FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
cavs.c
Go to the documentation of this file.
1 /*
2  * Chinese AVS video (AVS1-P2, JiZhun profile) decoder.
3  * Copyright (c) 2006 Stefan Gehrer <stefan.gehrer@gmx.de>
4  *
5  * This file is part of FFmpeg.
6  *
7  * FFmpeg is free software; you can redistribute it and/or
8  * modify it under the terms of the GNU Lesser General Public
9  * License as published by the Free Software Foundation; either
10  * version 2.1 of the License, or (at your option) any later version.
11  *
12  * FFmpeg is distributed in the hope that it will be useful,
13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15  * Lesser General Public License for more details.
16  *
17  * You should have received a copy of the GNU Lesser General Public
18  * License along with FFmpeg; if not, write to the Free Software
19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20  */
21 
22 /**
23  * @file
24  * Chinese AVS video (AVS1-P2, JiZhun profile) decoder
25  * @author Stefan Gehrer <stefan.gehrer@gmx.de>
26  */
27 
28 #include "avcodec.h"
29 #include "get_bits.h"
30 #include "golomb.h"
31 #include "h264chroma.h"
32 #include "idctdsp.h"
33 #include "internal.h"
34 #include "mathops.h"
35 #include "qpeldsp.h"
36 #include "cavs.h"
37 
38 static const uint8_t alpha_tab[64] = {
39  0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 2, 2, 2, 3, 3,
40  4, 4, 5, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 16, 18, 20,
41  22, 24, 26, 28, 30, 33, 33, 35, 35, 36, 37, 37, 39, 39, 42, 44,
42  46, 48, 50, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64
43 };
44 
45 static const uint8_t beta_tab[64] = {
46  0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2,
47  2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 6, 6,
48  6, 7, 7, 7, 8, 8, 8, 9, 9, 10, 10, 11, 11, 12, 13, 14,
49  15, 16, 17, 18, 19, 20, 21, 22, 23, 23, 24, 24, 25, 25, 26, 27
50 };
51 
52 static const uint8_t tc_tab[64] = {
53  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
54  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2,
55  2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4,
56  5, 5, 5, 6, 6, 6, 7, 7, 7, 7, 8, 8, 8, 9, 9, 9
57 };
58 
59 /** mark block as unavailable, i.e. out of picture
60  * or not yet decoded */
61 static const cavs_vector un_mv = { 0, 0, 1, NOT_AVAIL };
62 
63 static const int8_t left_modifier_l[8] = { 0, -1, 6, -1, -1, 7, 6, 7 };
64 static const int8_t top_modifier_l[8] = { -1, 1, 5, -1, -1, 5, 7, 7 };
65 static const int8_t left_modifier_c[7] = { 5, -1, 2, -1, 6, 5, 6 };
66 static const int8_t top_modifier_c[7] = { 4, 1, -1, -1, 4, 6, 6 };
67 
68 /*****************************************************************************
69  *
70  * in-loop deblocking filter
71  *
72  ****************************************************************************/
73 
74 static inline int get_bs(cavs_vector *mvP, cavs_vector *mvQ, int b)
75 {
76  if ((mvP->ref == REF_INTRA) || (mvQ->ref == REF_INTRA))
77  return 2;
78  if((abs(mvP->x - mvQ->x) >= 4) ||
79  (abs(mvP->y - mvQ->y) >= 4) ||
80  (mvP->ref != mvQ->ref))
81  return 1;
82  if (b) {
83  mvP += MV_BWD_OFFS;
84  mvQ += MV_BWD_OFFS;
85  if((abs(mvP->x - mvQ->x) >= 4) ||
86  (abs(mvP->y - mvQ->y) >= 4) ||
87  (mvP->ref != mvQ->ref))
88  return 1;
89  }
90  return 0;
91 }
92 
93 #define SET_PARAMS \
94  alpha = alpha_tab[av_clip_uintp2(qp_avg + h->alpha_offset, 6)]; \
