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 "mathops.h"
34 #include "qpeldsp.h"
35 #include "cavs.h"
36 
37 static const uint8_t alpha_tab[64] = {
38  0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 2, 2, 2, 3, 3,
39  4, 4, 5, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 16, 18, 20,
40  22, 24, 26, 28, 30, 33, 33, 35, 35, 36, 37, 37, 39, 39, 42, 44,
41  46, 48, 50, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64
42 };
43 
44 static const uint8_t beta_tab[64] = {
45  0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2,
46  2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 6, 6,
47  6, 7, 7, 7, 8, 8, 8, 9, 9, 10, 10, 11, 11, 12, 13, 14,
48  15, 16, 17, 18, 19, 20, 21, 22, 23, 23, 24, 24, 25, 25, 26, 27
49 };
50 
51 static const uint8_t tc_tab[64] = {
52  0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
53  1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2,
54  2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4,
55  5, 5, 5, 6, 6, 6, 7, 7, 7, 7, 8, 8, 8, 9, 9, 9
56 };
57 
58 /** mark block as unavailable, i.e. out of picture
59  * or not yet decoded */
60 static const cavs_vector un_mv = { 0, 0, 1, NOT_AVAIL };
61 
62 static const int8_t left_modifier_l[8] = { 0, -1, 6, -1, -1, 7, 6, 7 };
63 static const int8_t top_modifier_l[8] = { -1, 1, 5, -1, -1, 5, 7, 7 };
64 static const int8_t left_modifier_c[7] = { 5, -1, 2, -1, 6, 5, 6 };
65 static const int8_t top_modifier_c[7] = { 4, 1, -1, -1, 4, 6, 6 };
66 
67 /*****************************************************************************
68  *
69  * in-loop deblocking filter
70  *
71  ****************************************************************************/
72 
73 static inline int get_bs(cavs_vector *mvP, cavs_vector *mvQ, int b)
74 {
75  if ((mvP->ref == REF_INTRA) || (mvQ->ref == REF_INTRA))
76  return 2;
77  if((abs(mvP->x - mvQ->x) >= 4) ||
78  (abs(mvP->y - mvQ->y) >= 4) ||
79  (mvP->ref != mvQ->ref))
80  return 1;
81  if (b) {
82  mvP += MV_BWD_OFFS;
83  mvQ += MV_BWD_OFFS;
84  if((abs(mvP->x - mvQ->x) >= 4) ||
85  (abs(mvP->y - mvQ->y) >= 4) ||
86  (mvP->ref != mvQ->ref))
87  return 1;
88  }
89  return 0;
90 }
91 
92 #define SET_PARAMS \
93  alpha = alpha_tab[av_clip(qp_avg + h->alpha_offset, 0, 63)]; \
94  beta = beta_tab[av_clip(qp_avg + h->beta_offset, 0, 63)]; \
95  tc = tc_tab[av_clip(qp_avg + h->alpha_offset, 0, 63)];
96 
97 /**
98  * in-loop deblocking filter for a single macroblock
99  *
100  * boundary strength (bs) mapping:
101  *
102  * --4---5--
103  * 0 2 |
104  * | 6 | 7 |
105  * 1 3 |
106  * ---------
107  *
108  */
109 void ff_cavs_filter(AVSContext *h, enum cavs_mb mb_type)
110 {
111  uint8_t bs[8];
112  int qp_avg, alpha, beta, tc;
113  int i;
114 
115  /* save un-deblocked lines */
116  h->topleft_border_y = h->top_border_y[h->mbx * 16 + 15];
117  h->topleft_border_u = h->top_border_u[h->mbx * 10 + 8];
118  h->topleft_border_v = h->top_border_v[h->mbx * 10 + 8];
119  memcpy(&h->top_border_y[h->mbx * 16], h->cy + 15 * h->l_stride, 16);
120  memcpy(&h->top_border_u[h->mbx * 10 + 1], h->cu + 7 * h->c_stride, 8);
