my_i & 1) << 1) | (mx_i & 1);
794 
795  if(s->no_rounding){
796  s->dsp.put_no_rnd_pixels_tab[size][dxy](c->scratchpad, ref , stride, h);
797  }else{
798  s->dsp.put_pixels_tab [size][dxy](c->scratchpad, ref , stride, h);
799  }
800  dmin= s->dsp.mb_cmp[size](s, c->src[block][0], c->scratchpad, stride, h);
801  dmin+= (mv_penalty[mx_i-c->pred_x] + mv_penalty[my_i-c->pred_y] + 1)*c->mb_penalty_factor;
802  }else
803  dmin+= c->mb_penalty_factor; //field_select bits
804 
805  dmin += field_select != block; //slightly prefer same field
806 
807  if(dmin < best_dmin){
808  best_dmin= dmin;
809  best_field= field_select;
810  }
811  }
812  {
813  int16_t (*mv_table)[2]= mv_tables[block][best_field];
814 
815  if(mv_table[xy][0] != mx) same=0; //FIXME check if these checks work and are any good at all
816  if(mv_table[xy][1]&1) same=0;
817  if(mv_table[xy][1]*2 != my) same=0;
818  if(best_field != block) same=0;
819  }
820 
821  field_select_tables[block][xy]= best_field;
822  dmin_sum += best_dmin;
823  }
824 
825  c->ymin<<=1;
826  c->ymax<<=1;
827  c->stride>>=1;
828  c->uvstride>>=1;
829 
830  if(same)
831  return INT_MAX;
832 
833  switch(c->avctx->mb_cmp&0xFF){
834  /*case FF_CMP_SSE:
835  return dmin_sum+ 32*s->qscale*s->qscale;*/
836  case FF_CMP_RD:
837  return dmin_sum;
838  default:
839  return dmin_sum+ 11*c->mb_penalty_factor;
840  }
841 }
842 
843 static void clip_input_mv(MpegEncContext * s, int16_t *mv, int interlaced){
844  int ymax= s->me.ymax>>interlaced;
845  int ymin= s->me.ymin>>interlaced;
846 
847  if(mv[0] < s->me.xmin) mv[0] = s->me.xmin;
848  if(mv[0] > s->me.xmax) mv[0] = s->me.xmax;
849  if(mv[1] < ymin) mv[1] = ymin;
850  if(mv[1] > ymax) mv[1] = ymax;
851 }
852 
853 static inline int check_input_motion(MpegEncContext * s, int mb_x, int mb_y, int p_type){
854  MotionEstContext * const c= &s->me;
856  int mb_xy= mb_x + mb_y*s->mb_stride;
857  int xy= 2*mb_x + 2*mb_y*s->b8_stride;
858  int mb_type= s->current_picture.f.mb_type[mb_xy];
859  int flags= c->flags;
860  int shift= (flags&FLAG_QPEL) + 1;
861  int mask= (1<<shift)-1;
862  int x, y, i;
863  int d=0;
864  me_cmp_func cmpf= s->dsp.sse[0];
865  me_cmp_func chroma_cmpf= s->dsp.sse[1];
866 
867  if(p_type && USES_LIST(mb_type, 1)){
868  av_log(c->avctx, AV_LOG_ERROR, "backward motion vector in P frame\n");
869  return INT_MAX/2;
870  }
871  assert(IS_INTRA(mb_type) || USES_LIST(mb_type,0) || USES_LIST(mb_type,1));
872 
873  for(i=0; i<4; i++){
874  int xy= s->block_index[i];
875  clip_input_mv(s, p->f.motion_val[0][xy], !!IS_INTERLACED(mb_type));
876  clip_input_mv(s, p->f.motion_val[1][xy], !!IS_INTERLACED(mb_type));
877  }
878 
879  if(IS_INTERLACED(mb_type)){
880  int xy2= xy + s->b8_stride;
882  c->stride<<=1;
883  c->uvstride<<=1;
884 
885  if(!(s->flags & CODEC_FLAG_INTERLACED_ME)){
886  av_log(c->avctx, AV_LOG_ERROR, "Interlaced macroblock selected but interlaced motion estimation disabled\n");
887  return INT_MAX/2;
888  }
889 
890  if(USES_LIST(mb_type, 0)){
891  int field_select0= p->f.ref_index[0][4*mb_xy ];
892  int field_select1= p->f.ref_index[0][4*mb_xy+2];
893  assert(field_select0==0 ||field_select0==1);
894  assert(field_select1==0 ||field_select1==1);
895  init_interlaced_ref(s, 0);
896 
897  if(p_type){
898  s->p_field_select_table[0][mb_xy]= field_select0;
899  s->p_field_select_table[1][mb_xy]= field_select1;
900  *(uint32_t*)s->p_field_mv_table[0][field_select0][mb_xy] = *(uint32_t*)p->f.motion_val[0][xy ];
901  *(uint32_t*)s->p_field_mv_table[1][field_select1][mb_xy] = *(uint32_t*)p->f.motion_val[0][xy2];
903  }else{
904  s->b_field_select_table[0][0][mb_xy]= field_select0;
905  s->b_field_select_table[0][1][mb_xy]= field_select1;
906  *(uint32_t*)s->b_field_mv_table[0][0][field_select0][mb_xy] = *(uint32_t*)p->f.motion_val[0][xy ];
907  *(uint32_t*)s->b_field_mv_table[0][1][field_select1][mb_xy] = *(uint32_t*)p->f.motion_val[0][xy2];
909  }
910 
911  x = p->f.motion_val[0][xy ][0];
912  y = p->f.motion_val[0][xy ][1];
913  d = cmp(s, x>>shift, y>>shift, x&mask, y&mask, 0, 8, field_select0, 0, cmpf, chroma_cmpf, flags);
914  x = p->f.motion_val[0][xy2][0];
915  y = p->f.motion_val[0][xy2][1];
916  d+= cmp(s, x>>shift, y>>shift, x&mask, y&mask, 0, 8, field_select1, 1, cmpf, chroma_cmpf, flags);
917  }
918  if(USES_LIST(mb_type, 1)){
919  int field_select0 = p->f.ref_index[1][4 * mb_xy ];
920  int field_select1 = p->f.ref_index[1][4 * mb_xy + 2];
921  assert(field_select0==0 ||field_select0==1);
922  assert(field_select1==0 ||field_select1==1);
923  init_interlaced_ref(s, 2);
924 
925  s->b_field_select_table[1][0][mb_xy]= field_select0;
926  s->b_field_select_table[1][1][mb_xy]= field_select1;
927  *(uint32_t*)s->b_field_mv_table[1][0][field_select0][mb_xy] = *(uint32_t*)p->f.motion_val[1][xy ];
928  *(uint32_t*)s->b_field_mv_table[1][1][field_select1][mb_xy] = *(uint32_t*)p->f.motion_val[1][xy2];
929  if(USES_LIST(mb_type, 0)){
931  }else{
933  }
934 
935  x = p->f.motion_val[1][xy ][0];
936  y = p->f.motion_val[1][xy ][1];
937  d = cmp(s, x>>shift, y>>shift, x&mask, y&mask, 0, 8, field_select0+2, 0, cmpf, chroma_cmpf, flags);
938  x = p->f.motion_val[1][xy2][0];
939  y = p->f.motion_val[1][xy2][1];