95  beta = beta_tab[av_clip_uintp2(qp_avg + h->beta_offset, 6)]; \
96  tc = tc_tab[av_clip_uintp2(qp_avg + h->alpha_offset, 6)];
97 
98 /**
99  * in-loop deblocking filter for a single macroblock
100  *
101  * boundary strength (bs) mapping:
102  *
103  * --4---5--
104  * 0 2 |
105  * | 6 | 7 |
106  * 1 3 |
107  * ---------
108  *
109  */
110 void ff_cavs_filter(AVSContext *h, enum cavs_mb mb_type)
111 {
112  uint8_t bs[8];
113  int qp_avg, alpha, beta, tc;
114  int i;
115 
116  /* save un-deblocked lines */
117  h->topleft_border_y = h->top_border_y[h->mbx * 16 + 15];
118  h->topleft_border_u = h->top_border_u[h->mbx * 10 + 8];
119  h->topleft_border_v = h->top_border_v[h->mbx * 10 + 8];
120  memcpy(&h->top_border_y[h->mbx * 16], h->cy + 15 * h->l_stride, 16);
121  memcpy(&h->top_border_u[h->mbx * 10 + 1], h->cu + 7 * h->c_stride, 8);
122  memcpy(&h->top_border_v[h->mbx * 10 + 1], h->cv + 7 * h->c_stride, 8);
123  for (i = 0; i < 8; i++) {
124  h->left_border_y[i * 2 + 1] = *(h->cy + 15 + (i * 2 + 0) * h->l_stride);
125  h->left_border_y[i * 2 + 2] = *(h->cy + 15 + (i * 2 + 1) * h->l_stride);
126  h->left_border_u[i + 1] = *(h->cu + 7 + i * h->c_stride);
127  h->left_border_v[i + 1] = *(h->cv + 7 + i * h->c_stride);
128  }
129  if (!h->loop_filter_disable) {
130  /* determine bs */
131  if (mb_type == I_8X8)
132  memset(bs, 2, 8);
133  else {
134  memset(bs, 0, 8);
135  if (ff_cavs_partition_flags[mb_type] & SPLITV) {
136  bs[2] = get_bs(&h->mv[MV_FWD_X0], &h->mv[MV_FWD_X1], mb_type > P_8X8);
137  bs[3] = get_bs(&h->mv[MV_FWD_X2], &h->mv[MV_FWD_X3], mb_type > P_8X8);
138  }
139  if (ff_cavs_partition_flags[mb_type] & SPLITH) {
140  bs[6] = get_bs(&h->mv[MV_FWD_X0], &h->mv[MV_FWD_X2], mb_type > P_8X8);
141  bs[7] = get_bs(&h->mv[MV_FWD_X1], &h->mv[MV_FWD_X3], mb_type > P_8X8);
142  }
143  bs[0] = get_bs(&h->mv[MV_FWD_A1], &h->mv[MV_FWD_X0], mb_type > P_8X8);
144  bs[1] = get_bs(&h->mv[MV_FWD_A3], &h->mv[MV_FWD_X2], mb_type > P_8X8);
145  bs[4] = get_bs(&h->mv[MV_FWD_B2], &h->mv[MV_FWD_X0], mb_type > P_8X8);
146  bs[5] = get_bs(&h->mv[MV_FWD_B3], &h->mv[MV_FWD_X1], mb_type > P_8X8);
147  }
148  if (AV_RN64(bs)) {
149  if (h->flags & A_AVAIL) {
150  qp_avg = (h->qp + h->left_qp + 1) >> 1;
151  SET_PARAMS;
152  h->cdsp.cavs_filter_lv(h->cy, h->l_stride, alpha, beta, tc, bs[0], bs[1]);
153  qp_avg = (ff_cavs_chroma_qp[h->qp] + ff_cavs_chroma_qp[h->left_qp] + 1) >> 1;
154  SET_PARAMS;
155  h->cdsp.cavs_filter_cv(h->cu, h->c_stride, alpha, beta, tc, bs[0], bs[1]);
156  h->cdsp.cavs_filter_cv(h->cv, h->c_stride, alpha, beta, tc, bs[0], bs[1]);
157  }
158  qp_avg = h->qp;
159  SET_PARAMS;
160  h->cdsp.cavs_filter_lv(h->cy + 8, h->l_stride, alpha, beta, tc, bs[2], bs[3]);
161  h->cdsp.cavs_filter_lh(h->cy + 8 * h->l_stride, h->l_stride, alpha, beta, tc, bs[6], bs[7]);
162 
163  if (h->flags & B_AVAIL) {
164  qp_avg = (h->qp + h->top_qp[h->mbx] + 1) >> 1;
165  SET_PARAMS;
166  h->cdsp.cavs_filter_lh(h->cy, h->l_stride, alpha, beta, tc, bs[4], bs[5]);
167  qp_avg = (ff_cavs_chroma_qp[h->qp] + ff_cavs_chroma_qp[h->top_qp[h->mbx]] + 1) >> 1;
168  SET_PARAMS;
169  h->cdsp.cavs_filter_ch(h->cu, h->c_stride, alpha, beta, tc, bs[4], bs[5]);
170  h->cdsp.cavs_filter_ch(h->cv, h->c_stride, alpha, beta, tc, bs[4], bs[5]);
171  }
172  }
173  }
174  h->left_qp = h->qp;
175  h->top_qp[h->mbx] = h->qp;
176 }
177 
178 #undef SET_PARAMS
179 
180 /*****************************************************************************
181  *
