/src/ffmpeg/libavcodec/h264_refs.c
Line | Count | Source |
1 | | /* |
2 | | * H.26L/H.264/AVC/JVT/14496-10/... reference picture handling |
3 | | * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at> |
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 | | * H.264 / AVC / MPEG-4 part10 reference picture handling. |
25 | | * @author Michael Niedermayer <michaelni@gmx.at> |
26 | | */ |
27 | | |
28 | | #include <inttypes.h> |
29 | | |
30 | | #include "libavutil/avassert.h" |
31 | | #include "avcodec.h" |
32 | | #include "h264.h" |
33 | | #include "h264dec.h" |
34 | | #include "golomb.h" |
35 | | #include "mpegutils.h" |
36 | | |
37 | | #include <assert.h> |
38 | | |
39 | | static void pic_as_field(H264Ref *pic, const int parity) |
40 | 1.58M | { |
41 | 6.32M | for (int i = 0; i < FF_ARRAY_ELEMS(pic->data); ++i) { |
42 | 4.74M | if (parity == PICT_BOTTOM_FIELD) |
43 | 1.69M | pic->data[i] += pic->linesize[i]; |
44 | 4.74M | pic->reference = parity; |
45 | 4.74M | pic->linesize[i] *= 2; |
46 | 4.74M | } |
47 | 1.58M | pic->poc = pic->parent->field_poc[parity == PICT_BOTTOM_FIELD]; |
48 | 1.58M | } |
49 | | |
50 | | static void ref_from_h264pic(H264Ref *dst, const H264Picture *src) |
51 | 6.20M | { |
52 | 6.20M | memcpy(dst->data, src->f->data, sizeof(dst->data)); |
53 | 6.20M | memcpy(dst->linesize, src->f->linesize, sizeof(dst->linesize)); |
54 | 6.20M | dst->reference = src->reference; |
55 | 6.20M | dst->poc = src->poc; |
56 | 6.20M | dst->pic_id = src->pic_id; |
57 | 6.20M | dst->parent = src; |
58 | 6.20M | } |
59 | | |
60 | | static int split_field_copy(H264Ref *dest, const H264Picture *src, |
61 | | int parity, int id_add) |
62 | 6.15M | { |
63 | 6.15M | int match = !!(src->reference & parity); |
64 | | |
65 | 6.15M | if (match) { |
66 | 6.15M | ref_from_h264pic(dest, src); |
67 | 6.15M | if (parity != PICT_FRAME) { |
68 | 1.56M | pic_as_field(dest, parity); |
69 | 1.56M | dest->pic_id *= 2; |
70 | 1.56M | dest->pic_id += id_add; |
71 | 1.56M | } |
72 | 6.15M | } |
73 | | |
74 | 6.15M | return match; |
75 | 6.15M | } |
76 | | |
77 | | static int build_def_list(H264Ref *def, int def_len, |
78 | | H264Picture * const *in, int len, int is_long, int sel) |
79 | 6.26M | { |
80 | 6.26M | int i[2] = { 0 }; |
81 | 6.26M | int index = 0; |
82 | | |
83 | 15.3M | while (i[0] < len || i[1] < len) { |
84 | 58.3M | while (i[0] < len && !(in[i[0]] && (in[i[0]]->reference & sel))) |
85 | 49.2M | i[0]++; |
86 | 64.0M | while (i[1] < len && !(in[i[1]] && (in[i[1]]->reference & (sel ^ 3)))) |
87 | 54.9M | i[1]++; |
88 | 9.10M | if (i[0] < len) { |
89 | 5.93M | av_assert0(index < def_len); |
90 | 5.93M | in[i[0]]->pic_id = is_long ? i[0] : in[i[0]]->frame_num; |
91 | 5.93M | split_field_copy(&def[index++], in[i[0]++], sel, 1); |
92 | 5.93M | } |
93 | 9.10M | if (i[1] < len) { |
94 | 221k | av_assert0(index < def_len); |
95 | 221k | in[i[1]]->pic_id = is_long ? i[1] : in[i[1]]->frame_num; |
96 | 221k | split_field_copy(&def[index++], in[i[1]++], sel ^ 3, 0); |
97 | 221k | } |
98 | 9.10M | } |
99 | | |
100 | 6.26M | return index; |
101 | 6.26M | } |
102 | | |
103 | | static int add_sorted(H264Picture **sorted, H264Picture * const *src, |
104 | | int len, int limit, int dir) |
105 | 2.73M | { |
106 | 2.73M | int out_i = 0; |
107 | | |
108 | 5.47M | for (;;) { |
109 | 5.47M | int best_poc = dir ? INT_MIN : INT_MAX; |
110 | | |
111 | 23.5M | for (int i = 0; i < len; i++) { |
112 | 18.0M | const int poc = src[i]->poc; |
113 | 18.0M | if (((poc > limit) ^ dir) && ((poc < best_poc) ^ dir)) { |
114 | 4.69M | best_poc = poc; |
115 | 4.69M | sorted[out_i] = src[i]; |
116 | 4.69M | } |
117 | 18.0M | } |
118 | 5.47M | if (best_poc == (dir ? INT_MIN : INT_MAX)) |
119 | 2.73M | break; |
120 | 2.73M | limit = sorted[out_i++]->poc - dir; |
121 | 2.73M | } |
122 | 2.73M | return out_i; |
123 | 2.73M | } |
124 | | |
125 | | static int mismatches_ref(const H264Context *h, const H264Picture *pic) |
126 | 3.59M | { |
127 | 3.59M | const AVFrame *f = pic->f; |
128 | 3.59M | return (h->cur_pic_ptr->f->width != f->width || |
129 | 3.59M | h->cur_pic_ptr->f->height != f->height || |
130 | 3.59M | h->cur_pic_ptr->f->format != f->format); |
131 | 3.59M | } |
132 | | |
133 | | static void h264_initialise_ref_list(H264Context *h, H264SliceContext *sl) |
134 | 2.44M | { |
135 | 2.44M | int len; |
136 | | |
137 | 2.44M | if (sl->slice_type_nos == AV_PICTURE_TYPE_B) { |
138 | 684k | H264Picture *sorted[32]; |
139 | 684k | int cur_poc; |