121  memcpy(&h->top_border_v[h->mbx * 10 + 1], h->cv + 7 * h->c_stride, 8);
122  for (i = 0; i < 8; i++) {
123  h->left_border_y[i * 2 + 1] = *(h->cy + 15 + (i * 2 + 0) * h->l_stride);
124  h->left_border_y[i * 2 + 2] = *(h->cy + 15 + (i * 2 + 1) * h->l_stride);
125  h->left_border_u[i + 1] = *(h->cu + 7 + i * h->c_stride);
126  h->left_border_v[i + 1] = *(h->cv + 7 + i * h->c_stride);
127  }
128  if (!h->loop_filter_disable) {
129  /* determine bs */
130  if (mb_type == I_8X8)
131  memset(bs, 2, 8);
132  else {
133  memset(bs, 0, 8);
134  if (ff_cavs_partition_flags[mb_type] & SPLITV) {
135  bs[2] = get_bs(&h->mv[MV_FWD_X0], &h->mv[MV_FWD_X1], mb_type > P_8X8);
136  bs[3] = get_bs(&h->mv[MV_FWD_X2], &h->mv[MV_FWD_X3], mb_type > P_8X8);
137  }
138  if (ff_cavs_partition_flags[mb_type] & SPLITH) {
139  bs[6] = get_bs(&h->mv[MV_FWD_X0], &h->mv[MV_FWD_X2], mb_type > P_8X8);
140  bs[7] = get_bs(&h->mv[MV_FWD_X1], &h->mv[MV_FWD_X3], mb_type > P_8X8);
141  }
142  bs[0] = get_bs(&h->mv[MV_FWD_A1], &h->mv[MV_FWD_X0], mb_type > P_8X8);
143  bs[1] = get_bs(&h->mv[MV_FWD_A3], &h->mv[MV_FWD_X2], mb_type > P_8X8);
144  bs[4] = get_bs(&h->mv[MV_FWD_B2], &h->mv[MV_FWD_X0], mb_type > P_8X8);
145  bs[5] = get_bs(&h->mv[MV_FWD_B3], &h->mv[MV_FWD_X1], mb_type > P_8X8);
146  }
147  if (AV_RN64(bs)) {
148  if (h->flags & A_AVAIL) {
149  qp_avg = (h->qp + h->left_qp + 1) >> 1;
150  SET_PARAMS;
151  h->cdsp.cavs_filter_lv(h->cy, h->l_stride, alpha, beta, tc, bs[0], bs[1]);
152  qp_avg = (ff_cavs_chroma_qp[h->qp] + ff_cavs_chroma_qp[h->left_qp] + 1) >> 1;
153  SET_PARAMS;
154  h->cdsp.cavs_filter_cv(h->cu, h->c_stride, alpha, beta, tc, bs[0], bs[1]);
155  h->cdsp.cavs_filter_cv(h->cv, h->c_stride, alpha, beta, tc, bs[0], bs[1]);
156  }
157  qp_avg = h->qp;
158  SET_PARAMS;
159  h->cdsp.cavs_filter_lv(h->cy + 8, h->l_stride, alpha, beta, tc, bs[2], bs[3]);
160  h->cdsp.cavs_filter_lh(h->cy + 8 * h->l_stride, h->l_stride, alpha, beta, tc, bs[6], bs[7]);
161 
162  if (h->flags & B_AVAIL) {
163  qp_avg = (h->qp + h->top_qp[h->mbx] + 1) >> 1;
164  SET_PARAMS;
165  h->cdsp.cavs_filter_lh(h->cy, h->l_stride, alpha, beta, tc, bs[4], bs[5]);
166  qp_avg = (ff_cavs_chroma_qp[h->qp] + ff_cavs_chroma_qp[h->top_qp[h->mbx]] + 1) >> 1;
167  SET_PARAMS;
168  h->cdsp.cavs_filter_ch(h->cu, h->c_stride, alpha, beta, tc, bs[4], bs[5]);
169  h->cdsp.cavs_filter_ch(h->cv, h->c_stride, alpha, beta, tc, bs[4], bs[5]);
170  }
171  }
172  }
173  h->left_qp = h->qp;
174  h->top_qp[h->mbx] = h->qp;
175 }
176 
177 #undef SET_PARAMS
178 
179 /*****************************************************************************
180  *
181  * spatial intra prediction
182  *
183  ****************************************************************************/
184 
186  uint8_t **left, int block)
187 {
188  int i;
189 
190  switch (block) {
191  case 0:
192  *left = h->left_border_y;
193  h->left_border_y[0] = h->left_border_y[1];
194  memset(&h->left_border_y[17], h->left_border_y[16], 9);
195  memcpy(&top[1], &h->top_border_y[h->mbx * 16], 16);
196  top[17] = top[16];
197  top[0] = top[1];