182  * spatial intra prediction
183  *
184  ****************************************************************************/
185 
187  uint8_t **left, int block)
188 {
189  int i;
190 
191  switch (block) {
192  case 0:
193  *left = h->left_border_y;
194  h->left_border_y[0] = h->left_border_y[1];
195  memset(&h->left_border_y[17], h->left_border_y[16], 9);
196  memcpy(&top[1], &h->top_border_y[h->mbx * 16], 16);
197  top[17] = top[16];
198  top[0] = top[1];
199  if ((h->flags & A_AVAIL) && (h->flags & B_AVAIL))
200  h->left_border_y[0] = top[0] = h->topleft_border_y;
201  break;
202  case 1:
203  *left = h->intern_border_y;
204  for (i = 0; i < 8; i++)
205  h->intern_border_y[i + 1] = *(h->cy + 7 + i * h->l_stride);
206  memset(&h->intern_border_y[9], h->intern_border_y[8], 9);
207  h->intern_border_y[0] = h->intern_border_y[1];
208  memcpy(&top[1], &h->top_border_y[h->mbx * 16 + 8], 8);
209  if (h->flags & C_AVAIL)
210  memcpy(&top[9], &h->top_border_y[(h->mbx + 1) * 16], 8);
211  else
212  memset(&top[9], top[8], 9);
213  top[17] = top[16];
214  top[0] = top[1];
215  if (h->flags & B_AVAIL)
216  h->intern_border_y[0] = top[0] = h->top_border_y[h->mbx * 16 + 7];
217  break;
218  case 2:
219  *left = &h->left_border_y[8];
220  memcpy(&top[1], h->cy + 7 * h->l_stride, 16);
221  top[17] = top[16];
222  top[0] = top[1];
223  if (h->flags & A_AVAIL)
224  top[0] = h->left_border_y[8];
225  break;
226  case 3:
227  *left = &h->intern_border_y[8];
228  for (i = 0; i < 8; i++)
229  h->intern_border_y[i + 9] = *(h->cy + 7 + (i + 8) * h->l_stride);
230  memset(&h->intern_border_y[17], h->intern_border_y[16], 9);
231  memcpy(&top[0], h->cy + 7 + 7 * h->l_stride, 9);
232  memset(&top[9], top[8], 9);
233  break;
234  }
235 }
236 
238 {
239  /* extend borders by one pixel */
240  h->left_border_u[9] = h->left_border_u[8];
241  h->left_border_v[9] = h->left_border_v[8];
242  if(h->flags & C_AVAIL) {
243  h->top_border_u[h->mbx*10 + 9] = h->top_border_u[h->mbx*10 + 11];
244  h->top_border_v[h->mbx*10 + 9] = h->top_border_v[h->mbx*10 + 11];
245  } else {
246  h->top_border_u[h->mbx * 10 + 9] = h->top_border_u[h->mbx * 10 + 8];
247  h->top_border_v[h->mbx * 10 + 9] = h->top_border_v[h->mbx * 10 + 8];
248  }
249  if((h->flags & A_AVAIL) && (h->flags & B_AVAIL)) {
250  h->top_border_u[h->mbx * 10] = h->left_border_u[0] = h->topleft_border_u;
251  h->top_border_v[h->mbx * 10] = h->left_border_v[0] = h->topleft_border_v;
252  } else {
253  h->left_border_u[0] = h->left_border_u[1];
254  h->left_border_v[0] = h->left_border_v[1];
255  h->top_border_u[h->mbx * 10] = h->top_border_u[h->mbx * 10 + 1];
256  h->top_border_v[h->mbx * 10] = h->top_border_v[h->mbx * 10 + 1];
257  }
258 }
259 
260 static void intra_pred_vert(uint8_t *d, uint8_t *top, uint8_t *left, int stride)
261 {
262  int y;
263  uint64_t a = AV_RN64(&top[1]);
264  for (y = 0; y < 8; y++)
265  *((uint64_t *)(d + y * stride)) = a;
266 }
267 
268 static void intra_pred_horiz(uint8_t *d, uint8_t *top, uint8_t *left, int stride)
269 {
270  int y;
271  uint64_t a;
272  for (y = 0; y < 8; y++) {
273  a = left[y + 1] * 0x0101010101010101ULL;
274  *((uint64_t *)(d + y * stride)) = a;
275  }
276 }
277 
278 static void intra_pred_dc_128(uint8_t *d, uint8_t *top, uint8_t *left, int stride)
279 {
280  int y;
281  uint64_t a = 0x8080808080808080ULL;
282  for (y = 0; y < 8; y++)
283  *((uint64_t *)(d + y * stride)) = a;
284 }
285 