140 | 684k | int lens[2]; |
141 | | |
142 | 684k | if (FIELD_PICTURE(h)) |
143 | 158k | cur_poc = h->cur_pic_ptr->field_poc[h->picture_structure == PICT_BOTTOM_FIELD]; |
144 | 525k | else |
145 | 525k | cur_poc = h->cur_pic_ptr->poc; |
146 | | |
147 | 2.05M | for (int list = 0; list < 2; list++) { |
148 | 1.36M | len = add_sorted(sorted, h->short_ref, h->short_ref_count, cur_poc, 1 ^ list); |
149 | 1.36M | len += add_sorted(sorted + len, h->short_ref, h->short_ref_count, cur_poc, 0 ^ list); |
150 | 1.36M | av_assert0(len <= 32); |
151 | | |
152 | 1.36M | len = build_def_list(sl->ref_list[list], FF_ARRAY_ELEMS(sl->ref_list[0]), |
153 | 1.36M | sorted, len, 0, h->picture_structure); |
154 | 1.36M | len += build_def_list(sl->ref_list[list] + len, |
155 | 1.36M | FF_ARRAY_ELEMS(sl->ref_list[0]) - len, |
156 | 1.36M | h->long_ref, 16, 1, h->picture_structure); |
157 | 1.36M | av_assert0(len <= 32); |
158 | | |
159 | 1.36M | memset(&sl->ref_list[list][len], 0, sizeof(H264Ref) * (32 - len)); |
160 | 1.36M | lens[list] = len; |
161 | 1.36M | } |
162 | | |
163 | 684k | if (lens[0] == lens[1] && lens[1] > 1) { |
164 | 386k | int i; |
165 | 1.78M | for (i = 0; i < lens[0] && |
166 | 1.41M | sl->ref_list[0][i].parent->f->buf[0]->buffer == |
167 | 1.41M | sl->ref_list[1][i].parent->f->buf[0]->buffer; i++); |
168 | 386k | if (i == lens[0]) { |
169 | 371k | FFSWAP(H264Ref, sl->ref_list[1][0], sl->ref_list[1][1]); |
170 | 371k | } |
171 | 386k | } |
172 | 1.76M | } else { |
173 | 1.76M | len = build_def_list(sl->ref_list[0], FF_ARRAY_ELEMS(sl->ref_list[0]), |
174 | 1.76M | h->short_ref, h->short_ref_count, 0, h->picture_structure); |
175 | 1.76M | len += build_def_list(sl->ref_list[0] + len, |
176 | 1.76M | FF_ARRAY_ELEMS(sl->ref_list[0]) - len, |
177 | 1.76M | h-> long_ref, 16, 1, h->picture_structure); |
178 | 1.76M | av_assert0(len <= 32); |
179 | | |
180 | 1.76M | memset(&sl->ref_list[0][len], 0, sizeof(H264Ref) * (32 - len)); |
181 | 1.76M | } |
182 | | #ifdef TRACE |
183 | | for (int i = 0; i < sl->ref_count[0]; i++) { |
184 | | ff_tlog(h->avctx, "List0: %s fn:%d 0x%p\n", |
185 | | (sl->ref_list[0][i].parent ? (sl->ref_list[0][i].parent->long_ref ? "LT" : "ST") : "??"), |
186 | | sl->ref_list[0][i].pic_id, |
187 | | sl->ref_list[0][i].data[0]); |
188 | | } |
189 | | if (sl->slice_type_nos == AV_PICTURE_TYPE_B) { |
190 | | for (int i = 0; i < sl->ref_count[1]; i++) { |
191 | | ff_tlog(h->avctx, "List1: %s fn:%d 0x%p\n", |
192 | | (sl->ref_list[1][i].parent ? (sl->ref_list[1][i].parent->long_ref ? "LT" : "ST") : "??"), |
193 | | sl->ref_list[1][i].pic_id, |
194 | | sl->ref_list[1][i].data[0]); |
195 | | } |
196 | | } |
197 | | #endif |
198 | | |
199 | 5.58M | for (int j = 0; j < 1 + (sl->slice_type_nos == AV_PICTURE_TYPE_B); j++) { |
200 | 11.4M | for (int i = 0; i < sl->ref_count[j]; i++) { |
201 | 8.31M | if (sl->ref_list[j][i].parent) { |
202 | 3.54M | if (mismatches_ref(h, sl->ref_list[j][i].parent)) { |
203 | 0 | av_log(h->avctx, AV_LOG_ERROR, "Discarding mismatching reference\n"); |
204 | 0 | memset(&sl->ref_list[j][i], 0, sizeof(sl->ref_list[j][i])); |
205 | 0 | } |
206 | 3.54M | } |
207 | 8.31M | } |
208 | 3.13M | } |
209 | 4.61M | for (int i = 0; i < sl->list_count; i++) |
210 | 2.16M | h->default_ref[i] = sl->ref_list[i][0]; |
211 | 2.44M | } |
212 | | |
213 | | /** |
214 | | * print short term list |
215 | | */ |
216 | | static void print_short_term(const H264Context *h) |
217 | 5.47M | { |
218 | 5.47M | if (h->avctx->debug & FF_DEBUG_MMCO) { |
219 | 0 | av_log(h->avctx, AV_LOG_DEBUG, "short term list:\n"); |
220 | 0 | for (uint32_t i = 0; i < h->short_ref_count; i++) { |
221 | 0 | H264Picture *pic = h->short_ref[i]; |
222 | 0 | av_log(h->avctx, AV_LOG_DEBUG, "%"PRIu32" fn:%d poc:%d %p\n", |
223 | 0 | i, pic->frame_num, pic->poc, pic->f->data[0]); |
224 | 0 | } |
225 | 0 | } |
226 | 5.47M | } |
227 | | |
228 | | /** |
229 | | * print long term list |
230 | | */ |
231 | | static void print_long_term(const H264Context *h) |
232 | 5.47M | { |
233 | 5.47M | if (h->avctx->debug & FF_DEBUG_MMCO) { |
234 | 0 | av_log(h->avctx, AV_LOG_DEBUG, "long term list:\n"); |
235 | 0 | for (uint32_t i = 0; i < 16; i++) { |
236 | 0 | H264Picture *pic = h->long_ref[i]; |
237 | 0 | if (pic) { |
238 | 0 | av_log(h->avctx, AV_LOG_DEBUG, "%"PRIu32" fn:%d poc:%d %p\n", |
239 | 0 | i, pic->frame_num, pic->poc, pic->f->data[0]); |
240 | 0 | } |
241 | 0 | } |
242 | 0 | } |
243 | 5.47M | } |
244 | | |
245 | | /** |