940  d+= cmp(s, x>>shift, y>>shift, x&mask, y&mask, 0, 8, field_select1+2, 1, cmpf, chroma_cmpf, flags);
941  //FIXME bidir scores
942  }
943  c->stride>>=1;
944  c->uvstride>>=1;
945  }else if(IS_8X8(mb_type)){
946  if(!(s->flags & CODEC_FLAG_4MV)){
947  av_log(c->avctx, AV_LOG_ERROR, "4MV macroblock selected but 4MV encoding disabled\n");
948  return INT_MAX/2;
949  }
950  cmpf= s->dsp.sse[1];
951  chroma_cmpf= s->dsp.sse[1];
952  init_mv4_ref(c);
953  for(i=0; i<4; i++){
954  xy= s->block_index[i];
955  x= p->f.motion_val[0][xy][0];
956  y= p->f.motion_val[0][xy][1];
957  d+= cmp(s, x>>shift, y>>shift, x&mask, y&mask, 1, 8, i, i, cmpf, chroma_cmpf, flags);
958  }
960  }else{
961  if(USES_LIST(mb_type, 0)){
962  if(p_type){
963  *(uint32_t*)s->p_mv_table[mb_xy] = *(uint32_t*)p->f.motion_val[0][xy];
965  }else if(USES_LIST(mb_type, 1)){
966  *(uint32_t*)s->b_bidir_forw_mv_table[mb_xy] = *(uint32_t*)p->f.motion_val[0][xy];
967  *(uint32_t*)s->b_bidir_back_mv_table[mb_xy] = *(uint32_t*)p->f.motion_val[1][xy];
969  }else{
970  *(uint32_t*)s->b_forw_mv_table[mb_xy] = *(uint32_t*)p->f.motion_val[0][xy];
972  }
973  x = p->f.motion_val[0][xy][0];
974  y = p->f.motion_val[0][xy][1];
975  d = cmp(s, x>>shift, y>>shift, x&mask, y&mask, 0, 16, 0, 0, cmpf, chroma_cmpf, flags);
976  }else if(USES_LIST(mb_type, 1)){
977  *(uint32_t*)s->b_back_mv_table[mb_xy] = *(uint32_t*)p->f.motion_val[1][xy];
979 
980  x = p->f.motion_val[1][xy][0];
981  y = p->f.motion_val[1][xy][1];
982  d = cmp(s, x>>shift, y>>shift, x&mask, y&mask, 0, 16, 2, 0, cmpf, chroma_cmpf, flags);
983  }else
985  }
986  return d;
987 }
988 
990  int mb_x, int mb_y)
991 {
992  MotionEstContext * const c= &s->me;
993  uint8_t *pix, *ppix;
994  int sum, mx, my, dmin;
995  int varc;
996  int vard;
997  int P[10][2];
998  const int shift= 1+s->quarter_sample;
999  int mb_type=0;
1000  Picture * const pic= &s->current_picture;
1001 
1002  init_ref(c, s->new_picture.f.data, s->last_picture.f.data, NULL, 16*mb_x, 16*mb_y, 0);
1003 
1004  assert(s->quarter_sample==0 || s->quarter_sample==1);
1005  assert(s->linesize == c->stride);
1006  assert(s->uvlinesize == c->uvstride);
1007 
1012 
1013  get_limits(s, 16*mb_x, 16*mb_y);
1014  c->skip=0;
1015 
1016  /* intra / predictive decision */
1017  pix = c->src[0][0];
1018  sum = s->dsp.pix_sum(pix, s->linesize);
1019  varc = s->dsp.pix_norm1(pix, s->linesize) - (((unsigned)sum*sum)>>8) + 500;
1020 
1021  pic->mb_mean[s->mb_stride * mb_y + mb_x] = (sum+128)>>8;
1022  pic->mb_var [s->mb_stride * mb_y + mb_x] = (varc+128)>>8;
1023  c->mb_var_sum_temp += (varc+128)>>8;
1024 
1025  if(c->avctx->me_threshold){
1026  vard= check_input_motion(s, mb_x, mb_y, 1);
1027 
1028  if((vard+128)>>8 < c->avctx->me_threshold){
1029  int p_score= FFMIN(vard, varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*100);
1030  int i_score= varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*20;
1031  pic->mc_mb_var[s->mb_stride * mb_y + mb_x] = (vard+128)>>8;
1032  c->mc_mb_var_sum_temp += (vard+128)>>8;
1033  c->scene_change_score+= ff_sqrt(p_score) - ff_sqrt(i_score);
1034  return;
1035  }
1036  if((vard+128)>>8 < c->avctx->mb_threshold)
1037  mb_type= s->mb_type[mb_x + mb_y*s->mb_stride];
1038  }
1039 
1040  switch(s->me_method) {
1041  case ME_ZERO:
1042  default:
1043  no_motion_search(s, &mx, &my);
1044  mx-= mb_x*16;
1045  my-= mb_y*16;
1046  dmin = 0;
1047  break;
1048  case ME_X1:
1049  case ME_EPZS:
1050  {
1051  const int mot_stride = s->b8_stride;
1052  const int mot_xy = s->block_index[0];
1053 
1054  P_LEFT[0] = s->current_picture.f.motion_val[0][mot_xy - 1][0];
1055  P_LEFT[1] = s->current_picture.f.motion_val[0][mot_xy - 1][1];
1056 
1057  if(P_LEFT[0] > (c->xmax<<shift)) P_LEFT[0] = (c->xmax<<shift);
1058 
1059  if(!s->first_slice_line) {
1060  P_TOP[0] = s->current_picture.f.motion_val[0][mot_xy - mot_stride ][0];
1061  P_TOP[1] = s->current_picture.f.motion_val[0][mot_xy - mot_stride ][1];
1062  P_TOPRIGHT[0] = s->current_picture.f.motion_val[0][mot_xy - mot_stride + 2][0];
1063  P_TOPRIGHT[1] = s->current_picture.f.motion_val[0][mot_xy - mot_stride + 2][1];
1064  if(P_TOP[1] > (c->ymax<<shift)) P_TOP[1] = (c->ymax<<shift);
1065  if(P_TOPRIGHT[0] < (c->xmin<<shift)) P_TOPRIGHT[0]= (c->xmin<<shift);
1066  if(P_TOPRIGHT[1] > (c->ymax<<shift)) P_TOPRIGHT[1]= (c->ymax<<shift);
1067 
1068  P_MEDIAN[0]= mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
1069  P_MEDIAN[1]= mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
1070 
1071  if(s->out_format == FMT_H263){
1072  c->pred_x = P_MEDIAN[0];
1073  c->pred_y = P_MEDIAN[1];
1074  }else { /* mpeg1 at least */
1075  c->pred_x= P_LEFT[0];
1076  c->pred_y= P_LEFT[1];
1077  }
1078  }else{
1079  c->pred_x= P_LEFT[0];
1080  c->pred_y= P_LEFT[1];
1081  }
1082 
1083  }
1084  dmin = ff_epzs_motion_search(s, &mx, &my, P, 0, 0, s->p_mv_table, (1<<16)>>shift, 0, 16);
1085 
1086  break;
1087  }
1088 
1089  /* At this point (mx,my) are full-pell and the relative displacement */
1090  ppix = c->ref[0][0] + (my * s->linesize) + mx;
1091 
1092  vard = s->dsp.sse[0](NULL, pix, ppix, s->linesize, 16);
1093 
1094  pic->mc_mb_var[s->mb_stride * mb_y + mb_x] = (vard+128)>>8;
1095 // pic->mb_cmp_score[s->mb_stride * mb_y + mb_x] = dmin;