286 static void intra_pred_plane(uint8_t *d, uint8_t *top, uint8_t *left, int stride)
287 {
288  int x, y, ia;
289  int ih = 0;
290  int iv = 0;
291  const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
292 
293  for (x = 0; x < 4; x++) {
294  ih += (x + 1) * (top[5 + x] - top[3 - x]);
295  iv += (x + 1) * (left[5 + x] - left[3 - x]);
296  }
297  ia = (top[8] + left[8]) << 4;
298  ih = (17 * ih + 16) >> 5;
299  iv = (17 * iv + 16) >> 5;
300  for (y = 0; y < 8; y++)
301  for (x = 0; x < 8; x++)
302  d[y * stride + x] = cm[(ia + (x - 3) * ih + (y - 3) * iv + 16) >> 5];
303 }
304 
305 #define LOWPASS(ARRAY, INDEX) \
306  ((ARRAY[(INDEX) - 1] + 2 * ARRAY[(INDEX)] + ARRAY[(INDEX) + 1] + 2) >> 2)
307 
308 static void intra_pred_lp(uint8_t *d, uint8_t *top, uint8_t *left, int stride)
309 {
310  int x, y;
311  for (y = 0; y < 8; y++)
312  for (x = 0; x < 8; x++)
313  d[y * stride + x] = (LOWPASS(top, x + 1) + LOWPASS(left, y + 1)) >> 1;
314 }
315 
316 static void intra_pred_down_left(uint8_t *d, uint8_t *top, uint8_t *left, int stride)
317 {
318  int x, y;
319  for (y = 0; y < 8; y++)
320  for (x = 0; x < 8; x++)
321  d[y * stride + x] = (LOWPASS(top, x + y + 2) + LOWPASS(left, x + y + 2)) >> 1;
322 }
323 
324 static void intra_pred_down_right(uint8_t *d, uint8_t *top, uint8_t *left, int stride)
325 {
326  int x, y;
327  for (y = 0; y < 8; y++)
328  for (x = 0; x < 8; x++)
329  if (x == y)
330  d[y * stride + x] = (left[1] + 2 * top[0] + top[1] + 2) >> 2;
331  else if (x > y)
332  d[y * stride + x] = LOWPASS(top, x - y);
333  else
334  d[y * stride + x] = LOWPASS(left, y - x);
335 }
336 
337 static void intra_pred_lp_left(uint8_t *d, uint8_t *top, uint8_t *left, int stride)
338 {
339  int x, y;
340  for (y = 0; y < 8; y++)
341  for (x = 0; x < 8; x++)
342  d[y * stride + x] = LOWPASS(left, y + 1);
343 }
344 
345 static void intra_pred_lp_top(uint8_t *d, uint8_t *top, uint8_t *left, int stride)
346 {
347  int x, y;
348  for (y = 0; y < 8; y++)
349  for (x = 0; x < 8; x++)
350  d[y * stride + x] = LOWPASS(top, x + 1);
351 }
352 
353 #undef LOWPASS
354 
355 static inline void modify_pred(const int8_t *mod_table, int *mode)
356 {
357  *mode = mod_table[*mode];
358  if (*mode < 0) {
359  av_log(NULL, AV_LOG_ERROR, "Illegal intra prediction mode\n");
360  *mode = 0;
361  }
362 }
363 
364 void ff_cavs_modify_mb_i(AVSContext *h, int *pred_mode_uv)
365 {
366  /* save pred modes before they get modified */
367  h->pred_mode_Y[3] = h->pred_mode_Y[5];
368  h->pred_mode_Y[6] = h->pred_mode_Y[8];
369  h->top_pred_Y[h->mbx * 2 + 0] = h->pred_mode_Y[7];
370  h->top_pred_Y[h->mbx * 2 + 1] = h->pred_mode_Y[8];
371 
372  /* modify pred modes according to availability of neighbour samples */
373  if (!(h->flags & A_AVAIL)) {
376  modify_pred(left_modifier_c, pred_mode_uv);
377  }
378  if (!(h->flags & B_AVAIL)) {
381  modify_pred(top_modifier_c, pred_mode_uv);
382  }
383 }
384 
385 /*****************************************************************************
386  *
387  * motion compensation
388  *
389  ****************************************************************************/
390 
391 static inline void mc_dir_part(AVSContext *h, AVFrame *pic, int chroma_height,
392  int delta, int list, uint8_t *dest_y,
393  uint8_t *dest_cb, uint8_t *dest_cr,
394  int src_x_offset, int src_y_offset,
395  qpel_mc_func *qpix_op,
396  h264_chroma_mc_func chroma_op, cavs_vector *mv)
397 {
398  const int mx = mv->x + src_x_offset * 8;