246 | | * Extract structure information about the picture described by pic_num in |
247 | | * the current decoding context (frame or field). Note that pic_num is |
248 | | * picture number without wrapping (so, 0<=pic_num<max_pic_num). |
249 | | * @param pic_num picture number for which to extract structure information |
250 | | * @param structure one of PICT_XXX describing structure of picture |
251 | | * with pic_num |
252 | | * @return frame number (short term) or long term index of picture |
253 | | * described by pic_num |
254 | | */ |
255 | | static int pic_num_extract(const H264Context *h, int pic_num, int *structure) |
256 | 1.04M | { |
257 | 1.04M | *structure = h->picture_structure; |
258 | 1.04M | if (FIELD_PICTURE(h)) { |
259 | 398k | if (!(pic_num & 1)) |
260 | | /* opposite field */ |
261 | 274k | *structure ^= PICT_FRAME; |
262 | 398k | pic_num >>= 1; |
263 | 398k | } |
264 | | |
265 | 1.04M | return pic_num; |
266 | 1.04M | } |
267 | | |
268 | | static void h264_fill_mbaff_ref_list(H264SliceContext *sl) |
269 | 740k | { |
270 | 1.40M | for (int list = 0; list < sl->list_count; list++) { |
271 | 3.25M | for (int i = 0; i < sl->ref_count[list]; i++) { |
272 | 2.59M | const H264Ref *frame = &sl->ref_list[list][i]; |
273 | 2.59M | H264Ref *field = &sl->ref_list[list][16 + 2 * i]; |
274 | | |
275 | 2.59M | field[0] = *frame; |
276 | | |
277 | 10.3M | for (int j = 0; j < 3; j++) |
278 | 7.77M | field[0].linesize[j] <<= 1; |
279 | 2.59M | field[0].reference = PICT_TOP_FIELD; |
280 | 2.59M | field[0].poc = field[0].parent->field_poc[0]; |
281 | | |
282 | 2.59M | field[1] = field[0]; |
283 | | |
284 | 10.3M | for (int j = 0; j < 3; j++) |
285 | 7.77M | field[1].data[j] += frame->parent->f->linesize[j]; |
286 | 2.59M | field[1].reference = PICT_BOTTOM_FIELD; |
287 | 2.59M | field[1].poc = field[1].parent->field_poc[1]; |
288 | 2.59M | } |
289 | 663k | } |
290 | 740k | } |
291 | | |
292 | | int ff_h264_build_ref_list(H264Context *h, H264SliceContext *sl) |
293 | 2.44M | { |
294 | 2.44M | print_short_term(h); |
295 | 2.44M | print_long_term(h); |
296 | | |
297 | 2.44M | h264_initialise_ref_list(h, sl); |
298 | | |
299 | 4.60M | for (int list = 0; list < sl->list_count; list++) { |
300 | 2.16M | int pred = sl->curr_pic_num; |
301 | | |
302 | 2.53M | for (int index = 0; index < sl->nb_ref_modifications[list]; index++) { |
303 | 380k | unsigned int modification_of_pic_nums_idc = sl->ref_modifications[list][index].op; |
304 | 380k | unsigned int val = sl->ref_modifications[list][index].val; |
305 | 380k | unsigned int pic_id; |
306 | 380k | int i, pic_structure; |
307 | 380k | H264Picture *ref = NULL; |
308 | | |
309 | 380k | switch (modification_of_pic_nums_idc) { |
310 | 287k | case 0: |
311 | 301k | case 1: { |
312 | 301k | const unsigned int abs_diff_pic_num = val + 1; |
313 | 301k | int frame_num; |
314 | | |
315 | 301k | if (abs_diff_pic_num > sl->max_pic_num) { |
316 | 2.55k | av_log(h->avctx, AV_LOG_ERROR, |
317 | 2.55k | "abs_diff_pic_num overflow\n"); |
318 | 2.55k | return AVERROR_INVALIDDATA; |
319 | 2.55k | } |
320 | | |
321 | 299k | if (modification_of_pic_nums_idc == 0) |
322 | 285k | pred -= abs_diff_pic_num; |
323 | 13.7k | else |
324 | 13.7k | pred += abs_diff_pic_num; |
325 | 299k | pred &= sl->max_pic_num - 1; |
326 | | |
327 | 299k | frame_num = pic_num_extract(h, pred, &pic_structure); |
328 | | |
329 | 662k | for (i = h->short_ref_count - 1; i >= 0; i--) { |
330 | 413k | ref = h->short_ref[i]; |
331 | 413k | assert(ref->reference); |
332 | 413k | assert(!ref->long_ref); |
333 | 413k | if (ref->frame_num == frame_num && |
334 | 51.7k | (ref->reference & pic_structure)) |
335 | 50.1k | break; |
336 | 413k | } |
337 | 299k | if (i >= 0) |
338 | 50.1k | pic_id = pred; |
339 | 299k | break; |
340 | 301k | } |
341 | 79.0k | case 2: { |
342 | 79.0k | int long_idx; |
343 | 79.0k | pic_id = val; // long_term_pic_idx |
344 | | |
345 | 79.0k | long_idx = pic_num_extract(h, pic_id, &pic_structure); |
346 | | |
347 | 79.0k | if (long_idx > 31U) { |
348 | 45 | av_log(h->avctx, AV_LOG_ERROR, |
349 | 45 | "long_term_pic_idx overflow\n"); |
350 | 45 | return AVERROR_INVALIDDATA; |
351 | 45 | } |
352 | 78.9k | ref = h->long_ref[long_idx]; |
353 | 78.9k | assert(!(ref && !ref->reference)); |
354 | 78.9k | if (ref && (ref->reference & pic_structure)) { |
355 | 719 | assert(ref->long_ref); |
356 | 719 | i = 0; |
357 | 78.2k | } else { |
358 | 78.2k | i = -1; |
359 | 78.2k | } |
360 | 78.9k | break; |
361 | 79.0k | } |
362 | 0 | default: |