1096  c->mc_mb_var_sum_temp += (vard+128)>>8;
1097 
1098  if(mb_type){
1099  int p_score= FFMIN(vard, varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*100);
1100  int i_score= varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*20;
1101  c->scene_change_score+= ff_sqrt(p_score) - ff_sqrt(i_score);
1102 
1103  if(mb_type == CANDIDATE_MB_TYPE_INTER){
1104  c->sub_motion_search(s, &mx, &my, dmin, 0, 0, 0, 16);
1105  set_p_mv_tables(s, mx, my, 1);
1106  }else{
1107  mx <<=shift;
1108  my <<=shift;
1109  }
1110  if(mb_type == CANDIDATE_MB_TYPE_INTER4V){
1111  h263_mv4_search(s, mx, my, shift);
1112 
1113  set_p_mv_tables(s, mx, my, 0);
1114  }
1115  if(mb_type == CANDIDATE_MB_TYPE_INTER_I){
1116  interlaced_search(s, 0, s->p_field_mv_table, s->p_field_select_table, mx, my, 1);
1117  }
1118  }else if(c->avctx->mb_decision > FF_MB_DECISION_SIMPLE){
1119  int p_score= FFMIN(vard, varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*100);
1120  int i_score= varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*20;
1121  c->scene_change_score+= ff_sqrt(p_score) - ff_sqrt(i_score);
1122 
1123  if (vard*2 + 200*256 > varc)
1124  mb_type|= CANDIDATE_MB_TYPE_INTRA;
1125  if (varc*2 + 200*256 > vard || s->qscale > 24){
1126 // if (varc*2 + 200*256 + 50*(s->lambda2>>FF_LAMBDA_SHIFT) > vard){
1127  mb_type|= CANDIDATE_MB_TYPE_INTER;
1128  c->sub_motion_search(s, &mx, &my, dmin, 0, 0, 0, 16);
1129  if(s->flags&CODEC_FLAG_MV0)
1130  if(mx || my)
1131  mb_type |= CANDIDATE_MB_TYPE_SKIPPED; //FIXME check difference
1132  }else{
1133  mx <<=shift;
1134  my <<=shift;
1135  }
1136  if((s->flags&CODEC_FLAG_4MV)
1137  && !c->skip && varc>50<<8 && vard>10<<8){
1138  if(h263_mv4_search(s, mx, my, shift) < INT_MAX)
1139  mb_type|=CANDIDATE_MB_TYPE_INTER4V;
1140 
1141  set_p_mv_tables(s, mx, my, 0);
1142  }else
1143  set_p_mv_tables(s, mx, my, 1);
1145  && !c->skip){ //FIXME varc/d checks
1146  if(interlaced_search(s, 0, s->p_field_mv_table, s->p_field_select_table, mx, my, 0) < INT_MAX)
1147  mb_type |= CANDIDATE_MB_TYPE_INTER_I;
1148  }
1149  }else{
1150  int intra_score, i;
1151  mb_type= CANDIDATE_MB_TYPE_INTER;
1152 
1153  dmin= c->sub_motion_search(s, &mx, &my, dmin, 0, 0, 0, 16);
1154  if(c->avctx->me_sub_cmp != c->avctx->mb_cmp && !c->skip)
1155  dmin= ff_get_mb_score(s, mx, my, 0, 0, 0, 16, 1);
1156 
1157  if((s->flags&CODEC_FLAG_4MV)
1158  && !c->skip && varc>50<<8 && vard>10<<8){
1159  int dmin4= h263_mv4_search(s, mx, my, shift);
1160  if(dmin4 < dmin){
1161  mb_type= CANDIDATE_MB_TYPE_INTER4V;
1162  dmin=dmin4;
1163  }
1164  }
1166  && !c->skip){ //FIXME varc/d checks
1167  int dmin_i= interlaced_search(s, 0, s->p_field_mv_table, s->p_field_select_table, mx, my, 0);
1168  if(dmin_i < dmin){
1169  mb_type = CANDIDATE_MB_TYPE_INTER_I;
1170  dmin= dmin_i;
1171  }
1172  }
1173 
1174 // pic->mb_cmp_score[s->mb_stride * mb_y + mb_x] = dmin;
1175  set_p_mv_tables(s, mx, my, mb_type!=CANDIDATE_MB_TYPE_INTER4V);
1176 
1177  /* get intra luma score */
1178  if((c->avctx->mb_cmp&0xFF)==FF_CMP_SSE){
1179  intra_score= varc - 500;
1180  }else{
1181  unsigned mean = (sum+128)>>8;
1182  mean*= 0x01010101;
1183 
1184  for(i=0; i<16; i++){
1185  *(uint32_t*)(&c->scratchpad[i*s->linesize+ 0]) = mean;
1186  *(uint32_t*)(&c->scratchpad[i*s->linesize+ 4]) = mean;
1187  *(uint32_t*)(&c->scratchpad[i*s->linesize+ 8]) = mean;
1188  *(uint32_t*)(&c->scratchpad[i*s->linesize+12]) = mean;
1189  }
1190 
1191  intra_score= s->dsp.mb_cmp[0](s, c->scratchpad, pix, s->linesize, 16);
1192  }
1193  intra_score += c->mb_penalty_factor*16;
1194 
1195  if(intra_score < dmin){
1196  mb_type= CANDIDATE_MB_TYPE_INTRA;
1197  s->current_picture.f.mb_type[mb_y*s->mb_stride + mb_x] = CANDIDATE_MB_TYPE_INTRA; //FIXME cleanup
1198  }else
1199  s->current_picture.f.mb_type[mb_y*s->mb_stride + mb_x] = 0;
1200 
1201  {
1202  int p_score= FFMIN(vard, varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*100);
1203  int i_score= varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*20;
1204  c->scene_change_score+= ff_sqrt(p_score) - ff_sqrt(i_score);
1205  }
1206  }
1207 
1208  s->mb_type[mb_y*s->mb_stride + mb_x]= mb_type;
1209 }
1210 
1212  int mb_x, int mb_y)
1213 {
1214  MotionEstContext * const c= &s->me;
1215  int mx, my, dmin;
1216  int P[10][2];
1217  const int shift= 1+s->quarter_sample;
1218  const int xy= mb_x + mb_y*s->mb_stride;
1219  init_ref(c, s->new_picture.f.data, s->last_picture.f.data, NULL, 16*mb_x, 16*mb_y, 0);
1220 
1221  assert(s->quarter_sample==0 || s->quarter_sample==1);
1222 
1225 
1226  get_limits(s, 16*mb_x, 16*mb_y);
1227  c->skip=0;
1228 
1229  P_LEFT[0] = s->p_mv_table[xy + 1][0];
1230  P_LEFT[1] = s->p_mv_table[xy + 1][1];
1231 
1232  if(P_LEFT[0] < (c->xmin<<shift)) P_LEFT[0] = (c->xmin<<shift);
1233 
1234  /* special case for first line */
1235  if (s->first_slice_line) {
1236  c->pred_x= P_LEFT[0];
1237  c->pred_y= P_LEFT[1];
1238  P_TOP[0]= P_TOPRIGHT[0]= P_MEDIAN[0]=
1239  P_TOP[1]= P_TOPRIGHT[1]= P_MEDIAN[1]= 0; //FIXME
1240  } else {
1241  P_TOP[0] = s->p_mv_table[xy + s->mb_stride ][0];
1242  P_TOP[1] = s->p_mv_table[xy + s->mb_stride ][1];
1243  P_TOPRIGHT[0] = s->p_mv_table[xy + s->mb_stride - 1][0];
1244  P_TOPRIGHT[1] = s->p_mv_table[xy + s->mb_stride - 1][1];
