rtmp-av1.c 15 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635
  1. /******************************************************************************
  2. Copyright (C) 2023 by Lain Bailey <[email protected]>
  3. This program is free software: you can redistribute it and/or modify
  4. it under the terms of the GNU General Public License as published by
  5. the Free Software Foundation, either version 2 of the License, or
  6. (at your option) any later version.
  7. This program is distributed in the hope that it will be useful,
  8. but WITHOUT ANY WARRANTY; without even the implied warranty of
  9. MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  10. GNU General Public License for more details.
  11. You should have received a copy of the GNU General Public License
  12. along with this program. If not, see <http://www.gnu.org/licenses/>.
  13. ******************************************************************************/
  14. #include "rtmp-av1.h"
  15. #include "utils.h"
  16. #include <obs.h>
  17. #include <util/array-serializer.h>
  18. /* Adapted from FFmpeg's libavformat/av1.c for our FLV muxer. */
  19. #define AV1_OBU_SEQUENCE_HEADER 1
  20. #define AV1_OBU_TEMPORAL_DELIMITER 2
  21. #define AV1_OBU_REDUNDANT_FRAME_HEADER 7
  22. #define AV1_OBU_TILE_LIST 8
  23. #define AV1_OBU_PADDING 15
  24. #define AV1_OBU_METADATA 5
  25. #define AV1_OBU_TILE_GROUP 4
  26. #define AV1_OBU_TILE_LIST 8
  27. #define AV1_OBU_FRAME 6
  28. #define AV1_OBU_FRAME_HEADER 3
  29. #define FF_PROFILE_AV1_MAIN 0
  30. #define FF_PROFILE_AV1_HIGH 1
  31. #define FF_PROFILE_AV1_PROFESSIONAL 2
  32. enum frame_type {
  33. AV1_KEY_FRAME,
  34. AV1_INTER_FRAME,
  35. AV1_INTRA_FRAME,
  36. AV1_SWITCH_FRAME,
  37. };
  38. typedef struct AV1SequenceParameters {
  39. uint8_t profile;
  40. uint8_t level;
  41. uint8_t tier;
  42. uint8_t bitdepth;
  43. uint8_t monochrome;
  44. uint8_t chroma_subsampling_x;
  45. uint8_t chroma_subsampling_y;
  46. uint8_t chroma_sample_position;
  47. uint8_t color_description_present_flag;
  48. uint8_t color_primaries;
  49. uint8_t transfer_characteristics;
  50. uint8_t matrix_coefficients;
  51. uint8_t color_range;
  52. } AV1SequenceParameters;
  53. #define MAX_OBU_HEADER_SIZE (2 + 8)
  54. typedef struct Av1GetBitContext {
  55. const uint8_t *buffer, *buffer_end;
  56. int index;
  57. int size_in_bits;
  58. int size_in_bits_plus8;
  59. } Av1GetBitContext;
  60. static inline int init_get_bits_xe(Av1GetBitContext *s, const uint8_t *buffer,
  61. int bit_size)
  62. {
  63. int buffer_size;
  64. int ret = 0;
  65. if (bit_size >= INT_MAX - 64 * 8 || bit_size < 0 || !buffer) {
  66. bit_size = 0;
  67. buffer = NULL;
  68. ret = -1;
  69. }
  70. buffer_size = (bit_size + 7) >> 3;
  71. s->buffer = buffer;
  72. s->size_in_bits = bit_size;
  73. s->size_in_bits_plus8 = bit_size + 8;
  74. s->buffer_end = buffer + buffer_size;
  75. s->index = 0;
  76. return ret;
  77. }
  78. static inline int init_get_bits(Av1GetBitContext *s, const uint8_t *buffer,
  79. int bit_size)
  80. {
  81. return init_get_bits_xe(s, buffer, bit_size);
  82. }
  83. static inline int init_get_bits8(Av1GetBitContext *s, const uint8_t *buffer,