363 | 0 | av_assert0(0); |
364 | 380k | } |
365 | | |
366 | 377k | if (i < 0 || mismatches_ref(h, ref)) { |
367 | 327k | av_log(h->avctx, AV_LOG_ERROR, |
368 | 327k | i < 0 ? "reference picture missing during reorder\n" : |
369 | 327k | "mismatching reference\n" |
370 | 327k | ); |
371 | 327k | if (h->avctx->err_recognition & AV_EF_EXPLODE) { |
372 | 9.92k | return AVERROR_INVALIDDATA; |
373 | 9.92k | } |
374 | 317k | memset(&sl->ref_list[list][index], 0, sizeof(sl->ref_list[0][0])); // FIXME |
375 | 317k | } else { |
376 | 104k | for (i = index; i + 1 < sl->ref_count[list]; i++) { |
377 | 74.8k | if (sl->ref_list[list][i].parent && |
378 | 65.9k | ref->long_ref == sl->ref_list[list][i].parent->long_ref && |
379 | 64.0k | pic_id == sl->ref_list[list][i].pic_id) |
380 | 21.3k | break; |
381 | 74.8k | } |
382 | 104k | for (; i > index; i--) { |
383 | 53.4k | sl->ref_list[list][i] = sl->ref_list[list][i - 1]; |
384 | 53.4k | } |
385 | 50.8k | ref_from_h264pic(&sl->ref_list[list][index], ref); |
386 | 50.8k | if (FIELD_PICTURE(h)) { |
387 | 11.5k | pic_as_field(&sl->ref_list[list][index], pic_structure); |
388 | 11.5k | } |
389 | 50.8k | } |
390 | 377k | } |
391 | 2.16M | } |
392 | 4.43M | for (int list = 0; list < sl->list_count; list++) { |
393 | 9.79M | for (int index = 0; index < sl->ref_count[list]; index++) { |
394 | 7.80M | if ( !sl->ref_list[list][index].parent |
395 | 4.63M | || (!FIELD_PICTURE(h) && (sl->ref_list[list][index].reference&3) != 3)) { |
396 | 4.63M | if (h->avctx->err_recognition & AV_EF_EXPLODE) { |
397 | 17.4k | av_log(h->avctx, AV_LOG_ERROR, "Missing reference picture\n"); |
398 | 17.4k | return AVERROR_INVALIDDATA; |
399 | 17.4k | } |
400 | 4.61M | av_log(h->avctx, AV_LOG_ERROR, "Missing reference picture, default is %d\n", h->default_ref[list].poc); |
401 | | |
402 | 78.4M | for (int i = 0; i < FF_ARRAY_ELEMS(h->last_pocs); i++) |
403 | 73.8M | h->last_pocs[i] = INT_MIN; |
404 | 4.61M | if (h->default_ref[list].parent |
405 | 4.54M | && !(!FIELD_PICTURE(h) && (h->default_ref[list].reference&3) != 3)) |
406 | 4.52M | sl->ref_list[list][index] = h->default_ref[list]; |
407 | 94.7k | else |
408 | 94.7k | return -1; |
409 | 4.61M | } |
410 | 7.68M | if (h->noref_gray>0 && sl->ref_list[list][index].parent->gray && h->non_gray) { |
411 | 1.04M | for (int j=0; j<sl->list_count; j++) { |
412 | 652k | int list2 = (list+j)&1; |
413 | 652k | if (h->default_ref[list2].parent && !h->default_ref[list2].parent->gray |
414 | 3.96k | && !(!FIELD_PICTURE(h) && (h->default_ref[list2].reference&3) != 3)) { |
415 | 3.83k | sl->ref_list[list][index] = h->default_ref[list2]; |
416 | 3.83k | av_log(h->avctx, AV_LOG_DEBUG, "replacement of gray gap frame\n"); |
417 | 3.83k | break; |
418 | 3.83k | } |
419 | 652k | } |
420 | 400k | } |
421 | 7.68M | av_assert0(av_buffer_get_ref_count(sl->ref_list[list][index].parent->f->buf[0]) > 0); |
422 | 7.68M | } |
423 | 2.10M | } |
424 | | |
425 | 2.32M | if (FRAME_MBAFF(h)) |
426 | 740k | h264_fill_mbaff_ref_list(sl); |
427 | | |
428 | 2.32M | return 0; |
429 | 2.43M | } |
430 | | |
431 | | int ff_h264_decode_ref_pic_list_reordering(H264SliceContext *sl, void *logctx) |
432 | 2.26M | { |
433 | 2.26M | sl->nb_ref_modifications[0] = 0; |
434 | 2.26M | sl->nb_ref_modifications[1] = 0; |
435 | | |
436 | 5.07M | for (int list = 0; list < sl->list_count; list++) { |
437 | 3.10M | if (!get_bits1(&sl->gb)) // ref_pic_list_modification_flag_l[01] |
438 | 2.39M | continue; |
439 | | |
440 | 1.43M | for (int index = 0; ; index++) { |
441 | 1.43M | unsigned int op = get_ue_golomb_31(&sl->gb); |
442 | | |
443 | 1.43M | if (op == 3) |
444 | 417k | break; |
445 | | |
446 | 1.01M | if (index >= sl->ref_count[list]) { |
447 | 153k | av_log(logctx, AV_LOG_ERROR, "reference count overflow\n"); |
448 | 153k | return AVERROR_INVALIDDATA; |
449 | 863k | } else if (op > 2) { |
450 | 131k | av_log(logctx, AV_LOG_ERROR, |
451 | 131k | "illegal modification_of_pic_nums_idc %u\n", |
452 | 131k | op); |
453 | 131k | return AVERROR_INVALIDDATA; |
454 | 131k | } |
455 | 732k | sl->ref_modifications[list][index].val = get_ue_golomb_long(&sl->gb); |
456 | 732k | sl->ref_modifications[list][index].op = op; |
457 | 732k | sl->nb_ref_modifications[list]++; |
458 | 732k | } |
459 | 702k | } |
460 | | |
461 | 1.97M | return 0; |
462 | 2.26M | } |
463 | | |
464 | | /** |
465 | | * Mark a picture as no longer needed for reference. The refmask |
466 | | * argument allows unreferencing of individual fields or the whole frame. |