399  const int my = mv->y + src_y_offset * 8;
400  const int luma_xy = (mx & 3) + ((my & 3) << 2);
401  uint8_t *src_y = pic->data[0] + (mx >> 2) + (my >> 2) * h->l_stride;
402  uint8_t *src_cb = pic->data[1] + (mx >> 3) + (my >> 3) * h->c_stride;
403  uint8_t *src_cr = pic->data[2] + (mx >> 3) + (my >> 3) * h->c_stride;
404  int extra_width = 0;
405  int extra_height = extra_width;
406  const int full_mx = mx >> 2;
407  const int full_my = my >> 2;
408  const int pic_width = 16 * h->mb_width;
409  const int pic_height = 16 * h->mb_height;
410  int emu = 0;
411 
412  if (!pic->data[0])
413  return;
414  if (mx & 7)
415  extra_width -= 3;
416  if (my & 7)
417  extra_height -= 3;
418 
419  if (full_mx < 0 - extra_width ||
420  full_my < 0 - extra_height ||
421  full_mx + 16 /* FIXME */ > pic_width + extra_width ||
422  full_my + 16 /* FIXME */ > pic_height + extra_height) {
424  src_y - 2 - 2 * h->l_stride,
425  h->l_stride, h->l_stride,
426  16 + 5, 16 + 5 /* FIXME */,
427  full_mx - 2, full_my - 2,
428  pic_width, pic_height);
429  src_y = h->edge_emu_buffer + 2 + 2 * h->l_stride;
430  emu = 1;
431  }
432 
433  // FIXME try variable height perhaps?
434  qpix_op[luma_xy](dest_y, src_y, h->l_stride);
435 
436  if (emu) {
437  h->vdsp.emulated_edge_mc(h->edge_emu_buffer, src_cb,
438  h->c_stride, h->c_stride,
439  9, 9 /* FIXME */,
440  mx >> 3, my >> 3,
441  pic_width >> 1, pic_height >> 1);
442  src_cb = h->edge_emu_buffer;
443  }
444  chroma_op(dest_cb, src_cb, h->c_stride, chroma_height, mx & 7, my & 7);
445 
446  if (emu) {
447  h->vdsp.emulated_edge_mc(h->edge_emu_buffer, src_cr,
448  h->c_stride, h->c_stride,
449  9, 9 /* FIXME */,
450  mx >> 3, my >> 3,
451  pic_width >> 1, pic_height >> 1);
452  src_cr = h->edge_emu_buffer;
453  }
454  chroma_op(dest_cr, src_cr, h->c_stride, chroma_height, mx & 7, my & 7);
455 }
456 
457 static inline void mc_part_std(AVSContext *h, int chroma_height, int delta,
458  uint8_t *dest_y,
459  uint8_t *dest_cb,
460  uint8_t *dest_cr,
461  int x_offset, int y_offset,
462  qpel_mc_func *qpix_put,
463  h264_chroma_mc_func chroma_put,
464  qpel_mc_func *qpix_avg,
465  h264_chroma_mc_func chroma_avg,
466  cavs_vector *mv)
467 {
468  qpel_mc_func *qpix_op = qpix_put;
469  h264_chroma_mc_func chroma_op = chroma_put;
470 
471  dest_y += x_offset * 2 + y_offset * h->l_stride * 2;
472  dest_cb += x_offset + y_offset * h->c_stride;
473  dest_cr += x_offset + y_offset * h->c_stride;
474  x_offset += 8 * h->mbx;
475  y_offset += 8 * h->mby;
476 
477  if (mv->ref >= 0) {
478  AVFrame *ref = h->DPB[mv->ref].f;
479  mc_dir_part(h, ref, chroma_height, delta, 0,
480  dest_y, dest_cb, dest_cr, x_offset, y_offset,
481  qpix_op, chroma_op, mv);
482 
483  qpix_op = qpix_avg;
484  chroma_op = chroma_avg;
485  }
486 
487  if ((mv + MV_BWD_OFFS)->ref >= 0) {
488  AVFrame *ref = h->DPB[0].f;
489  mc_dir_part(h, ref, chroma_height, delta, 1,
490  dest_y, dest_cb, dest_cr, x_offset, y_offset,
491  qpix_op, chroma_op, mv + MV_BWD_OFFS);
492  }
493 }
494 
495 void ff_cavs_inter(AVSContext *h, enum cavs_mb mb_type)
496 {
497  if (ff_cavs_partition_flags[mb_type] == 0) { // 16x16
498  mc_part_std(h, 8, 0, h->cy, h->cu, h->cv, 0, 0,
503  &h->mv[MV_FWD_X0]);
504  } else {
505  mc_part_std(h, 4, 0, h->cy, h->cu, h->cv, 0, 0,
510  &h->mv[MV_FWD_X0]);
511  mc_part_std(h, 4, 0, h->cy, h->cu, h->cv, 4, 0,
516  &h->mv[MV_FWD_X1]);
517  mc_part_std(h, 4, 0, h->cy, h->cu, h->cv, 0, 4,
522  &h->mv[MV_FWD_X2]);