198  if ((h->flags & A_AVAIL) && (h->flags & B_AVAIL))
199  h->left_border_y[0] = top[0] = h->topleft_border_y;
200  break;
201  case 1:
202  *left = h->intern_border_y;
203  for (i = 0; i < 8; i++)
204  h->intern_border_y[i + 1] = *(h->cy + 7 + i * h->l_stride);
205  memset(&h->intern_border_y[9], h->intern_border_y[8], 9);
206  h->intern_border_y[0] = h->intern_border_y[1];
207  memcpy(&top[1], &h->top_border_y[h->mbx * 16 + 8], 8);
208  if (h->flags & C_AVAIL)
209  memcpy(&top[9], &h->top_border_y[(h->mbx + 1) * 16], 8);
210  else
211  memset(&top[9], top[8], 9);
212  top[17] = top[16];
213  top[0] = top[1];
214  if (h->flags & B_AVAIL)
215  h->intern_border_y[0] = top[0] = h->top_border_y[h->mbx * 16 + 7];
216  break;
217  case 2:
218  *left = &h->left_border_y[8];
219  memcpy(&top[1], h->cy + 7 * h->l_stride, 16);
220  top[17] = top[16];
221  top[0] = top[1];
222  if (h->flags & A_AVAIL)
223  top[0] = h->left_border_y[8];
224  break;
225  case 3:
226  *left = &h->intern_border_y[8];
227  for (i = 0; i < 8; i++)
228  h->intern_border_y[i + 9] = *(h->cy + 7 + (i + 8) * h->l_stride);
229  memset(&h->intern_border_y[17], h->intern_border_y[16], 9);
230  memcpy(&top[0], h->cy + 7 + 7 * h->l_stride, 9);
231  memset(&top[9], top[8], 9);
232  break;
233  }
234 }
235 
237 {
238  /* extend borders by one pixel */
239  h->left_border_u[9] = h->left_border_u[8];
240  h->left_border_v[9] = h->left_border_v[8];
241  if(h->flags & C_AVAIL) {
242  h->top_border_u[h->mbx*10 + 9] = h->top_border_u[h->mbx*10 + 11];
243  h->top_border_v[h->mbx*10 + 9] = h->top_border_v[h->mbx*10 + 11];
244  } else {
245  h->top_border_u[h->mbx * 10 + 9] = h->top_border_u[h->mbx * 10 + 8];
246  h->top_border_v[h->mbx * 10 + 9] = h->top_border_v[h->mbx * 10 + 8];
247  }
248  if((h->flags & A_AVAIL) && (h->flags & B_AVAIL)) {
249  h->top_border_u[h->mbx * 10] = h->left_border_u[0] = h->topleft_border_u;
250  h->top_border_v[h->mbx * 10] = h->left_border_v[0] = h->topleft_border_v;
251  } else {
252  h->left_border_u[0] = h->left_border_u[1];
253  h->left_border_v[0] = h->left_border_v[1];
254  h->top_border_u[h->mbx * 10] = h->top_border_u[h->mbx * 10 + 1];
255  h->top_border_v[h->mbx * 10] = h->top_border_v[h->mbx * 10 + 1];
256  }
257 }
258 
259 static void intra_pred_vert(uint8_t *d, uint8_t *top, uint8_t *left, int stride)
260 {
261  int y;
262  uint64_t a = AV_RN64(&top[1]);
263  for (y = 0; y < 8; y++)
264  *((uint64_t *)(d + y * stride)) = a;
265 }
266 
267 static void intra_pred_horiz(uint8_t *d, uint8_t *top, uint8_t *left, int stride)
268 {
269  int y;
270  uint64_t a;
271  for (y = 0; y < 8; y++) {
272  a = left[y + 1] * 0x0101010101010101ULL;
273  *((uint64_t *)(d + y * stride)) = a;
274  }
275 }
276 
277 static void intra_pred_dc_128(uint8_t *d, uint8_t *top, uint8_t *left, int stride)
278 {
279  int y;
280  uint64_t a = 0x8080808080808080ULL;
281  for (y = 0; y < 8; y++)
282  *((uint64_t *)(d + y * stride)) = a;
283 }
284 
285 static void intra_pred_plane(uint8_t *d, uint8_t *top, uint8_t *left, int stride)
286 {
287  int x, y, ia;
288  int ih = 0;