467 | | * If the picture becomes entirely unreferenced, but is being held for |
468 | | * display purposes, it is marked as such. |
469 | | * @param refmask mask of fields to unreference; the mask is bitwise |
470 | | * anded with the reference marking of pic |
471 | | * @return non-zero if pic becomes entirely unreferenced (except possibly |
472 | | * for display purposes) zero if one of the fields remains in |
473 | | * reference |
474 | | */ |
475 | | static inline int unreference_pic(H264Context *h, H264Picture *pic, int refmask) |
476 | 3.57M | { |
477 | 3.57M | if (pic->reference &= refmask) { |
478 | 101k | return 0; |
479 | 3.47M | } else { |
480 | 7.97M | for (int i = 0; h->delayed_pic[i]; i++) |
481 | 5.34M | if(pic == h->delayed_pic[i]){ |
482 | 851k | pic->reference = DELAYED_PIC_REF; |
483 | 851k | break; |
484 | 851k | } |
485 | 3.47M | return 1; |
486 | 3.47M | } |
487 | 3.57M | } |
488 | | |
489 | | /** |
490 | | * Find a H264Picture in the short term reference list by frame number. |
491 | | * @param frame_num frame number to search for |
492 | | * @param idx the index into h->short_ref where returned picture is found |
493 | | * undefined if no picture found. |
494 | | * @return pointer to the found picture, or NULL if no pic with the provided |
495 | | * frame number is found |
496 | | */ |
497 | | static H264Picture *find_short(H264Context *h, int frame_num, int *idx) |
498 | 3.90M | { |
499 | 11.0M | for (int i = 0; i < h->short_ref_count; i++) { |
500 | 8.84M | H264Picture *pic = h->short_ref[i]; |
501 | 8.84M | if (h->avctx->debug & FF_DEBUG_MMCO) |
502 | 0 | av_log(h->avctx, AV_LOG_DEBUG, "%d %d %p\n", i, pic->frame_num, pic); |
503 | 8.84M | if (pic->frame_num == frame_num) { |
504 | 1.67M | *idx = i; |
505 | 1.67M | return pic; |
506 | 1.67M | } |
507 | 8.84M | } |
508 | 2.22M | return NULL; |
509 | 3.90M | } |
510 | | |
511 | | /** |
512 | | * Remove a picture from the short term reference list by its index in |
513 | | * that list. This does no checking on the provided index; it is assumed |
514 | | * to be valid. Other list entries are shifted down. |
515 | | * @param i index into h->short_ref of picture to remove. |
516 | | */ |
517 | | static void remove_short_at_index(H264Context *h, int i) |
518 | 1.16M | { |
519 | 1.16M | assert(i >= 0 && i < h->short_ref_count); |
520 | 1.16M | h->short_ref[i] = NULL; |
521 | 1.16M | if (--h->short_ref_count) |
522 | 647k | memmove(&h->short_ref[i], &h->short_ref[i + 1], |
523 | 647k | (h->short_ref_count - i) * sizeof(H264Picture*)); |
524 | 1.16M | } |
525 | | |
526 | | /** |
527 | | * @return the removed picture or NULL if an error occurs |
528 | | */ |
529 | | static H264Picture *remove_short(H264Context *h, int frame_num, int ref_mask) |
530 | 3.34M | { |
531 | 3.34M | H264Picture *pic; |
532 | 3.34M | int i; |
533 | | |
534 | 3.34M | if (h->avctx->debug & FF_DEBUG_MMCO) |
535 | 0 | av_log(h->avctx, AV_LOG_DEBUG, "remove short %d count %d\n", frame_num, h->short_ref_count); |
536 | | |
537 | 3.34M | pic = find_short(h, frame_num, &i); |
538 | 3.34M | if (pic) { |
539 | 1.25M | if (unreference_pic(h, pic, ref_mask)) |
540 | 1.15M | remove_short_at_index(h, i); |
541 | 1.25M | } |
542 | | |
543 | 3.34M | return pic; |
544 | 3.34M | } |
545 | | |
546 | | /** |
547 | | * Remove a picture from the long term reference list by its index in |
548 | | * that list. |
549 | | * @return the removed picture or NULL if an error occurs |
550 | | */ |
551 | | static H264Picture *remove_long(H264Context *h, int i, int ref_mask) |
552 | 28.9M | { |
553 | 28.9M | H264Picture *pic; |
554 | | |
555 | 28.9M | pic = h->long_ref[i]; |
556 | 28.9M | if (pic) { |
557 | 939k | if (unreference_pic(h, pic, ref_mask)) { |
558 | 938k | assert(h->long_ref[i]->long_ref == 1); |
559 | 938k | h->long_ref[i]->long_ref = 0; |
560 | 938k | h->long_ref[i] = NULL; |
561 | 938k | h->long_ref_count--; |
562 | 938k | } |
563 | 939k | } |
564 | | |
565 | 28.9M | return pic; |
566 | 28.9M | } |
567 | | |
568 | | void ff_h264_remove_all_refs(H264Context *h) |
569 | 1.68M | { |
570 | 28.5M | for (int i = 0; i < 16; i++) |
571 | 26.9M | remove_long(h, i, 0); |
572 | 1.68M | assert(h->long_ref_count == 0); |
573 | | |
574 | 1.68M | if (h->short_ref_count && !h->last_pic_for_ec.f->data[0]) { |
575 | 682k | ff_h264_unref_picture(&h->last_pic_for_ec); |
576 | 682k | ff_h264_ref_picture(&h->last_pic_for_ec, h->short_ref[0]); |
577 | 682k | } |
578 | | |