  84. int byte_size)
  85. {
  86. if (byte_size > INT_MAX / 8 || byte_size < 0)
  87. byte_size = -1;
  88. return init_get_bits(s, buffer, byte_size * 8);
  89. }
  90. static inline unsigned int get_bit1(Av1GetBitContext *s)
  91. {
  92. unsigned int index = s->index;
  93. uint8_t result = s->buffer[index >> 3];
  94. result <<= index & 7;
  95. result >>= 8 - 1;
  96. if (s->index < s->size_in_bits_plus8)
  97. index++;
  98. s->index = index;
  99. return result;
  100. }
  101. static inline unsigned int get_bits(Av1GetBitContext *s, unsigned int n)
  102. {
  103. unsigned int out = 0;
  104. for (unsigned int i = 0; i < n; i++)
  105. out = (out << 1) | get_bit1(s);
  106. return out;
  107. }
  108. #define skip_bits get_bits
  109. static inline int get_bits_count(Av1GetBitContext *s)
  110. {
  111. return s->index;
  112. }
  113. static inline int get_bits_left(Av1GetBitContext *gb)
  114. {
  115. return gb->size_in_bits - get_bits_count(gb);
  116. }
  117. #define get_bits_long get_bits
  118. #define skip_bits_long get_bits_long
  119. static inline int64_t leb128(Av1GetBitContext *gb)
  120. {
  121. int64_t ret = 0;
  122. int i;
  123. for (i = 0; i < 8; i++) {
  124. int byte = get_bits(gb, 8);
  125. ret |= (int64_t)(byte & 0x7f) << (i * 7);
  126. if (!(byte & 0x80))
  127. break;
  128. }
  129. return ret;
  130. }
  131. static inline void uvlc(Av1GetBitContext *gb)
  132. {
  133. int leading_zeros = 0;
  134. while (get_bits_left(gb)) {
  135. if (get_bits(gb, 1))
  136. break;
  137. leading_zeros++;
  138. }
  139. if (leading_zeros >= 32)
  140. return;
  141. skip_bits_long(gb, leading_zeros);
  142. }
  143. static inline int parse_obu_header(const uint8_t *buf, int buf_size,
  144. int64_t *obu_size, int *start_pos, int *type,
  145. int *temporal_id, int *spatial_id)
  146. {
  147. Av1GetBitContext gb;
  148. int ret, extension_flag, has_size_flag;
  149. size_t size;
  150. ret = init_get_bits8(&gb, buf, min_i32(buf_size, MAX_OBU_HEADER_SIZE));
  151. if (ret < 0)
  152. return ret;
  153. if (get_bits(&gb, 1) != 0) // obu_forbidden_bit
  154. return -1;
  155. *type = get_bits(&gb, 4);
  156. extension_flag = get_bits(&gb, 1);
  157. has_size_flag = get_bits(&gb, 1);
  158. skip_bits(&gb, 1); // obu_reserved_1bit
  159. if (extension_flag) {
  160. *temporal_id = get_bits(&gb, 3);
  161. *spatial_id = get_bits(&gb, 2);
  162. skip_bits(&gb, 3); // extension_header_reserved_3bits
  163. } else {
  164. *temporal_id = *spatial_id = 0;
  165. }
  166. *obu_size = has_size_flag ? leb128(&gb) : buf_size - 1 - extension_flag;
  167. if (get_bits_left(&gb) < 0)
  168. return -1;
  169. *start_pos = get_bits_count(&gb) / 8;
  170. size = (size_t)(*obu_size + *start_pos);
  171. if (size > (size_t)buf_size)
  172. return -1;
  173. assert(size <= INT_MAX);
  174. return (int)size;
  175. }
  176. static inline int get_obu_bit_length(const uint8_t *buf, int size, int type)
  177. {
  178. int v;
  179. /* There are no trailing bits on these */
  180. if (type == AV1_OBU_TILE_GROUP || type == AV1_OBU_TILE_LIST ||
  181. type == AV1_OBU_FRAME) {
  182. if (size > INT_MAX / 8)
  183. return -1;
  184. else