1245  if(P_TOP[1] < (c->ymin<<shift)) P_TOP[1] = (c->ymin<<shift);
1246  if(P_TOPRIGHT[0] > (c->xmax<<shift)) P_TOPRIGHT[0]= (c->xmax<<shift);
1247  if(P_TOPRIGHT[1] < (c->ymin<<shift)) P_TOPRIGHT[1]= (c->ymin<<shift);
1248 
1249  P_MEDIAN[0]= mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
1250  P_MEDIAN[1]= mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
1251 
1252  c->pred_x = P_MEDIAN[0];
1253  c->pred_y = P_MEDIAN[1];
1254  }
1255 
1256  dmin = ff_epzs_motion_search(s, &mx, &my, P, 0, 0, s->p_mv_table, (1<<16)>>shift, 0, 16);
1257 
1258  s->p_mv_table[xy][0] = mx<<shift;
1259  s->p_mv_table[xy][1] = my<<shift;
1260 
1261  return dmin;
1262 }
1263 
1265  int mb_x, int mb_y, int16_t (*mv_table)[2], int ref_index, int f_code)
1266 {
1267  MotionEstContext * const c= &s->me;
1268  int mx, my, dmin;
1269  int P[10][2];
1270  const int shift= 1+s->quarter_sample;
1271  const int mot_stride = s->mb_stride;
1272  const int mot_xy = mb_y*mot_stride + mb_x;
1273  uint8_t * const mv_penalty= c->mv_penalty[f_code] + MAX_MV;
1274  int mv_scale;
1275 
1280 
1281  get_limits(s, 16*mb_x, 16*mb_y);
1282 
1283  switch(s->me_method) {
1284  case ME_ZERO:
1285  default:
1286  no_motion_search(s, &mx, &my);
1287  dmin = 0;
1288  mx-= mb_x*16;
1289  my-= mb_y*16;
1290  break;
1291  case ME_X1:
1292  case ME_EPZS:
1293  {
1294  P_LEFT[0] = mv_table[mot_xy - 1][0];
1295  P_LEFT[1] = mv_table[mot_xy - 1][1];
1296 
1297  if(P_LEFT[0] > (c->xmax<<shift)) P_LEFT[0] = (c->xmax<<shift);
1298 
1299  /* special case for first line */
1300  if (!s->first_slice_line) {
1301  P_TOP[0] = mv_table[mot_xy - mot_stride ][0];
1302  P_TOP[1] = mv_table[mot_xy - mot_stride ][1];
1303  P_TOPRIGHT[0] = mv_table[mot_xy - mot_stride + 1 ][0];
1304  P_TOPRIGHT[1] = mv_table[mot_xy - mot_stride + 1 ][1];
1305  if(P_TOP[1] > (c->ymax<<shift)) P_TOP[1]= (c->ymax<<shift);
1306  if(P_TOPRIGHT[0] < (c->xmin<<shift)) P_TOPRIGHT[0]= (c->xmin<<shift);
1307  if(P_TOPRIGHT[1] > (c->ymax<<shift)) P_TOPRIGHT[1]= (c->ymax<<shift);
1308 
1309  P_MEDIAN[0]= mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
1310  P_MEDIAN[1]= mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
1311  }
1312  c->pred_x= P_LEFT[0];
1313  c->pred_y= P_LEFT[1];
1314  }
1315 
1316  if(mv_table == s->b_forw_mv_table){
1317  mv_scale= (s->pb_time<<16) / (s->pp_time<<shift);
1318  }else{
1319  mv_scale= ((s->pb_time - s->pp_time)<<16) / (s->pp_time<<shift);
1320  }
1321 
1322  dmin = ff_epzs_motion_search(s, &mx, &my, P, 0, ref_index, s->p_mv_table, mv_scale, 0, 16);
1323 
1324  break;
1325  }
1326 
1327  dmin= c->sub_motion_search(s, &mx, &my, dmin, 0, ref_index, 0, 16);
1328 
1329  if(c->avctx->me_sub_cmp != c->avctx->mb_cmp && !c->skip)
1330  dmin= ff_get_mb_score(s, mx, my, 0, ref_index, 0, 16, 1);
1331 
1332 //printf("%d %d %d %d//", s->mb_x, s->mb_y, mx, my);
1333 // s->mb_type[mb_y*s->mb_width + mb_x]= mb_type;
1334  mv_table[mot_xy][0]= mx;
1335  mv_table[mot_xy][1]= my;
1336 
1337  return dmin;
1338 }
1339 
1340 static inline int check_bidir_mv(MpegEncContext * s,
1341  int motion_fx, int motion_fy,
1342  int motion_bx, int motion_by,
1343  int pred_fx, int pred_fy,
1344  int pred_bx, int pred_by,
1345  int size, int h)
1346 {
1347  //FIXME optimize?
1348  //FIXME better f_code prediction (max mv & distance)
1349  //FIXME pointers
1350  MotionEstContext * const c= &s->me;
1351  uint8_t * const mv_penalty_f= c->mv_penalty[s->f_code] + MAX_MV; // f_code of the prev frame
1352  uint8_t * const mv_penalty_b= c->mv_penalty[s->b_code] + MAX_MV; // f_code of the prev frame
1353  int stride= c->stride;
1354  uint8_t *dest_y = c->scratchpad;
1355  uint8_t *ptr;
1356  int dxy;
1357  int src_x, src_y;
1358  int fbmin;
1359  uint8_t **src_data= c->src[0];
1360  uint8_t **ref_data= c->ref[0];
1361  uint8_t **ref2_data= c->ref[2];
1362 
1363  if(s->quarter_sample){
1364  dxy = ((motion_fy & 3) << 2) | (motion_fx & 3);
1365  src_x = motion_fx >> 2;
1366  src_y = motion_fy >> 2;
1367 
1368  ptr = ref_data[0] + (src_y * stride) + src_x;
1369  s->dsp.put_qpel_pixels_tab[0][dxy](dest_y , ptr , stride);
1370 
1371  dxy = ((motion_by & 3) << 2) | (motion_bx & 3);
1372  src_x = motion_bx >> 2;
1373  src_y = motion_by >> 2;
1374 
1375  ptr = ref2_data[0] + (src_y * stride) + src_x;
1376  s->dsp.avg_qpel_pixels_tab[size][dxy](dest_y , ptr , stride);
1377  }else{
1378  dxy = ((motion_fy & 1) << 1) | (motion_fx & 1);
1379  src_x = motion_fx >> 1;
1380  src_y = motion_fy >> 1;
1381 
1382  ptr = ref_data[0] + (src_y * stride) + src_x;
1383  s->dsp.put_pixels_tab[size][dxy](dest_y , ptr , stride, h);
1384 
1385  dxy = ((motion_by & 1) << 1) | (motion_bx & 1);
1386  src_x = motion_bx >> 1;
1387  src_y = motion_by >> 1;
1388 
1389  ptr = ref2_data[0] + (src_y * stride) + src_x;
1390  s->dsp.avg_pixels_tab[size][dxy](dest_y , ptr , stride, h);
1391  }
1392 
1393  fbmin = (mv_penalty_f[motion_fx-pred_fx] + mv_penalty_f[motion_fy-pred_fy])*c->mb_penalty_factor
1394  +(mv_penalty_b[motion_bx-pred_bx] + mv_penalty_b[motion_by-pred_by])*c->mb_penalty_factor
1395  + s->dsp.mb_cmp[size](s, src_data[0], dest_y, stride, h); //FIXME new_pic
1396 
1397  if(c->avctx->mb_cmp&FF_CMP_CHROMA){
1398  }
1399  //FIXME CHROMA !!!