579 | 3.06M | for (int i = 0; i < h->short_ref_count; i++) { |
580 | 1.38M | unreference_pic(h, h->short_ref[i], 0); |
581 | 1.38M | h->short_ref[i] = NULL; |
582 | 1.38M | } |
583 | 1.68M | h->short_ref_count = 0; |
584 | | |
585 | 1.68M | memset(h->default_ref, 0, sizeof(h->default_ref)); |
586 | 1.68M | } |
587 | | |
588 | | static void generate_sliding_window_mmcos(H264Context *h) |
589 | 1.84M | { |
590 | 1.84M | MMCO *mmco = h->mmco; |
591 | 1.84M | int nb_mmco = 0; |
592 | | |
593 | 1.84M | if (h->short_ref_count && |
594 | 1.44M | h->long_ref_count + h->short_ref_count >= h->ps.sps->ref_frame_count && |
595 | 308k | !(FIELD_PICTURE(h) && !h->first_field && h->cur_pic_ptr->reference)) { |
596 | 307k | mmco[0].opcode = MMCO_SHORT2UNUSED; |
597 | 307k | mmco[0].short_pic_num = h->short_ref[h->short_ref_count - 1]->frame_num; |
598 | 307k | nb_mmco = 1; |
599 | 307k | if (FIELD_PICTURE(h)) { |
600 | 58.4k | mmco[0].short_pic_num *= 2; |
601 | 58.4k | mmco[1].opcode = MMCO_SHORT2UNUSED; |
602 | 58.4k | mmco[1].short_pic_num = mmco[0].short_pic_num + 1; |
603 | 58.4k | nb_mmco = 2; |
604 | 58.4k | } |
605 | 307k | } |
606 | | |
607 | 1.84M | h->nb_mmco = nb_mmco; |
608 | 1.84M | } |
609 | | |
610 | | int ff_h264_execute_ref_pic_marking(H264Context *h) |
611 | 3.02M | { |
612 | 3.02M | MMCO *mmco = h->mmco; |
613 | 3.02M | int mmco_count; |
614 | 3.02M | int pps_ref_count[2] = {0}; |
615 | 3.02M | int current_ref_assigned = 0, err = 0; |
616 | | |
617 | 3.02M | if (!h->ps.sps) { |
618 | 0 | av_log(h->avctx, AV_LOG_ERROR, "SPS is unset\n"); |
619 | 0 | err = AVERROR_INVALIDDATA; |
620 | 0 | goto out; |
621 | 0 | } |
622 | | |
623 | 3.02M | if (!h->explicit_ref_marking) |
624 | 1.84M | generate_sliding_window_mmcos(h); |
625 | 3.02M | mmco_count = h->nb_mmco; |
626 | | |
627 | 3.02M | if ((h->avctx->debug & FF_DEBUG_MMCO) && mmco_count == 0) |
628 | 0 | av_log(h->avctx, AV_LOG_DEBUG, "no mmco here\n"); |
629 | | |
630 | 4.65M | for (int i = 0; i < mmco_count; i++) { |
631 | 1.63M | if (h->avctx->debug & FF_DEBUG_MMCO) |
632 | 0 | av_log(h->avctx, AV_LOG_DEBUG, "mmco:%d %d %d\n", h->mmco[i].opcode, |
633 | 0 | h->mmco[i].short_pic_num, h->mmco[i].long_arg); |
634 | | |
635 | 1.63M | switch (mmco[i].opcode) { |
636 | 519k | case MMCO_SHORT2UNUSED: |
637 | 563k | case MMCO_SHORT2LONG: { |
638 | 563k | int structure, j; |
639 | 563k | int frame_num = pic_num_extract(h, mmco[i].short_pic_num, &structure); |
640 | 563k | H264Picture *pic = find_short(h, frame_num, &j); |
641 | | |
642 | 563k | if (!pic) { |
643 | 138k | if (mmco[i].opcode != MMCO_SHORT2LONG || |
644 | 33.5k | !h->long_ref[mmco[i].long_arg] || |
645 | 134k | h->long_ref[mmco[i].long_arg]->frame_num != frame_num) { |
646 | 134k | av_log(h->avctx, h->short_ref_count ? AV_LOG_ERROR : AV_LOG_DEBUG, "mmco: unref short failure\n"); |
647 | 134k | err = AVERROR_INVALIDDATA; |
648 | 134k | } |
649 | 138k | continue; |
650 | 138k | } |
651 | 425k | if (mmco[i].opcode == MMCO_SHORT2UNUSED) { |
652 | 415k | if (h->avctx->debug & FF_DEBUG_MMCO) |
653 | 0 | av_log(h->avctx, AV_LOG_DEBUG, "mmco: unref short %d count %d\n", |
654 | 0 | h->mmco[i].short_pic_num, h->short_ref_count); |
655 | 415k | remove_short(h, frame_num, structure ^ PICT_FRAME); |
656 | 415k | } else { |
657 | 9.85k | if (h->long_ref[mmco[i].long_arg] != pic) |
658 | 9.85k | remove_long(h, mmco[i].long_arg, 0); |
659 | | |
660 | 9.85k | remove_short_at_index(h, j); |
661 | 9.85k | h->long_ref[ mmco[i].long_arg ] = pic; |
662 | 9.85k | if (h->long_ref[mmco[i].long_arg]) { |
663 | 9.85k | h->long_ref[mmco[i].long_arg]->long_ref = 1; |
664 | 9.85k | h->long_ref_count++; |
665 | 9.85k | } |
666 | 9.85k | } |
667 | 425k | break; |
668 | 563k | } |
669 | 100k | case MMCO_LONG2UNUSED: { |
670 | 100k | int structure, j = pic_num_extract(h, mmco[i].long_arg, &structure); |
671 | 100k | H264Picture *pic = h->long_ref[j]; |
672 | 100k | if (pic) { |
673 | 3.60k | remove_long(h, j, structure ^ PICT_FRAME); |
674 | 97.1k | } else if (h->avctx->debug & FF_DEBUG_MMCO) |
675 | 0 | av_log(h->avctx, AV_LOG_DEBUG, "mmco: unref long failure\n"); |
676 | 100k | break; |
677 | 563k | } |
678 | 928k | case MMCO_LONG: |
679 | | // Comment below left from previous code as it is an interesting note. |
680 | | /* First field in pair is in short term list or |
681 | | * at a different long term index. |
682 | | * This is not allowed; see 7.4.3.3, notes 2 and 3. |
683 | | * Report the problem and keep the pair where it is, |