  185. return size * 8;
  186. }
  187. while (size > 0 && buf[size - 1] == 0)
  188. size--;
  189. if (!size)
  190. return 0;
  191. v = buf[size - 1];
  192. if (size > INT_MAX / 8)
  193. return -1;
  194. size *= 8;
  195. /* Remove the trailing_one_bit and following trailing zeros */
  196. if (v)
  197. size -= ctz32(v) + 1;
  198. return size;
  199. }
  200. static int parse_color_config(AV1SequenceParameters *seq_params,
  201. Av1GetBitContext *gb)
  202. {
  203. int twelve_bit = 0;
  204. int high_bitdepth = get_bits(gb, 1);
  205. if (seq_params->profile == FF_PROFILE_AV1_PROFESSIONAL && high_bitdepth)
  206. twelve_bit = get_bits(gb, 1);
  207. seq_params->bitdepth = 8 + (high_bitdepth * 2) + (twelve_bit * 2);
  208. if (seq_params->profile == FF_PROFILE_AV1_HIGH)
  209. seq_params->monochrome = 0;
  210. else
  211. seq_params->monochrome = get_bits(gb, 1);
  212. seq_params->color_description_present_flag = get_bits(gb, 1);
  213. if (seq_params->color_description_present_flag) {
  214. seq_params->color_primaries = get_bits(gb, 8);
  215. seq_params->transfer_characteristics = get_bits(gb, 8);
  216. seq_params->matrix_coefficients = get_bits(gb, 8);
  217. } else {
  218. seq_params->color_primaries = 2;
  219. seq_params->transfer_characteristics = 2;
  220. seq_params->matrix_coefficients = 2;
  221. }
  222. if (seq_params->monochrome) {
  223. seq_params->color_range = get_bits(gb, 1);
  224. seq_params->chroma_subsampling_x = 1;
  225. seq_params->chroma_subsampling_y = 1;
  226. seq_params->chroma_sample_position = 0;
  227. return 0;
  228. } else if (seq_params->color_primaries == 1 &&
  229. seq_params->transfer_characteristics == 13 &&
  230. seq_params->matrix_coefficients == 0) {
  231. seq_params->chroma_subsampling_x = 0;
  232. seq_params->chroma_subsampling_y = 0;
  233. } else {
  234. seq_params->color_range = get_bits(gb, 1);
  235. if (seq_params->profile == FF_PROFILE_AV1_MAIN) {
  236. seq_params->chroma_subsampling_x = 1;
  237. seq_params->chroma_subsampling_y = 1;
  238. } else if (seq_params->profile == FF_PROFILE_AV1_HIGH) {
  239. seq_params->chroma_subsampling_x = 0;
  240. seq_params->chroma_subsampling_y = 0;
  241. } else {
  242. if (twelve_bit) {
  243. seq_params->chroma_subsampling_x =
  244. get_bits(gb, 1);
  245. if (seq_params->chroma_subsampling_x)
  246. seq_params->chroma_subsampling_y =
  247. get_bits(gb, 1);
  248. else
  249. seq_params->chroma_subsampling_y = 0;
  250. } else {
  251. seq_params->chroma_subsampling_x = 1;
  252. seq_params->chroma_subsampling_y = 0;
  253. }
  254. }
  255. if (seq_params->chroma_subsampling_x &&
  256. seq_params->chroma_subsampling_y)
  257. seq_params->chroma_sample_position = get_bits(gb, 2);
  258. }
  259. skip_bits(gb, 1); // separate_uv_delta_q
  260. return 0;
  261. }
  262. static int parse_sequence_header(AV1SequenceParameters *seq_params,
  263. const uint8_t *buf, int size)
  264. {
  265. Av1GetBitContext gb;
  266. int reduced_still_picture_header;
  267. int frame_width_bits_minus_1, frame_height_bits_minus_1;
  268. int size_bits, ret;
  269. size_bits = get_obu_bit_length(buf, size, AV1_OBU_SEQUENCE_HEADER);
  270. if (size_bits < 0)