289  int iv = 0;
290  const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;
291 
292  for (x = 0; x < 4; x++) {
293  ih += (x + 1) * (top[5 + x] - top[3 - x]);
294  iv += (x + 1) * (left[5 + x] - left[3 - x]);
295  }
296  ia = (top[8] + left[8]) << 4;
297  ih = (17 * ih + 16) >> 5;
298  iv = (17 * iv + 16) >> 5;
299  for (y = 0; y < 8; y++)
300  for (x = 0; x < 8; x++)
301  d[y * stride + x] = cm[(ia + (x - 3) * ih + (y - 3) * iv + 16) >> 5];
302 }
303 
304 #define LOWPASS(ARRAY, INDEX) \
305  ((ARRAY[(INDEX) - 1] + 2 * ARRAY[(INDEX)] + ARRAY[(INDEX) + 1] + 2) >> 2)
306 
307 static void intra_pred_lp(uint8_t *d, uint8_t *top, uint8_t *left, int stride)
308 {
309  int x, y;
310  for (y = 0; y < 8; y++)
311  for (x = 0; x < 8; x++)
312  d[y * stride + x] = (LOWPASS(top, x + 1) + LOWPASS(left, y + 1)) >> 1;
313 }
314 
315 static void intra_pred_down_left(uint8_t *d, uint8_t *top, uint8_t *left, int stride)
316 {
317  int x, y;
318  for (y = 0; y < 8; y++)
319  for (x = 0; x < 8; x++)
320  d[y * stride + x] = (LOWPASS(top, x + y + 2) + LOWPASS(left, x + y + 2)) >> 1;
321 }
322 
323 static void intra_pred_down_right(uint8_t *d, uint8_t *top, uint8_t *left, int stride)
324 {
325  int x, y;
326  for (y = 0; y < 8; y++)
327  for (x = 0; x < 8; x++)
328  if (x == y)
329  d[y * stride + x] = (left[1] + 2 * top[0] + top[1] + 2) >> 2;
330  else if (x > y)
331  d[y * stride + x] = LOWPASS(top, x - y);
332  else
333  d[y * stride + x] = LOWPASS(left, y - x);
334 }
335 
336 static void intra_pred_lp_left(uint8_t *d, uint8_t *top, uint8_t *left, int stride)
337 {
338  int x, y;
339  for (y = 0; y < 8; y++)
340  for (x = 0; x < 8; x++)
341  d[y * stride + x] = LOWPASS(left, y + 1);
342 }
343 
344 static void intra_pred_lp_top(uint8_t *d, uint8_t *top, uint8_t *left, int stride)
345 {
346  int x, y;
347  for (y = 0; y < 8; y++)
348  for (x = 0; x < 8; x++)
349  d[y * stride + x] = LOWPASS(top, x + 1);
350 }
351 
352 #undef LOWPASS
353 
354 static inline void modify_pred(const int8_t *mod_table, int *mode)
355 {
356  *mode = mod_table[*mode];
357  if (*mode < 0) {
358  av_log(NULL, AV_LOG_ERROR, "Illegal intra prediction mode\n");
359  *mode = 0;
360  }
361 }
362 
363 void ff_cavs_modify_mb_i(AVSContext *h, int *pred_mode_uv)
364 {
365  /* save pred modes before they get modified */
366  h->pred_mode_Y[3] = h->pred_mode_Y[5];
367  h->pred_mode_Y[6] = h->pred_mode_Y[8];
368  h->top_pred_Y[h->mbx * 2 + 0] = h->pred_mode_Y[7];
369  h->top_pred_Y[h->mbx * 2 + 1] = h->pred_mode_Y[8];
370 
371  /* modify pred modes according to availability of neighbour samples */
372  if (!(h->flags & A_AVAIL)) {
375  modify_pred(left_modifier_c, pred_mode_uv);
376  }
377  if (!(h->flags & B_AVAIL)) {
380  modify_pred(top_modifier_c, pred_mode_uv);
381  }
382 }
383 
384 /*****************************************************************************
385  *
386  * motion compensation
387  *
388  ****************************************************************************/
389 
390 static inline void mc_dir_part(AVSContext *h, AVFrame *pic, int chroma_height,