684 | | * and mark this field valid. |
685 | | */ |
686 | 928k | if (h->short_ref[0] == h->cur_pic_ptr) { |
687 | 211 | av_log(h->avctx, AV_LOG_ERROR, "mmco: cannot assign current picture to short and long at the same time\n"); |
688 | 211 | remove_short_at_index(h, 0); |
689 | 211 | } |
690 | | |
691 | | /* make sure the current picture is not already assigned as a long ref */ |
692 | 928k | if (h->cur_pic_ptr->long_ref) { |
693 | 15.6M | for (int j = 0; j < FF_ARRAY_ELEMS(h->long_ref); j++) { |
694 | 15.2M | if (h->long_ref[j] == h->cur_pic_ptr) { |
695 | 475k | if (j != mmco[i].long_arg) |
696 | 20.1k | av_log(h->avctx, AV_LOG_ERROR, "mmco: cannot assign current picture to 2 long term references\n"); |
697 | 475k | remove_long(h, j, 0); |
698 | 475k | } |
699 | 15.2M | } |
700 | 475k | } |
701 | | |
702 | 928k | if (h->long_ref[mmco[i].long_arg] != h->cur_pic_ptr) { |
703 | 928k | av_assert0(!h->cur_pic_ptr->long_ref); |
704 | 928k | remove_long(h, mmco[i].long_arg, 0); |
705 | | |
706 | 928k | h->long_ref[mmco[i].long_arg] = h->cur_pic_ptr; |
707 | 928k | h->long_ref[mmco[i].long_arg]->long_ref = 1; |
708 | 928k | h->long_ref_count++; |
709 | 928k | } |
710 | | |
711 | 928k | h->cur_pic_ptr->reference |= h->picture_structure; |
712 | 928k | current_ref_assigned = 1; |
713 | 928k | break; |
714 | 15.6k | case MMCO_SET_MAX_LONG: |
715 | 15.6k | assert(mmco[i].long_arg <= 16); |
716 | | // just remove the long term which index is greater than new max |
717 | 221k | for (int j = mmco[i].long_arg; j < 16; j++) |
718 | 205k | remove_long(h, j, 0); |
719 | 15.6k | break; |
720 | 27.2k | case MMCO_RESET: |
721 | 59.0k | while (h->short_ref_count) { |
722 | 31.8k | remove_short(h, h->short_ref[0]->frame_num, 0); |
723 | 31.8k | } |
724 | 462k | for (int j = 0; j < 16; j++) |
725 | 435k | remove_long(h, j, 0); |
726 | 27.2k | h->poc.frame_num = h->cur_pic_ptr->frame_num = 0; |
727 | 27.2k | h->mmco_reset = 1; |
728 | 27.2k | h->cur_pic_ptr->mmco_reset = 1; |
729 | 462k | for (int j = 0; j < FF_ARRAY_ELEMS(h->last_pocs); j++) |
730 | 435k | h->last_pocs[j] = INT_MIN; |
731 | 27.2k | break; |
732 | 0 | default: av_assert0(0); |
733 | 1.63M | } |
734 | 1.63M | } |
735 | | |
736 | 3.02M | if (!current_ref_assigned) { |
737 | | /* Second field of complementary field pair; the first field of |
738 | | * which is already referenced. If short referenced, it |
739 | | * should be first entry in short_ref. If not, it must exist |
740 | | * in long_ref; trying to put it on the short list here is an |
741 | | * error in the encoded bit stream (ref: 7.4.3.3, NOTE 2 and 3). |
742 | | */ |
743 | 2.56M | if (h->short_ref_count && h->short_ref[0] == h->cur_pic_ptr) { |
744 | | /* Just mark the second field valid */ |
745 | 18.0k | h->cur_pic_ptr->reference |= h->picture_structure; |
746 | 2.54M | } else if (h->cur_pic_ptr->long_ref) { |
747 | 315 | av_log(h->avctx, AV_LOG_ERROR, "illegal short term reference " |
748 | 315 | "assignment for second field " |
749 | 315 | "in complementary field pair " |
750 | 315 | "(first field is long term)\n"); |
751 | 315 | err = AVERROR_INVALIDDATA; |
752 | 2.54M | } else { |
753 | 2.54M | H264Picture *pic = remove_short(h, h->cur_pic_ptr->frame_num, 0); |
754 | 2.54M | if (pic) { |
755 | 455k | av_log(h->avctx, AV_LOG_ERROR, "illegal short term buffer state detected\n"); |
756 | 455k | err = AVERROR_INVALIDDATA; |
757 | 455k | } |
758 | | |
759 | 2.54M | if (h->short_ref_count) |
760 | 1.34M | memmove(&h->short_ref[1], &h->short_ref[0], |
761 | 1.34M | h->short_ref_count * sizeof(H264Picture*)); |
762 | | |
763 | 2.54M | h->short_ref[0] = h->cur_pic_ptr; |
764 | 2.54M | h->short_ref_count++; |
765 | 2.54M | h->cur_pic_ptr->reference |= h->picture_structure; |
766 | 2.54M | } |
767 | 2.56M | } |
768 | | |
769 | 3.02M | if (h->long_ref_count + h->short_ref_count > FFMAX(h->ps.sps->ref_frame_count, 1)) { |
770 | | |
771 | | /* We have too many reference frames, probably due to corrupted |
772 | | * stream. Need to discard one frame. Prevents overrun of the |
773 | | * short_ref and long_ref buffers. |
774 | | */ |
775 | 151k | av_log(h->avctx, AV_LOG_ERROR, |
776 | 151k | "number of reference frames (%d+%d) exceeds max (%d; probably " |
777 | 151k | "corrupt input), discarding one\n", |
778 | 151k | h->long_ref_count, h->short_ref_count, h->ps.sps->ref_frame_count); |
779 | 151k | err = AVERROR_INVALIDDATA; |
780 | | |
781 | 151k | if (h->long_ref_count && !h->short_ref_count) { |