523  mc_part_std(h, 4, 0, h->cy, h->cu, h->cv, 4, 4,
528  &h->mv[MV_FWD_X3]);
529  }
530 }
531 
532 /*****************************************************************************
533  *
534  * motion vector prediction
535  *
536  ****************************************************************************/
537 
538 static inline void scale_mv(AVSContext *h, int *d_x, int *d_y,
539  cavs_vector *src, int distp)
540 {
541  int den = h->scale_den[FFMAX(src->ref, 0)];
542 
543  *d_x = (src->x * distp * den + 256 + FF_SIGNBIT(src->x)) >> 9;
544  *d_y = (src->y * distp * den + 256 + FF_SIGNBIT(src->y)) >> 9;
545 }
546 
547 static inline void mv_pred_median(AVSContext *h,
548  cavs_vector *mvP,
549  cavs_vector *mvA,
550  cavs_vector *mvB,
551  cavs_vector *mvC)
552 {
553  int ax, ay, bx, by, cx, cy;
554  int len_ab, len_bc, len_ca, len_mid;
555 
556  /* scale candidates according to their temporal span */
557  scale_mv(h, &ax, &ay, mvA, mvP->dist);
558  scale_mv(h, &bx, &by, mvB, mvP->dist);
559  scale_mv(h, &cx, &cy, mvC, mvP->dist);
560  /* find the geometrical median of the three candidates */
561  len_ab = abs(ax - bx) + abs(ay - by);
562  len_bc = abs(bx - cx) + abs(by - cy);
563  len_ca = abs(cx - ax) + abs(cy - ay);
564  len_mid = mid_pred(len_ab, len_bc, len_ca);
565  if (len_mid == len_ab) {
566  mvP->x = cx;
567  mvP->y = cy;
568  } else if (len_mid == len_bc) {
569  mvP->x = ax;
570  mvP->y = ay;
571  } else {
572  mvP->x = bx;
573  mvP->y = by;
574  }
575 }
576 
577 void ff_cavs_mv(AVSContext *h, enum cavs_mv_loc nP, enum cavs_mv_loc nC,
578  enum cavs_mv_pred mode, enum cavs_block size, int ref)
579 {
580  cavs_vector *mvP = &h->mv[nP];
581  cavs_vector *mvA = &h->mv[nP-1];
582  cavs_vector *mvB = &h->mv[nP-4];
583  cavs_vector *mvC = &h->mv[nC];
584  const cavs_vector *mvP2 = NULL;
585 
586  mvP->ref = ref;
587  mvP->dist = h->dist[mvP->ref];
588  if (mvC->ref == NOT_AVAIL || (nP == MV_FWD_X3) || (nP == MV_BWD_X3 ))
589  mvC = &h->mv[nP - 5]; // set to top-left (mvD)
590  if (mode == MV_PRED_PSKIP &&
591  (mvA->ref == NOT_AVAIL ||
592  mvB->ref == NOT_AVAIL ||
593  (mvA->x | mvA->y | mvA->ref) == 0 ||
594  (mvB->x | mvB->y | mvB->ref) == 0)) {
595  mvP2 = &un_mv;
596  /* if there is only one suitable candidate, take it */
597  } else if (mvA->ref >= 0 && mvB->ref < 0 && mvC->ref < 0) {
598  mvP2 = mvA;
599  } else if (mvA->ref < 0 && mvB->ref >= 0 && mvC->ref < 0) {
600  mvP2 = mvB;
601  } else if (mvA->ref < 0 && mvB->ref < 0 && mvC->ref >= 0) {
602  mvP2 = mvC;
603  } else if (mode == MV_PRED_LEFT && mvA->ref == ref) {
604  mvP2 = mvA;
605  } else if (mode == MV_PRED_TOP && mvB->ref == ref) {
606  mvP2 = mvB;
607  } else if (mode == MV_PRED_TOPRIGHT && mvC->ref == ref) {
608  mvP2 = mvC;
609  }
610  if (mvP2) {
611  mvP->x = mvP2->x;
612  mvP->y = mvP2->y;
613  } else
614  mv_pred_median(h, mvP, mvA, mvB, mvC);
615 
616  if (mode < MV_PRED_PSKIP) {
617  mvP->x += get_se_golomb(&h->gb);
618  mvP->y += get_se_golomb(&h->gb);
619  }
620  set_mvs(mvP, size);
621 }
622 
623 /*****************************************************************************
624  *
625  * macroblock level
626  *
627  ****************************************************************************/
628 
629 /**
630  * initialise predictors for motion vectors and intra prediction
631  */
633 {
634  int i;
635 
636  /* copy predictors from top line (MB B and C) into cache */
637  for (i = 0; i < 3; i++) {