1400 
1401  return fbmin;
1402 }
1403 
1404 /* refine the bidir vectors in hq mode and return the score in both lq & hq mode*/
1405 static inline int bidir_refine(MpegEncContext * s, int mb_x, int mb_y)
1406 {
1407  MotionEstContext * const c= &s->me;
1408  const int mot_stride = s->mb_stride;
1409  const int xy = mb_y *mot_stride + mb_x;
1410  int fbmin;
1411  int pred_fx= s->b_bidir_forw_mv_table[xy-1][0];
1412  int pred_fy= s->b_bidir_forw_mv_table[xy-1][1];
1413  int pred_bx= s->b_bidir_back_mv_table[xy-1][0];
1414  int pred_by= s->b_bidir_back_mv_table[xy-1][1];
1415  int motion_fx= s->b_bidir_forw_mv_table[xy][0]= s->b_forw_mv_table[xy][0];
1416  int motion_fy= s->b_bidir_forw_mv_table[xy][1]= s->b_forw_mv_table[xy][1];
1417  int motion_bx= s->b_bidir_back_mv_table[xy][0]= s->b_back_mv_table[xy][0];
1418  int motion_by= s->b_bidir_back_mv_table[xy][1]= s->b_back_mv_table[xy][1];
1419  const int flags= c->sub_flags;
1420  const int qpel= flags&FLAG_QPEL;
1421  const int shift= 1+qpel;
1422  const int xmin= c->xmin<<shift;
1423  const int ymin= c->ymin<<shift;
1424  const int xmax= c->xmax<<shift;
1425  const int ymax= c->ymax<<shift;
1426 #define HASH(fx,fy,bx,by) ((fx)+17*(fy)+63*(bx)+117*(by))
1427 #define HASH8(fx,fy,bx,by) ((uint8_t)HASH(fx,fy,bx,by))
1428  int hashidx= HASH(motion_fx,motion_fy, motion_bx, motion_by);
1429  uint8_t map[256];
1430 
1431  memset(map,0,sizeof(map));
1432  map[hashidx&255] = 1;
1433 
1434  fbmin= check_bidir_mv(s, motion_fx, motion_fy,
1435  motion_bx, motion_by,
1436  pred_fx, pred_fy,
1437  pred_bx, pred_by,
1438  0, 16);
1439 
1440  if(s->avctx->bidir_refine){
1441  int end;
1442  static const uint8_t limittab[5]={0,8,32,64,80};
1443  const int limit= limittab[s->avctx->bidir_refine];
1444  static const int8_t vect[][4]={
1445 { 0, 0, 0, 1}, { 0, 0, 0,-1}, { 0, 0, 1, 0}, { 0, 0,-1, 0}, { 0, 1, 0, 0}, { 0,-1, 0, 0}, { 1, 0, 0, 0}, {-1, 0, 0, 0},
1446 
1447 { 0, 0, 1, 1}, { 0, 0,-1,-1}, { 0, 1, 1, 0}, { 0,-1,-1, 0}, { 1, 1, 0, 0}, {-1,-1, 0, 0}, { 1, 0, 0, 1}, {-1, 0, 0,-1},
1448 { 0, 1, 0, 1}, { 0,-1, 0,-1}, { 1, 0, 1, 0}, {-1, 0,-1, 0},
1449 { 0, 0,-1, 1}, { 0, 0, 1,-1}, { 0,-1, 1, 0}, { 0, 1,-1, 0}, {-1, 1, 0, 0}, { 1,-1, 0, 0}, { 1, 0, 0,-1}, {-1, 0, 0, 1},
1450 { 0,-1, 0, 1}, { 0, 1, 0,-1}, {-1, 0, 1, 0}, { 1, 0,-1, 0},
1451 
1452 { 0, 1, 1, 1}, { 0,-1,-1,-1}, { 1, 1, 1, 0}, {-1,-1,-1, 0}, { 1, 1, 0, 1}, {-1,-1, 0,-1}, { 1, 0, 1, 1}, {-1, 0,-1,-1},
1453 { 0,-1, 1, 1}, { 0, 1,-1,-1}, {-1, 1, 1, 0}, { 1,-1,-1, 0}, { 1, 1, 0,-1}, {-1,-1, 0, 1}, { 1, 0,-1, 1}, {-1, 0, 1,-1},
1454 { 0, 1,-1, 1}, { 0,-1, 1,-1}, { 1,-1, 1, 0}, {-1, 1,-1, 0}, {-1, 1, 0, 1}, { 1,-1, 0,-1}, { 1, 0, 1,-1}, {-1, 0,-1, 1},
1455 { 0, 1, 1,-1}, { 0,-1,-1, 1}, { 1, 1,-1, 0}, {-1,-1, 1, 0}, { 1,-1, 0, 1}, {-1, 1, 0,-1}, {-1, 0, 1, 1}, { 1, 0,-1,-1},
1456 
1457 { 1, 1, 1, 1}, {-1,-1,-1,-1},
1458 { 1, 1, 1,-1}, {-1,-1,-1, 1}, { 1, 1,-1, 1}, {-1,-1, 1,-1}, { 1,-1, 1, 1}, {-1, 1,-1,-1}, {-1, 1, 1, 1}, { 1,-1,-1,-1},
1459 { 1, 1,-1,-1}, {-1,-1, 1, 1}, { 1,-1,-1, 1}, {-1, 1, 1,-1}, { 1,-1, 1,-1}, {-1, 1,-1, 1},
1460  };
1461  static const uint8_t hash[]={
1462 HASH8( 0, 0, 0, 1), HASH8( 0, 0, 0,-1), HASH8( 0, 0, 1, 0), HASH8( 0, 0,-1, 0), HASH8( 0, 1, 0, 0), HASH8( 0,-1, 0, 0), HASH8( 1, 0, 0, 0), HASH8(-1, 0, 0, 0),
1463 
1464 HASH8( 0, 0, 1, 1), HASH8( 0, 0,-1,-1), HASH8( 0, 1, 1, 0), HASH8( 0,-1,-1, 0), HASH8( 1, 1, 0, 0), HASH8(-1,-1, 0, 0), HASH8( 1, 0, 0, 1), HASH8(-1, 0, 0,-1),
1465 HASH8( 0, 1, 0, 1), HASH8( 0,-1, 0,-1), HASH8( 1, 0, 1, 0), HASH8(-1, 0,-1, 0),
1466 HASH8( 0, 0,-1, 1), HASH8( 0, 0, 1,-1), HASH8( 0,-1, 1, 0), HASH8( 0, 1,-1, 0), HASH8(-1, 1, 0, 0), HASH8( 1,-1, 0, 0), HASH8( 1, 0, 0,-1), HASH8(-1, 0, 0, 1),
1467 HASH8( 0,-1, 0, 1), HASH8( 0, 1, 0,-1), HASH8(-1, 0, 1, 0), HASH8( 1, 0,-1, 0),
1468 
1469 HASH8( 0, 1, 1, 1), HASH8( 0,-1,-1,-1), HASH8( 1, 1, 1, 0), HASH8(-1,-1,-1, 0), HASH8( 1, 1, 0, 1), HASH8(-1,-1, 0,-1), HASH8( 1, 0, 1, 1), HASH8(-1, 0,-1,-1),
1470 HASH8( 0,-1, 1, 1), HASH8( 0, 1,-1,-1), HASH8(-1, 1, 1, 0), HASH8( 1,-1,-1, 0), HASH8( 1, 1, 0,-1), HASH8(-1,-1, 0, 1), HASH8( 1, 0,-1, 1), HASH8(-1, 0, 1,-1),
1471 HASH8( 0, 1,-1, 1), HASH8( 0,-1, 1,-1), HASH8( 1,-1, 1, 0), HASH8(-1, 1,-1, 0), HASH8(-1, 1, 0, 1), HASH8( 1,-1, 0,-1), HASH8( 1, 0, 1,-1), HASH8(-1, 0,-1, 1),
1472 HASH8( 0, 1, 1,-1), HASH8( 0,-1,-1, 1), HASH8( 1, 1,-1, 0), HASH8(-1,-1, 1, 0), HASH8( 1,-1, 0, 1), HASH8(-1, 1, 0,-1), HASH8(-1, 0, 1, 1), HASH8( 1, 0,-1,-1),
1473 
1474 HASH8( 1, 1, 1, 1), HASH8(-1,-1,-1,-1),
1475 HASH8( 1, 1, 1,-1), HASH8(-1,-1,-1, 1), HASH8( 1, 1,-1, 1), HASH8(-1,-1, 1,-1), HASH8( 1,-1, 1, 1), HASH8(-1, 1,-1,-1), HASH8(-1, 1, 1, 1), HASH8( 1,-1,-1,-1),
1476 HASH8( 1, 1,-1,-1), HASH8(-1,-1, 1, 1), HASH8( 1,-1,-1, 1), HASH8(-1, 1, 1,-1), HASH8( 1,-1, 1,-1), HASH8(-1, 1,-1, 1),
1477 };
1478 
1479 #define CHECK_BIDIR(fx,fy,bx,by)\
1480  if( !map[(hashidx+HASH(fx,fy,bx,by))&255]\
1481  &&(fx<=0 || motion_fx+fx<=xmax) && (fy<=0 || motion_fy+fy<=ymax) && (bx<=0 || motion_bx+bx<=xmax) && (by<=0 || motion_by+by<=ymax)\