391  int delta, int list, uint8_t *dest_y,
392  uint8_t *dest_cb, uint8_t *dest_cr,
393  int src_x_offset, int src_y_offset,
394  qpel_mc_func *qpix_op,
395  h264_chroma_mc_func chroma_op, cavs_vector *mv)
396 {
397  const int mx = mv->x + src_x_offset * 8;
398  const int my = mv->y + src_y_offset * 8;
399  const int luma_xy = (mx & 3) + ((my & 3) << 2);
400  uint8_t *src_y = pic->data[0] + (mx >> 2) + (my >> 2) * h->l_stride;
401  uint8_t *src_cb = pic->data[1] + (mx >> 3) + (my >> 3) * h->c_stride;
402  uint8_t *src_cr = pic->data[2] + (mx >> 3) + (my >> 3) * h->c_stride;
403  int extra_width = 0;
404  int extra_height = extra_width;
405  const int full_mx = mx >> 2;
406  const int full_my = my >> 2;
407  const int pic_width = 16 * h->mb_width;
408  const int pic_height = 16 * h->mb_height;
409  int emu = 0;
410 
411  if (!pic->data[0])
412  return;
413  if (mx & 7)
414  extra_width -= 3;
415  if (my & 7)
416  extra_height -= 3;
417 
418  if (full_mx < 0 - extra_width ||
419  full_my < 0 - extra_height ||
420  full_mx + 16 /* FIXME */ > pic_width + extra_width ||
421  full_my + 16 /* FIXME */ > pic_height + extra_height) {
423  src_y - 2 - 2 * h->l_stride,
424  h->l_stride, h->l_stride,
425  16 + 5, 16 + 5 /* FIXME */,
426  full_mx - 2, full_my - 2,
427  pic_width, pic_height);
428  src_y = h->edge_emu_buffer + 2 + 2 * h->l_stride;
429  emu = 1;
430  }
431 
432  // FIXME try variable height perhaps?
433  qpix_op[luma_xy](dest_y, src_y, h->l_stride);
434 
435  if (emu) {
436  h->vdsp.emulated_edge_mc(h->edge_emu_buffer, src_cb,
437  h->c_stride, h->c_stride,
438  9, 9 /* FIXME */,
439  mx >> 3, my >> 3,
440  pic_width >> 1, pic_height >> 1);
441  src_cb = h->edge_emu_buffer;
442  }
443  chroma_op(dest_cb, src_cb, h->c_stride, chroma_height, mx & 7, my & 7);
444 
445  if (emu) {
446  h->vdsp.emulated_edge_mc(h->edge_emu_buffer, src_cr,
447  h->c_stride, h->c_stride,
448  9, 9 /* FIXME */,
449  mx >> 3, my >> 3,
450  pic_width >> 1, pic_height >> 1);
451  src_cr = h->edge_emu_buffer;
452  }
453  chroma_op(dest_cr, src_cr, h->c_stride, chroma_height, mx & 7, my & 7);
454 }
455 
456 static inline void mc_part_std(AVSContext *h, int chroma_height, int delta,
457  uint8_t *dest_y,
458  uint8_t *dest_cb,
459  uint8_t *dest_cr,
460  int x_offset, int y_offset,
461  qpel_mc_func *qpix_put,
462  h264_chroma_mc_func chroma_put,
463  qpel_mc_func *qpix_avg,
464  h264_chroma_mc_func chroma_avg,
465  cavs_vector *mv)
466 {
467  qpel_mc_func *qpix_op = qpix_put;
468  h264_chroma_mc_func chroma_op = chroma_put;
469 
470  dest_y += x_offset * 2 + y_offset * h->l_stride * 2;
471  dest_cb += x_offset + y_offset * h->c_stride;
472  dest_cr += x_offset + y_offset * h->c_stride;
473  x_offset += 8 * h->mbx;
474  y_offset += 8 * h->mby;
475 
476  if (mv->ref >= 0) {
477  AVFrame *ref = h->DPB[mv->ref].f;
478  mc_dir_part(h, ref, chroma_height, delta, 0,
479  dest_y, dest_cb, dest_cr, x_offset, y_offset,
480  qpix_op, chroma_op, mv);
481 
482  qpix_op = qpix_avg;
483  chroma_op = chroma_avg;
484  }
485 
486  if ((mv + MV_BWD_OFFS)->ref >= 0) {
487  AVFrame *ref = h->DPB[0].f;