638  h->mv[MV_FWD_B2 + i] = h->top_mv[0][h->mbx * 2 + i];
639  h->mv[MV_BWD_B2 + i] = h->top_mv[1][h->mbx * 2 + i];
640  }
641  h->pred_mode_Y[1] = h->top_pred_Y[h->mbx * 2 + 0];
642  h->pred_mode_Y[2] = h->top_pred_Y[h->mbx * 2 + 1];
643  /* clear top predictors if MB B is not available */
644  if (!(h->flags & B_AVAIL)) {
645  h->mv[MV_FWD_B2] = un_mv;
646  h->mv[MV_FWD_B3] = un_mv;
647  h->mv[MV_BWD_B2] = un_mv;
648  h->mv[MV_BWD_B3] = un_mv;
649  h->pred_mode_Y[1] = h->pred_mode_Y[2] = NOT_AVAIL;
650  h->flags &= ~(C_AVAIL | D_AVAIL);
651  } else if (h->mbx) {
652  h->flags |= D_AVAIL;
653  }
654  if (h->mbx == h->mb_width - 1) // MB C not available
655  h->flags &= ~C_AVAIL;
656  /* clear top-right predictors if MB C is not available */
657  if (!(h->flags & C_AVAIL)) {
658  h->mv[MV_FWD_C2] = un_mv;
659  h->mv[MV_BWD_C2] = un_mv;
660  }
661  /* clear top-left predictors if MB D is not available */
662  if (!(h->flags & D_AVAIL)) {
663  h->mv[MV_FWD_D3] = un_mv;
664  h->mv[MV_BWD_D3] = un_mv;
665  }
666 }
667 
668 /**
669  * save predictors for later macroblocks and increase
670  * macroblock address
671  * @return 0 if end of frame is reached, 1 otherwise
672  */
674 {
675  int i;
676 
677  h->flags |= A_AVAIL;
678  h->cy += 16;
679  h->cu += 8;
680  h->cv += 8;
681  /* copy mvs as predictors to the left */
682  for (i = 0; i <= 20; i += 4)
683  h->mv[i] = h->mv[i + 2];
684  /* copy bottom mvs from cache to top line */
685  h->top_mv[0][h->mbx * 2 + 0] = h->mv[MV_FWD_X2];
686  h->top_mv[0][h->mbx * 2 + 1] = h->mv[MV_FWD_X3];
687  h->top_mv[1][h->mbx * 2 + 0] = h->mv[MV_BWD_X2];
688  h->top_mv[1][h->mbx * 2 + 1] = h->mv[MV_BWD_X3];
689  /* next MB address */
690  h->mbidx++;
691  h->mbx++;
692  if (h->mbx == h->mb_width) { // New mb line
693  h->flags = B_AVAIL | C_AVAIL;
694  /* clear left pred_modes */
695  h->pred_mode_Y[3] = h->pred_mode_Y[6] = NOT_AVAIL;
696  /* clear left mv predictors */
697  for (i = 0; i <= 20; i += 4)
698  h->mv[i] = un_mv;
699  h->mbx = 0;
700  h->mby++;
701  /* re-calculate sample pointers */
702  h->cy = h->cur.f->data[0] + h->mby * 16 * h->l_stride;
703  h->cu = h->cur.f->data[1] + h->mby * 8 * h->c_stride;
704  h->cv = h->cur.f->data[2] + h->mby * 8 * h->c_stride;
705  if (h->mby == h->mb_height) { // Frame end
706  return 0;
707  }
708  }
709  return 1;
710 }
711 
712 /*****************************************************************************
713  *
714  * frame level
715  *
716  ****************************************************************************/
717 
719 {
720  int i;
721 
722  /* clear some predictors */
723  for (i = 0; i <= 20; i += 4)
724  h->mv[i] = un_mv;
726  set_mvs(&h->mv[MV_BWD_X0], BLK_16X16);
728  set_mvs(&h->mv[MV_FWD_X0], BLK_16X16);
729  h->pred_mode_Y[3] = h->pred_mode_Y[6] = NOT_AVAIL;
730  h->cy = h->cur.f->data[0];
731  h->cu = h->cur.f->data[1];
732  h->cv = h->cur.f->data[2];
733  h->l_stride = h->cur.f->linesize[0];
734  h->c_stride = h->cur.f->linesize[1];
735  h->luma_scan[2] = 8 * h->l_stride;
736  h->luma_scan[3] = 8 * h->l_stride + 8;
737  h->mbx = h->mby = h->mbidx = 0;
738  h->flags = 0;
739 
740  return 0;
741 }
742 
743 /*****************************************************************************
744  *
745  * headers and interface
746  *
747  ****************************************************************************/
748 
749 /**
750  * some predictions require data from the top-neighbouring macroblock.