1482  &&(fx>=0 || motion_fx+fx>=xmin) && (fy>=0 || motion_fy+fy>=ymin) && (bx>=0 || motion_bx+bx>=xmin) && (by>=0 || motion_by+by>=ymin)){\
1483  int score;\
1484  map[(hashidx+HASH(fx,fy,bx,by))&255] = 1;\
1485  score= check_bidir_mv(s, motion_fx+fx, motion_fy+fy, motion_bx+bx, motion_by+by, pred_fx, pred_fy, pred_bx, pred_by, 0, 16);\
1486  if(score < fbmin){\
1487  hashidx += HASH(fx,fy,bx,by);\
1488  fbmin= score;\
1489  motion_fx+=fx;\
1490  motion_fy+=fy;\
1491  motion_bx+=bx;\
1492  motion_by+=by;\
1493  end=0;\
1494  }\
1495  }
1496 #define CHECK_BIDIR2(a,b,c,d)\
1497 CHECK_BIDIR(a,b,c,d)\
1498 CHECK_BIDIR(-(a),-(b),-(c),-(d))
1499 
1500  do{
1501  int i;
1502  int borderdist=0;
1503  end=1;
1504 
1505  CHECK_BIDIR2(0,0,0,1)
1506  CHECK_BIDIR2(0,0,1,0)
1507  CHECK_BIDIR2(0,1,0,0)
1508  CHECK_BIDIR2(1,0,0,0)
1509 
1510  for(i=8; i<limit; i++){
1511  int fx= motion_fx+vect[i][0];
1512  int fy= motion_fy+vect[i][1];
1513  int bx= motion_bx+vect[i][2];
1514  int by= motion_by+vect[i][3];
1515  if(borderdist<=0){
1516  int a= (xmax - FFMAX(fx,bx))|(FFMIN(fx,bx) - xmin);
1517  int b= (ymax - FFMAX(fy,by))|(FFMIN(fy,by) - ymin);
1518  if((a|b) < 0)
1519  map[(hashidx+hash[i])&255] = 1;
1520  }
1521  if(!map[(hashidx+hash[i])&255]){
1522  int score;
1523  map[(hashidx+hash[i])&255] = 1;
1524  score= check_bidir_mv(s, fx, fy, bx, by, pred_fx, pred_fy, pred_bx, pred_by, 0, 16);
1525  if(score < fbmin){
1526  hashidx += hash[i];
1527  fbmin= score;
1528  motion_fx=fx;
1529  motion_fy=fy;
1530  motion_bx=bx;
1531  motion_by=by;
1532  end=0;
1533  borderdist--;
1534  if(borderdist<=0){
1535  int a= FFMIN(xmax - FFMAX(fx,bx), FFMIN(fx,bx) - xmin);
1536  int b= FFMIN(ymax - FFMAX(fy,by), FFMIN(fy,by) - ymin);
1537  borderdist= FFMIN(a,b);
1538  }
1539  }
1540  }
1541  }
1542  }while(!end);
1543  }
1544 
1545  s->b_bidir_forw_mv_table[xy][0]= motion_fx;
1546  s->b_bidir_forw_mv_table[xy][1]= motion_fy;
1547  s->b_bidir_back_mv_table[xy][0]= motion_bx;
1548  s->b_bidir_back_mv_table[xy][1]= motion_by;
1549 
1550  return fbmin;
1551 }
1552 
1553 static inline int direct_search(MpegEncContext * s, int mb_x, int mb_y)
1554 {
1555  MotionEstContext * const c= &s->me;
1556  int P[10][2];
1557  const int mot_stride = s->mb_stride;
1558  const int mot_xy = mb_y*mot_stride + mb_x;
1559  const int shift= 1+s->quarter_sample;
1560  int dmin, i;
1561  const int time_pp= s->pp_time;
1562  const int time_pb= s->pb_time;
1563  int mx, my, xmin, xmax, ymin, ymax;
1564  int16_t (*mv_table)[2]= s->b_direct_mv_table;
1565 
1566  c->current_mv_penalty= c->mv_penalty[1] + MAX_MV;
1567  ymin= xmin=(-32)>>shift;
1568  ymax= xmax= 31>>shift;
1569 
1570  if (IS_8X8(s->next_picture.f.mb_type[mot_xy])) {
1571  s->mv_type= MV_TYPE_8X8;
1572  }else{
1573  s->mv_type= MV_TYPE_16X16;
1574  }
1575 
1576  for(i=0; i<4; i++){
1577  int index= s->block_index[i];
1578  int min, max;
1579 
1580  c->co_located_mv[i][0] = s->next_picture.f.motion_val[0][index][0];
1581  c->co_located_mv[i][1] = s->next_picture.f.motion_val[0][index][1];
1582  c->direct_basis_mv[i][0]= c->co_located_mv[i][0]*time_pb/time_pp + ((i& 1)<<(shift+3));
1583  c->direct_basis_mv[i][1]= c->co_located_mv[i][1]*time_pb/time_pp + ((i>>1)<<(shift+3));
1584 // c->direct_basis_mv[1][i][0]= c->co_located_mv[i][0]*(time_pb - time_pp)/time_pp + ((i &1)<<(shift+3);
1585 // c->direct_basis_mv[1][i][1]= c->co_located_mv[i][1]*(time_pb - time_pp)/time_pp + ((i>>1)<<(shift+3);
1586 
1587  max= FFMAX(c->direct_basis_mv[i][0], c->direct_basis_mv[i][0] - c->co_located_mv[i][0])>>shift;
1588  min= FFMIN(c->direct_basis_mv[i][0], c->direct_basis_mv[i][0] - c->co_located_mv[i][0])>>shift;
1589  max+= 16*mb_x + 1; // +-1 is for the simpler rounding
1590  min+= 16*mb_x - 1;
1591  xmax= FFMIN(xmax, s->width - max);
1592  xmin= FFMAX(xmin, - 16 - min);
1593 
1594  max= FFMAX(c->direct_basis_mv[i][1], c->direct_basis_mv[i][1] - c->co_located_mv[i][1])>>shift;
1595  min= FFMIN(c->direct_basis_mv[i][1], c->direct_basis_mv[i][1] - c->co_located_mv[i][1])>>shift;
1596  max+= 16*mb_y + 1; // +-1 is for the simpler rounding
1597  min+= 16*mb_y - 1;
1598  ymax= FFMIN(ymax, s->height - max);
1599  ymin= FFMAX(ymin, - 16 - min);
1600 
1601  if(s->mv_type == MV_TYPE_16X16) break;
1602  }
1603 
1604  assert(xmax <= 15 && ymax <= 15 && xmin >= -16 && ymin >= -16);
1605 
1606  if(xmax < 0 || xmin >0 || ymax < 0 || ymin > 0){
1607  s->b_direct_mv_table[mot_xy][0]= 0;
1608  s->b_direct_mv_table[mot_xy][1]= 0;
1609 
1610  return 256*256*256*64;
1611  }
1612 
1613  c->xmin= xmin;
1614  c->ymin= ymin;
1615  c->xmax= xmax;
1616  c->ymax= ymax;
1617  c->flags |= FLAG_DIRECT;
1618  c->sub_flags |= FLAG_DIRECT;
1619  c->pred_x=0;
1620  c->pred_y=0;
1621 
1622  P_LEFT[0] = av_clip(mv_table[mot_xy - 1][0], xmin<<shift, xmax<<shift);
1623  P_LEFT[1] = av_clip(mv_table[mot_xy - 1][1], ymin<<shift, ymax<<shift);
1624 
1625  /* special case for first line */