488  mc_dir_part(h, ref, chroma_height, delta, 1,
489  dest_y, dest_cb, dest_cr, x_offset, y_offset,
490  qpix_op, chroma_op, mv + MV_BWD_OFFS);
491  }
492 }
493 
494 void ff_cavs_inter(AVSContext *h, enum cavs_mb mb_type)
495 {
496  if (ff_cavs_partition_flags[mb_type] == 0) { // 16x16
497  mc_part_std(h, 8, 0, h->cy, h->cu, h->cv, 0, 0,
502  &h->mv[MV_FWD_X0]);
503  } else {
504  mc_part_std(h, 4, 0, h->cy, h->cu, h->cv, 0, 0,
509  &h->mv[MV_FWD_X0]);
510  mc_part_std(h, 4, 0, h->cy, h->cu, h->cv, 4, 0,
515  &h->mv[MV_FWD_X1]);
516  mc_part_std(h, 4, 0, h->cy, h->cu, h->cv, 0, 4,
521  &h->mv[MV_FWD_X2]);
522  mc_part_std(h, 4, 0, h->cy, h->cu, h->cv, 4, 4,
527  &h->mv[MV_FWD_X3]);
528  }
529 }
530 
531 /*****************************************************************************
532  *
533  * motion vector prediction
534  *
535  ****************************************************************************/
536 
537 static inline void scale_mv(AVSContext *h, int *d_x, int *d_y,
538  cavs_vector *src, int distp)
539 {
540  int den = h->scale_den[FFMAX(src->ref, 0)];
541 
542  *d_x = (src->x * distp * den + 256 + (src->x >> 31)) >> 9;
543  *d_y = (src->y * distp * den + 256 + (src->y >> 31)) >> 9;
544 }
545 
546 static inline void mv_pred_median(AVSContext *h,
547  cavs_vector *mvP,
548  cavs_vector *mvA,
549  cavs_vector *mvB,
550  cavs_vector *mvC)
551 {
552  int ax, ay, bx, by, cx, cy;
553  int len_ab, len_bc, len_ca, len_mid;
554 
555  /* scale candidates according to their temporal span */
556  scale_mv(h, &ax, &ay, mvA, mvP->dist);
557  scale_mv(h, &bx, &by, mvB, mvP->dist);
558  scale_mv(h, &cx, &cy, mvC, mvP->dist);
559  /* find the geometrical median of the three candidates */
560  len_ab = abs(ax - bx) + abs(ay - by);
561  len_bc = abs(bx - cx) + abs(by - cy);
562  len_ca = abs(cx - ax) + abs(cy - ay);
563  len_mid = mid_pred(len_ab, len_bc, len_ca);
564  if (len_mid == len_ab) {
565  mvP->x = cx;
566  mvP->y = cy;
567  } else if (len_mid == len_bc) {
568  mvP->x = ax;
569  mvP->y = ay;
570  } else {
571  mvP->x = bx;
572  mvP->y = by;
573  }
574 }
575 
576 void ff_cavs_mv(AVSContext *h, enum cavs_mv_loc nP, enum cavs_mv_loc nC,
577  enum cavs_mv_pred mode, enum cavs_block size, int ref)
578 {
579  cavs_vector *mvP = &h->mv[nP];
580  cavs_vector *mvA = &h->mv[nP-1];
581  cavs_vector *mvB = &h->mv[nP-4];
582  cavs_vector *mvC = &h->mv[nC];
583  const cavs_vector *mvP2 = NULL;
584 
585  mvP->ref = ref;
586  mvP->dist = h->dist[mvP->ref];
587  if (mvC->ref == NOT_AVAIL || (nP == MV_FWD_X3) || (nP == MV_BWD_X3 ))
588  mvC = &h->mv[nP - 5]; // set to top-left (mvD)
589  if (mode == MV_PRED_PSKIP &&
590  (mvA->ref == NOT_AVAIL ||
591  mvB->ref == NOT_AVAIL ||
592  (mvA->x | mvA->y | mvA->ref) == 0 ||
593  (mvB->x | mvB->y | mvB->ref) == 0)) {
594  mvP2 = &un_mv;
595  /* if there is only one suitable candidate, take it */
596  } else if (mvA->ref >= 0 && mvB->ref < 0 && mvC->ref < 0) {
597  mvP2 = mvA;
598  } else if (mvA->ref < 0 && mvB->ref >= 0 && mvC->ref < 0) {
599  mvP2 = mvB;
600  } else if (mvA->ref < 0 && mvB->ref < 0 && mvC->ref >= 0) {