751  * this data has to be stored for one complete row of macroblocks
752  * and this storage space is allocated here
753  */
755 {
756  /* alloc top line of predictors */
757  h->top_qp = av_mallocz(h->mb_width);
758  h->top_mv[0] = av_mallocz_array(h->mb_width * 2 + 1, sizeof(cavs_vector));
759  h->top_mv[1] = av_mallocz_array(h->mb_width * 2 + 1, sizeof(cavs_vector));
760  h->top_pred_Y = av_mallocz_array(h->mb_width * 2, sizeof(*h->top_pred_Y));
761  h->top_border_y = av_mallocz_array(h->mb_width + 1, 16);
764 
765  /* alloc space for co-located MVs and types */
767  4 * sizeof(cavs_vector));
769  h->block = av_mallocz(64 * sizeof(int16_t));
770 
771  if (!h->top_qp || !h->top_mv[0] || !h->top_mv[1] || !h->top_pred_Y ||
772  !h->top_border_y || !h->top_border_u || !h->top_border_v ||
773  !h->col_mv || !h->col_type_base || !h->block) {
774  av_freep(&h->top_qp);
775  av_freep(&h->top_mv[0]);
776  av_freep(&h->top_mv[1]);
777  av_freep(&h->top_pred_Y);
778  av_freep(&h->top_border_y);
779  av_freep(&h->top_border_u);
780  av_freep(&h->top_border_v);
781  av_freep(&h->col_mv);
782  av_freep(&h->col_type_base);
783  av_freep(&h->block);
784  return AVERROR(ENOMEM);
785  }
786  return 0;
787 }
788 
790 {
791  AVSContext *h = avctx->priv_data;
792 
793  ff_blockdsp_init(&h->bdsp, avctx);
795  ff_idctdsp_init(&h->idsp, avctx);
796  ff_videodsp_init(&h->vdsp, 8);
797  ff_cavsdsp_init(&h->cdsp, avctx);
799  h->cdsp.idct_perm);
801 
802  h->avctx = avctx;
803  avctx->pix_fmt = AV_PIX_FMT_YUV420P;
804 
805  h->cur.f = av_frame_alloc();
806  h->DPB[0].f = av_frame_alloc();
807  h->DPB[1].f = av_frame_alloc();
808  if (!h->cur.f || !h->DPB[0].f || !h->DPB[1].f) {
809  ff_cavs_end(avctx);
810  return AVERROR(ENOMEM);
811  }
812 
813  h->luma_scan[0] = 0;
814  h->luma_scan[1] = 8;
830  h->mv[7] = un_mv;
831  h->mv[19] = un_mv;
832  return 0;
833 }
834 
836 {
837  AVSContext *h = avctx->priv_data;
838 
839  av_frame_free(&h->cur.f);
840  av_frame_free(&h->DPB[0].f);
841  av_frame_free(&h->DPB[1].f);
842 
843  av_freep(&h->top_qp);
844  av_freep(&h->top_mv[0]);
845  av_freep(&h->top_mv[1]);
846  av_freep(&h->top_pred_Y);
847  av_freep(&h->top_border_y);
848  av_freep(&h->top_border_u);
849  av_freep(&h->top_border_v);
850  av_freep(&h->col_mv);
851  av_freep(&h->col_type_base);
852  av_freep(&h->block);
854  return 0;
855 }