1626  if (!s->first_slice_line) { //FIXME maybe allow this over thread boundary as it is clipped
1627  P_TOP[0] = av_clip(mv_table[mot_xy - mot_stride ][0], xmin<<shift, xmax<<shift);
1628  P_TOP[1] = av_clip(mv_table[mot_xy - mot_stride ][1], ymin<<shift, ymax<<shift);
1629  P_TOPRIGHT[0] = av_clip(mv_table[mot_xy - mot_stride + 1 ][0], xmin<<shift, xmax<<shift);
1630  P_TOPRIGHT[1] = av_clip(mv_table[mot_xy - mot_stride + 1 ][1], ymin<<shift, ymax<<shift);
1631 
1632  P_MEDIAN[0]= mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
1633  P_MEDIAN[1]= mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
1634  }
1635 
1636  dmin = ff_epzs_motion_search(s, &mx, &my, P, 0, 0, mv_table, 1<<(16-shift), 0, 16);
1637  if(c->sub_flags&FLAG_QPEL)
1638  dmin = qpel_motion_search(s, &mx, &my, dmin, 0, 0, 0, 16);
1639  else
1640  dmin = hpel_motion_search(s, &mx, &my, dmin, 0, 0, 0, 16);
1641 
1642  if(c->avctx->me_sub_cmp != c->avctx->mb_cmp && !c->skip)
1643  dmin= ff_get_mb_score(s, mx, my, 0, 0, 0, 16, 1);
1644 
1645  get_limits(s, 16*mb_x, 16*mb_y); //restore c->?min/max, maybe not needed
1646 
1647  mv_table[mot_xy][0]= mx;
1648  mv_table[mot_xy][1]= my;
1649  c->flags &= ~FLAG_DIRECT;
1650  c->sub_flags &= ~FLAG_DIRECT;
1651 
1652  return dmin;
1653 }
1654 
1656  int mb_x, int mb_y)
1657 {
1658  MotionEstContext * const c= &s->me;
1659  const int penalty_factor= c->mb_penalty_factor;
1660  int fmin, bmin, dmin, fbmin, bimin, fimin;
1661  int type=0;
1662  const int xy = mb_y*s->mb_stride + mb_x;
1664  s->next_picture.f.data, 16 * mb_x, 16 * mb_y, 2);
1665 
1666  get_limits(s, 16*mb_x, 16*mb_y);
1667 
1668  c->skip=0;
1669 
1670  if (s->codec_id == CODEC_ID_MPEG4 && s->next_picture.f.mbskip_table[xy]) {
1671  int score= direct_search(s, mb_x, mb_y); //FIXME just check 0,0
1672 
1673  score= ((unsigned)(score*score + 128*256))>>16;
1674  c->mc_mb_var_sum_temp += score;
1675  s->current_picture.mc_mb_var[mb_y*s->mb_stride + mb_x] = score; //FIXME use SSE
1676  s->mb_type[mb_y*s->mb_stride + mb_x]= CANDIDATE_MB_TYPE_DIRECT0;
1677 
1678  return;
1679  }
1680 
1681  if(c->avctx->me_threshold){
1682  int vard= check_input_motion(s, mb_x, mb_y, 0);
1683 
1684  if((vard+128)>>8 < c->avctx->me_threshold){
1685 // pix = c->src[0][0];
1686 // sum = s->dsp.pix_sum(pix, s->linesize);
1687 // varc = s->dsp.pix_norm1(pix, s->linesize) - (((unsigned)(sum*sum))>>8) + 500;
1688 
1689 // pic->mb_var [s->mb_stride * mb_y + mb_x] = (varc+128)>>8;
1690  s->current_picture.mc_mb_var[s->mb_stride * mb_y + mb_x] = (vard+128)>>8;
1691 /* pic->mb_mean [s->mb_stride * mb_y + mb_x] = (sum+128)>>8;
1692  c->mb_var_sum_temp += (varc+128)>>8;*/
1693  c->mc_mb_var_sum_temp += (vard+128)>>8;
1694 /* if (vard <= 64<<8 || vard < varc) {
1695  c->scene_change_score+= ff_sqrt(vard) - ff_sqrt(varc);
1696  }else{
1697  c->scene_change_score+= s->qscale * s->avctx->scenechange_factor;
1698  }*/
1699  return;
1700  }
1701  if((vard+128)>>8 < c->avctx->mb_threshold){
1702  type= s->mb_type[mb_y*s->mb_stride + mb_x];
1703  if(type == CANDIDATE_MB_TYPE_DIRECT){
1704  direct_search(s, mb_x, mb_y);
1705  }
1706  if(type == CANDIDATE_MB_TYPE_FORWARD || type == CANDIDATE_MB_TYPE_BIDIR){
1707  c->skip=0;
1708  ff_estimate_motion_b(s, mb_x, mb_y, s->b_forw_mv_table, 0, s->f_code);
1709  }
1710  if(type == CANDIDATE_MB_TYPE_BACKWARD || type == CANDIDATE_MB_TYPE_BIDIR){
1711  c->skip=0;
1712  ff_estimate_motion_b(s, mb_x, mb_y, s->b_back_mv_table, 2, s->b_code);
1713  }
1715  c->skip=0;
1717  interlaced_search(s, 0,
1719  s->b_forw_mv_table[xy][0], s->b_forw_mv_table[xy][1], 1);
1720  }
1722  c->skip=0;
1724  interlaced_search(s, 2,
1726  s->b_back_mv_table[xy][0], s->b_back_mv_table[xy][1], 1);
1727  }
1728  return;
1729  }
1730  }
1731 
1732  if (s->codec_id == CODEC_ID_MPEG4)
1733  dmin= direct_search(s, mb_x, mb_y);
1734  else
1735  dmin= INT_MAX;
1736 //FIXME penalty stuff for non mpeg4
1737  c->skip=0;
1738  fmin= ff_estimate_motion_b(s, mb_x, mb_y, s->b_forw_mv_table, 0, s->f_code) + 3*penalty_factor;
1739 
1740  c->skip=0;
1741  bmin= ff_estimate_motion_b(s, mb_x, mb_y, s->b_back_mv_table, 2, s->b_code) + 2*penalty_factor;
1742 //printf(" %d %d ", s->b_forw_mv_table[xy][0], s->b_forw_mv_table[xy][1]);
1743 
1744  c->skip=0;
1745  fbmin= bidir_refine(s, mb_x, mb_y) + penalty_factor;
1746 //printf("%d %d %d %d\n", dmin, fmin, bmin, fbmin);
1747 
1749 //FIXME mb type penalty
1750  c->skip=0;
1752  fimin= interlaced_search(s, 0,
1754  s->b_forw_mv_table[xy][0], s->b_forw_mv_table[xy][1], 0);
1756  bimin= interlaced_search(s, 2,
1758  s->b_back_mv_table[xy][0], s->b_back_mv_table[xy][1], 0);
1759  }else
1760  fimin= bimin= INT_MAX;
1761 
1762  {
1763  int score= fmin;
1765 
1766  if (dmin <= score){
1767  score = dmin;
1768  type = CANDIDATE_MB_TYPE_DIRECT;
1769  }
1770  if(bmin<score){
1771  score=bmin;
1773  }
1774  if(fbmin<score){
1775  score=fbmin;
1777  }
1778  if(fimin<score){
1779  score=fimin;
1781  }
1782  if(bimin<score){
1783  score=bimin;
1785  }
1786 