601  mvP2 = mvC;
602  } else if (mode == MV_PRED_LEFT && mvA->ref == ref) {
603  mvP2 = mvA;
604  } else if (mode == MV_PRED_TOP && mvB->ref == ref) {
605  mvP2 = mvB;
606  } else if (mode == MV_PRED_TOPRIGHT && mvC->ref == ref) {
607  mvP2 = mvC;
608  }
609  if (mvP2) {
610  mvP->x = mvP2->x;
611  mvP->y = mvP2->y;
612  } else
613  mv_pred_median(h, mvP, mvA, mvB, mvC);
614 
615  if (mode < MV_PRED_PSKIP) {
616  mvP->x += get_se_golomb(&h->gb);
617  mvP->y += get_se_golomb(&h->gb);
618  }
619  set_mvs(mvP, size);
620 }
621 
622 /*****************************************************************************
623  *
624  * macroblock level
625  *
626  ****************************************************************************/
627 
628 /**
629  * initialise predictors for motion vectors and intra prediction
630  */
632 {
633  int i;
634 
635  /* copy predictors from top line (MB B and C) into cache */
636  for (i = 0; i < 3; i++) {
637  h->mv[MV_FWD_B2 + i] = h->top_mv[0][h->mbx * 2 + i];
638  h->mv[MV_BWD_B2 + i] = h->top_mv[1][h->mbx * 2 + i];
639  }
640  h->pred_mode_Y[1] = h->top_pred_Y[h->mbx * 2 + 0];
641  h->pred_mode_Y[2] = h->top_pred_Y[h->mbx * 2 + 1];
642  /* clear top predictors if MB B is not available */
643  if (!(h->flags & B_AVAIL)) {
644  h->mv[MV_FWD_B2] = un_mv;
645  h->mv[MV_FWD_B3] = un_mv;
646  h->mv[MV_BWD_B2] = un_mv;
647  h->mv[MV_BWD_B3] = un_mv;
648  h->pred_mode_Y[1] = h->pred_mode_Y[2] = NOT_AVAIL;
649  h->flags &= ~(C_AVAIL | D_AVAIL);
650  } else if (h->mbx) {
651  h->flags |= D_AVAIL;
652  }
653  if (h->mbx == h->mb_width - 1) // MB C not available
654  h->flags &= ~C_AVAIL;
655  /* clear top-right predictors if MB C is not available */
656  if (!(h->flags & C_AVAIL)) {
657  h->mv[MV_FWD_C2] = un_mv;
658  h->mv[MV_BWD_C2] = un_mv;
659  }
660  /* clear top-left predictors if MB D is not available */
661  if (!(h->flags & D_AVAIL)) {
662  h->mv[MV_FWD_D3] = un_mv;
663  h->mv[MV_BWD_D3] = un_mv;
664  }
665 }
666 
667 /**
668  * save predictors for later macroblocks and increase
669  * macroblock address
670  * @return 0 if end of frame is reached, 1 otherwise
671  */
673 {
674  int i;
675 
676  h->flags |= A_AVAIL;
677  h->cy += 16;
678  h->cu += 8;
679  h->cv += 8;
680  /* copy mvs as predictors to the left */
681  for (i = 0; i <= 20; i += 4)
682  h->mv[i] = h->mv[i + 2];
683  /* copy bottom mvs from cache to top line */
684  h->top_mv[0][h->mbx * 2 + 0] = h->mv[MV_FWD_X2];
685  h->top_mv[0][h->mbx * 2 + 1] = h->mv[MV_FWD_X3];
686  h->top_mv[1][h->mbx * 2 + 0] = h->mv[MV_BWD_X2];
687  h->top_mv[1][h->mbx * 2 + 1] = h->mv[MV_BWD_X3];
688  /* next MB address */
689  h->mbidx++;
690  h->mbx++;
691  if (h->mbx == h->mb_width) { // New mb line
692  h->flags = B_AVAIL | C_AVAIL;
693  /* clear left pred_modes */
694  h->pred_mode_Y[3] = h->pred_mode_Y[6] = NOT_AVAIL;
695  /* clear left mv predictors */
696  for (i = 0; i <= 20; i += 4)
697  h->mv[i] = un_mv;
698  h->mbx = 0;
699  h->mby++;
700  /* re-calculate sample pointers */
701  h->cy = h->cur.f->data[0] + h->mby * 16 * h->l_stride;