  271. return size_bits;
  272. ret = init_get_bits(&gb, buf, size_bits);
  273. if (ret < 0)
  274. return ret;
  275. memset(seq_params, 0, sizeof(*seq_params));
  276. seq_params->profile = get_bits(&gb, 3);
  277. skip_bits(&gb, 1); // still_picture
  278. reduced_still_picture_header = get_bits(&gb, 1);
  279. if (reduced_still_picture_header) {
  280. seq_params->level = get_bits(&gb, 5);
  281. seq_params->tier = 0;
  282. } else {
  283. int initial_display_delay_present_flag,
  284. operating_points_cnt_minus_1;
  285. int decoder_model_info_present_flag,
  286. buffer_delay_length_minus_1;
  287. if (get_bits(&gb, 1)) { // timing_info_present_flag
  288. skip_bits_long(&gb, 32); // num_units_in_display_tick
  289. skip_bits_long(&gb, 32); // time_scale
  290. if (get_bits(&gb, 1)) // equal_picture_interval
  291. uvlc(&gb); // num_ticks_per_picture_minus_1
  292. decoder_model_info_present_flag = get_bits(&gb, 1);
  293. if (decoder_model_info_present_flag) {
  294. buffer_delay_length_minus_1 = get_bits(&gb, 5);
  295. skip_bits_long(&gb, 32);
  296. skip_bits(&gb, 10);
  297. }
  298. } else
  299. decoder_model_info_present_flag = 0;
  300. initial_display_delay_present_flag = get_bits(&gb, 1);
  301. operating_points_cnt_minus_1 = get_bits(&gb, 5);
  302. for (int i = 0; i <= operating_points_cnt_minus_1; i++) {
  303. int seq_level_idx, seq_tier;
  304. skip_bits(&gb, 12);
  305. seq_level_idx = get_bits(&gb, 5);
  306. if (seq_level_idx > 7)
  307. seq_tier = get_bits(&gb, 1);
  308. else
  309. seq_tier = 0;
  310. if (decoder_model_info_present_flag) {
  311. if (get_bits(&gb, 1)) {
  312. skip_bits_long(
  313. &gb,
  314. buffer_delay_length_minus_1 +
  315. 1);
  316. skip_bits_long(
  317. &gb,
  318. buffer_delay_length_minus_1 +
  319. 1);
  320. skip_bits(&gb, 1);
  321. }
  322. }
  323. if (initial_display_delay_present_flag) {
  324. if (get_bits(&gb, 1))
  325. skip_bits(&gb, 4);
  326. }
  327. if (i == 0) {
  328. seq_params->level = seq_level_idx;
  329. seq_params->tier = seq_tier;
  330. }
  331. }
  332. }
  333. frame_width_bits_minus_1 = get_bits(&gb, 4);
  334. frame_height_bits_minus_1 = get_bits(&gb, 4);
  335. skip_bits(&gb, frame_width_bits_minus_1 + 1); // max_frame_width_minus_1
  336. skip_bits(&gb,
  337. frame_height_bits_minus_1 + 1); // max_frame_height_minus_1
  338. if (!reduced_still_picture_header) {
  339. if (get_bits(&gb, 1)) // frame_id_numbers_present_flag
  340. skip_bits(&gb, 7);
  341. }
  342. skip_bits(
  343. &gb,
  344. 3); // use_128x128_superblock (1), enable_filter_intra (1), enable_intra_edge_filter (1)
  345. if (!reduced_still_picture_header) {
  346. int enable_order_hint, seq_force_screen_content_tools;
  347. skip_bits(&gb, 4);
  348. enable_order_hint = get_bits(&gb, 1);
  349. if (enable_order_hint)
  350. skip_bits(&gb, 2);
  351. if (get_bits(&gb, 1)) // seq_choose_screen_content_tools
  352. seq_force_screen_content_tools = 2;
  353. else
  354. seq_force_screen_content_tools = get_bits(&gb, 1);
  355. if (seq_force_screen_content_tools) {
  356. if (!get_bits(&gb, 1)) // seq_choose_integer_mv
  357. skip_bits(&gb, 1); // seq_force_integer_mv
  358. }
  359. if (enable_order_hint)