782 | 2.49k | int i; |
783 | 7.37k | for (i = 0; i < 16; ++i) |
784 | 7.37k | if (h->long_ref[i]) |
785 | 2.49k | break; |
786 | | |
787 | 2.49k | assert(i < 16); |
788 | 2.49k | remove_long(h, i, 0); |
789 | 148k | } else { |
790 | 148k | H264Picture *pic = h->short_ref[h->short_ref_count - 1]; |
791 | 148k | remove_short(h, pic->frame_num, 0); |
792 | 148k | } |
793 | 151k | } |
794 | | |
795 | 10.1M | for (int i = 0; i < h->short_ref_count; i++) { |
796 | 7.07M | H264Picture *pic = h->short_ref[i]; |
797 | 7.07M | if (pic->invalid_gap) { |
798 | 5.01M | int d = av_zero_extend(h->cur_pic_ptr->frame_num - pic->frame_num, h->ps.sps->log2_max_frame_num); |
799 | 5.01M | if (d > h->ps.sps->ref_frame_count) |
800 | 200k | remove_short(h, pic->frame_num, 0); |
801 | 5.01M | } |
802 | 7.07M | } |
803 | | |
804 | 3.02M | print_short_term(h); |
805 | 3.02M | print_long_term(h); |
806 | | |
807 | 777M | for (int i = 0; i < FF_ARRAY_ELEMS(h->ps.pps_list); i++) { |
808 | 774M | if (h->ps.pps_list[i]) { |
809 | 3.46M | const PPS *pps = h->ps.pps_list[i]; |
810 | 3.46M | pps_ref_count[0] = FFMAX(pps_ref_count[0], pps->ref_count[0]); |
811 | 3.46M | pps_ref_count[1] = FFMAX(pps_ref_count[1], pps->ref_count[1]); |
812 | 3.46M | } |
813 | 774M | } |
814 | | |
815 | | // Detect unmarked random access points |
816 | 3.02M | if ( err >= 0 |
817 | 2.38M | && h->long_ref_count==0 |
818 | 1.70M | && ( h->short_ref_count<=2 |
819 | 669k | || pps_ref_count[0] <= 2 && pps_ref_count[1] <= 1 && h->avctx->has_b_frames |
820 | 603k | || pps_ref_count[0] <= 1 + (h->picture_structure != PICT_FRAME) && pps_ref_count[1] <= 1) |
821 | 1.12M | && pps_ref_count[0]<=2 + (h->picture_structure != PICT_FRAME) + (2*!h->has_recovery_point) |
822 | 443k | && h->cur_pic_ptr->f->pict_type == AV_PICTURE_TYPE_I){ |
823 | 238k | h->cur_pic_ptr->recovered |= FRAME_RECOVERED_HEURISTIC; |
824 | 238k | if(!h->avctx->has_b_frames) |
825 | 73.3k | h->frame_recovered |= FRAME_RECOVERED_HEURISTIC; |
826 | 238k | } |
827 | | |
828 | 3.02M | out: |
829 | 3.02M | return (h->avctx->err_recognition & AV_EF_EXPLODE) ? err : 0; |
830 | 3.02M | } |
831 | | |
832 | | int ff_h264_decode_ref_pic_marking(H264SliceContext *sl, GetBitContext *gb, |
833 | | const H2645NAL *nal, void *logctx) |
834 | 1.97M | { |
835 | 1.97M | MMCO *mmco = sl->mmco; |
836 | 1.97M | int nb_mmco = 0; |
837 | | |
838 | 1.97M | if (nal->type == H264_NAL_IDR_SLICE) { // FIXME fields |
839 | 1.07M | skip_bits1(gb); // broken_link |
840 | 1.07M | if (get_bits1(gb)) { |
841 | 457k | mmco[0].opcode = MMCO_LONG; |
842 | 457k | mmco[0].long_arg = 0; |
843 | 457k | nb_mmco = 1; |
844 | 457k | } |
845 | 1.07M | sl->explicit_ref_marking = 1; |
846 | 1.07M | } else { |
847 | 897k | sl->explicit_ref_marking = get_bits1(gb); |
848 | 897k | if (sl->explicit_ref_marking) { |
849 | 367k | int i; |
850 | 1.44M | for (i = 0; i < FF_ARRAY_ELEMS(sl->mmco); i++) { |
851 | 1.44M | MMCOOpcode opcode = get_ue_golomb_31(gb); |
852 | | |
853 | 1.44M | mmco[i].opcode = opcode; |
854 | 1.44M | if (opcode == MMCO_SHORT2UNUSED || opcode == MMCO_SHORT2LONG) { |
855 | 248k | mmco[i].short_pic_num = |
856 | 248k | (sl->curr_pic_num - get_ue_golomb_long(gb) - 1) & |
857 | 248k | (sl->max_pic_num - 1); |
858 | 248k | } |
859 | 1.44M | if (opcode == MMCO_SHORT2LONG || opcode == MMCO_LONG2UNUSED || |
860 | 1.23M | opcode == MMCO_LONG || opcode == MMCO_SET_MAX_LONG) { |
861 | 861k | unsigned int long_arg = get_ue_golomb_31(gb); |
862 | 861k | if (long_arg >= 32 || |
863 | 851k | (long_arg >= 16 && !(opcode == MMCO_SET_MAX_LONG && |
864 | 1.76k | long_arg == 16) && |
865 | 13.5k | !(opcode == MMCO_LONG2UNUSED && FIELD_PICTURE(sl)))) { |
866 | 13.5k | av_log(logctx, AV_LOG_ERROR, |
867 | 13.5k | "illegal long ref in memory management control " |
868 | 13.5k | "operation %d\n", opcode); |
869 | 13.5k | sl->nb_mmco = i; |
870 | 13.5k | return -1; |
871 | 13.5k | } |
872 | 848k | mmco[i].long_arg = long_arg; |
873 | 848k | } |
874 | | |
875 | 1.42M | if (opcode > (unsigned) MMCO_LONG) { |
876 | 90.3k | av_log(logctx, AV_LOG_ERROR, |
877 | 90.3k | "illegal memory management control operation %d\n", |
878 | 90.3k | opcode); |
879 | 90.3k | sl->nb_mmco = i; |
880 | 90.3k | return -1; |
881 | 90.3k | } |
882 | 1.33M | if (opcode == MMCO_END) |
883 | 259k | break; |
884 | 1.33M | } |
885 | 263k | nb_mmco = i; |
886 | 263k | } |
887 | 897k | } |
888 | | |
889 | 1.87M | sl->nb_mmco = nb_mmco; |
890 | | |
891 | 1.87M | return 0; |
892 | 1.97M | } |