702  h->cu = h->cur.f->data[1] + h->mby * 8 * h->c_stride;
703  h->cv = h->cur.f->data[2] + h->mby * 8 * h->c_stride;
704  if (h->mby == h->mb_height) { // Frame end
705  return 0;
706  }
707  }
708  return 1;
709 }
710 
711 /*****************************************************************************
712  *
713  * frame level
714  *
715  ****************************************************************************/
716 
718 {
719  int i;
720 
721  /* clear some predictors */
722  for (i = 0; i <= 20; i += 4)
723  h->mv[i] = un_mv;
725  set_mvs(&h->mv[MV_BWD_X0], BLK_16X16);
727  set_mvs(&h->mv[MV_FWD_X0], BLK_16X16);
728  h->pred_mode_Y[3] = h->pred_mode_Y[6] = NOT_AVAIL;
729  h->cy = h->cur.f->data[0];
730  h->cu = h->cur.f->data[1];
731  h->cv = h->cur.f->data[2];
732  h->l_stride = h->cur.f->linesize[0];
733  h->c_stride = h->cur.f->linesize[1];
734  h->luma_scan[2] = 8 * h->l_stride;
735  h->luma_scan[3] = 8 * h->l_stride + 8;
736  h->mbx = h->mby = h->mbidx = 0;
737  h->flags = 0;
738 
739  return 0;
740 }
741 
742 /*****************************************************************************
743  *
744  * headers and interface
745  *
746  ****************************************************************************/
747 
748 /**
749  * some predictions require data from the top-neighbouring macroblock.
750  * this data has to be stored for one complete row of macroblocks
751  * and this storage space is allocated here
752  */
754 {
755  /* alloc top line of predictors */
756  h->top_qp = av_mallocz(h->mb_width);
757  h->top_mv[0] = av_mallocz_array(h->mb_width * 2 + 1, sizeof(cavs_vector));
758  h->top_mv[1] = av_mallocz_array(h->mb_width * 2 + 1, sizeof(cavs_vector));
759  h->top_pred_Y = av_mallocz_array(h->mb_width * 2, sizeof(*h->top_pred_Y));
760  h->top_border_y = av_mallocz_array(h->mb_width + 1, 16);
763 
764  /* alloc space for co-located MVs and types */
766  4 * sizeof(cavs_vector));
768  h->block = av_mallocz(64 * sizeof(int16_t));
769 }
770 
772 {
773  AVSContext *h = avctx->priv_data;
774 
775  ff_blockdsp_init(&h->bdsp, avctx);
777  ff_idctdsp_init(&h->idsp, avctx);
778  ff_videodsp_init(&h->vdsp, 8);
779  ff_cavsdsp_init(&h->cdsp, avctx);
781  h->cdsp.idct_perm);
783 
784  h->avctx = avctx;
785  avctx->pix_fmt = AV_PIX_FMT_YUV420P;
786 
787  h->cur.f = av_frame_alloc();
788  h->DPB[0].f = av_frame_alloc();
789  h->DPB[1].f = av_frame_alloc();
790  if (!h->cur.f || !h->DPB[0].f || !h->DPB[1].f) {
791  ff_cavs_end(avctx);
792  return AVERROR(ENOMEM);
793  }
794 
795  h->luma_scan[0] = 0;
796  h->luma_scan[1] = 8;
812  h->mv[7] = un_mv;
813  h->mv[19] = un_mv;
814  return 0;
815 }
816 
818 {
819  AVSContext *h = avctx->priv_data;
820 
821  av_frame_free(&h->cur.f);
822  av_frame_free(&h->DPB[0].f);
823  av_frame_free(&h->DPB[1].f);
824 
825  av_free(h->top_qp);
826  av_free(h->top_mv[0]);
827  av_free(h->top_mv[1]);
828  av_free(h->top_pred_Y);
829  av_free(h->top_border_y);
830  av_free(h->top_border_u);
831  av_free(h->top_border_v);
832  av_free(h->col_mv);
834  av_free(h->block);
836  return 0;
837 }