1787  score= ((unsigned)(score*score + 128*256))>>16;
1788  c->mc_mb_var_sum_temp += score;
1789  s->current_picture.mc_mb_var[mb_y*s->mb_stride + mb_x] = score; //FIXME use SSE
1790  }
1791 
1794  if(fimin < INT_MAX)
1796  if(bimin < INT_MAX)
1798  if(fimin < INT_MAX && bimin < INT_MAX){
1799  type |= CANDIDATE_MB_TYPE_BIDIR_I;
1800  }
1801  //FIXME something smarter
1802  if(dmin>256*256*16) type&= ~CANDIDATE_MB_TYPE_DIRECT; //do not try direct mode if it is invalid for this MB
1803  if(s->codec_id == CODEC_ID_MPEG4 && type&CANDIDATE_MB_TYPE_DIRECT && s->flags&CODEC_FLAG_MV0 && *(uint32_t*)s->b_direct_mv_table[xy])
1804  type |= CANDIDATE_MB_TYPE_DIRECT0;
1805  }
1806 
1807  s->mb_type[mb_y*s->mb_stride + mb_x]= type;
1808 }
1809 
1810 /* find best f_code for ME which do unlimited searches */
1811 int ff_get_best_fcode(MpegEncContext * s, int16_t (*mv_table)[2], int type)
1812 {
1813  if(s->me_method>=ME_EPZS){
1814  int score[8];
1815  int i, y, range= s->avctx->me_range ? s->avctx->me_range : (INT_MAX/2);
1816  uint8_t * fcode_tab= s->fcode_tab;
1817  int best_fcode=-1;
1818  int best_score=-10000000;
1819 
1820  if(s->msmpeg4_version)
1821  range= FFMIN(range, 16);
1823  range= FFMIN(range, 256);
1824 
1825  for(i=0; i<8; i++) score[i]= s->mb_num*(8-i);
1826 
1827  for(y=0; y<s->mb_height; y++){
1828  int x;
1829  int xy= y*s->mb_stride;
1830  for(x=0; x<s->mb_width; x++){
1831  if(s->mb_type[xy] & type){
1832  int mx= mv_table[xy][0];
1833  int my= mv_table[xy][1];
1834  int fcode= FFMAX(fcode_tab[mx + MAX_MV],
1835  fcode_tab[my + MAX_MV]);
1836  int j;
1837 
1838  if(mx >= range || mx < -range ||
1839  my >= range || my < -range)
1840  continue;
1841 
1842  for(j=0; j<fcode && j<8; j++){
1844  score[j]-= 170;
1845  }
1846  }
1847  xy++;
1848  }
1849  }
1850 
1851  for(i=1; i<8; i++){
1852  if(score[i] > best_score){
1853  best_score= score[i];
1854  best_fcode= i;
1855  }
1856 // printf("%d %d\n", i, score[i]);
1857  }
1858 
1859 // printf("fcode: %d type: %d\n", i, s->pict_type);
1860  return best_fcode;
1861 /* for(i=0; i<=MAX_FCODE; i++){
1862  printf("%d ", mv_num[i]);
1863  }
1864  printf("\n");*/
1865  }else{
1866  return 1;
1867  }
1868 }
1869 
1871 {
1872  MotionEstContext * const c= &s->me;
1873  const int f_code= s->f_code;
1874  int y, range;
1875  assert(s->pict_type==AV_PICTURE_TYPE_P);
1876 
1877  range = (((s->out_format == FMT_MPEG1 || s->msmpeg4_version) ? 8 : 16) << f_code);
1878 
1879  assert(range <= 16 || !s->msmpeg4_version);
1880  assert(range <=256 || !(s->codec_id == CODEC_ID_MPEG2VIDEO && s->avctx->strict_std_compliance >= FF_COMPLIANCE_NORMAL));
1881 
1882  if(c->avctx->me_range && range > c->avctx->me_range) range= c->avctx->me_range;
1883 
1884 //printf("%d no:%d %d//\n", clip, noclip, f_code);
1885  if(s->flags&CODEC_FLAG_4MV){
1886  const int wrap= s->b8_stride;
1887 
1888  /* clip / convert to intra 8x8 type MVs */
1889  for(y=0; y<s->mb_height; y++){
1890  int xy= y*2*wrap;
1891  int i= y*s->mb_stride;
1892  int x;
1893 
1894  for(x=0; x<s->mb_width; x++){
1896  int block;
1897  for(block=0; block<4; block++){
1898  int off= (block& 1) + (block>>1)*wrap;
1899  int mx = s->current_picture.f.motion_val[0][ xy + off ][0];
1900  int my = s->current_picture.f.motion_val[0][ xy + off ][1];
1901 
1902  if( mx >=range || mx <-range
1903  || my >=range || my <-range){
1904  s->mb_type[i] &= ~CANDIDATE_MB_TYPE_INTER4V;
1907  }
1908  }
1909  }
1910  xy+=2;
1911  i++;
1912  }
1913  }
1914  }
1915 }
1916 
1921 void ff_fix_long_mvs(MpegEncContext * s, uint8_t *field_select_table, int field_select,
1922  int16_t (*mv_table)[2], int f_code, int type, int truncate)
1923 {
1924  MotionEstContext * const c= &s->me;
1925  int y, h_range, v_range;
1926 
1927  // RAL: 8 in MPEG-1, 16 in MPEG-4
1928  int range = (((s->out_format == FMT_MPEG1 || s->msmpeg4_version) ? 8 : 16) << f_code);
1929 
1930  if(c->avctx->me_range && range > c->avctx->me_range) range= c->avctx->me_range;
1931 
1932  h_range= range;
1933  v_range= field_select_table ? range>>1 : range;
1934 
1935  /* clip / convert to intra 16x16 type MVs */
1936  for(y=0; y<s->mb_height; y++){
1937  int x;
1938  int xy= y*s->mb_stride;
1939  for(x=0; x<s->mb_width; x++){
1940  if (s->mb_type[xy] & type){ // RAL: "type" test added...
1941  if(field_select_table==NULL || field_select_table[xy] == field_select){
1942  if( mv_table[xy][0] >=h_range || mv_table[xy][0] <-h_range
1943  || mv_table[xy][1] >=v_range || mv_table[xy][1] <-v_range){
1944 
1945  if(truncate){
1946  if (mv_table[xy][0] > h_range-1) mv_table[xy][0]= h_range-1;
1947  else if(mv_table[xy][0] < -h_range ) mv_table[xy][0]= -h_range;
1948  if (mv_table[xy][1] > v_range-1) mv_table[xy][1]= v_range-1;
1949  else if(mv_table[xy][1] < -v_range ) mv_table[xy][1]= -v_range;
1950  }else{
1951  s->mb_type[xy] &= ~type;
1952  s->mb_type[xy] |= CANDIDATE_MB_TYPE_INTRA;
1953  mv_table[xy][0]=
1954  mv_table[xy][1]= 0;
1955  }
1956  }
1957  }
1958  }
1959  xy++;
1960  }
1961  }
1962 }