  360. skip_bits(&gb, 3); // order_hint_bits_minus_1
  361. }
  362. skip_bits(&gb, 3);
  363. parse_color_config(seq_params, &gb);
  364. skip_bits(&gb, 1); // film_grain_params_present
  365. if (get_bits_left(&gb))
  366. return -1;
  367. return 0;
  368. }
  369. size_t obs_parse_av1_header(uint8_t **header, const uint8_t *data, size_t size)
  370. {
  371. if (data[0] & 0x80) {
  372. int config_record_version = data[0] & 0x7f;
  373. if (config_record_version != 1 || size < 4)
  374. return 0;
  375. *header = bmemdup(data, size);
  376. return size;
  377. }
  378. // AV1S init
  379. AV1SequenceParameters seq_params;
  380. int nb_seq = 0, seq_size = 0, meta_size = 0;
  381. const uint8_t *seq = 0, *meta = 0;
  382. uint8_t *buf = (uint8_t *)data;
  383. while (size > 0) {
  384. int64_t obu_size;
  385. int start_pos, type, temporal_id, spatial_id;
  386. assert(size <= INT_MAX);
  387. int len = parse_obu_header(buf, (int)size, &obu_size,
  388. &start_pos, &type, &temporal_id,
  389. &spatial_id);
  390. if (len < 0)
  391. return 0;
  392. switch (type) {
  393. case AV1_OBU_SEQUENCE_HEADER:
  394. nb_seq++;
  395. if (!obu_size || nb_seq > 1) {
  396. return 0;
  397. }
  398. assert(obu_size <= INT_MAX);
  399. if (parse_sequence_header(&seq_params, buf + start_pos,
  400. (int)obu_size) < 0)
  401. return 0;
  402. seq = buf;
  403. seq_size = len;
  404. break;
  405. case AV1_OBU_METADATA:
  406. if (!obu_size)
  407. return 0;
  408. meta = buf;
  409. meta_size = len;
  410. break;
  411. default:
  412. break;
  413. }
  414. size -= len;
  415. buf += len;
  416. }
  417. if (!nb_seq)
  418. return 0;
  419. uint8_t av1header[4];
  420. av1header[0] = (1 << 7) | 1; // marker and version
  421. av1header[1] = (seq_params.profile << 5) | (seq_params.level);
  422. av1header[2] = (seq_params.tier << 7) | (seq_params.bitdepth > 8) << 6 |
  423. (seq_params.bitdepth == 12) << 5 |
  424. (seq_params.monochrome) << 4 |
  425. (seq_params.chroma_subsampling_x) << 3 |
  426. (seq_params.chroma_subsampling_y) << 2 |
  427. (seq_params.chroma_sample_position);
  428. av1header[3] = 0;
  429. struct array_output_data output;
  430. struct serializer s;
  431. array_output_serializer_init(&s, &output);
  432. s_write(&s, av1header, sizeof(av1header));
  433. if (seq_size)
  434. s_write(&s, seq, seq_size);
  435. if (meta_size)
  436. s_write(&s, meta, meta_size);
  437. *header = output.bytes.array;
  438. return output.bytes.num;
  439. }
  440. static void compute_av1_keyframe_priority(const uint8_t *buf, bool *is_keyframe,
  441. int *priority)
  442. {
  443. /* Skip if the packet already has a priority, e.g., assigned by the
  444. * encoder implementation (currently QSV/AMF). */
  445. if (*priority)
  446. return;
  447. Av1GetBitContext gb;
  448. init_get_bits8(&gb, buf, 1);
  449. // show_existing_frame
  450. if (get_bit1(&gb))
  451. return;
  452. enum frame_type type = get_bits(&gb, 2);
  453. bool show_frame = get_bit1(&gb);
  454. switch (type) {
  455. case AV1_KEY_FRAME:
  456. *is_keyframe = true;
  457. *priority = 3;
  458. break;
  459. case AV1_INTER_FRAME:
  460. *priority = show_frame ? 1 : 2;
  461. break;
  462. case AV1_INTRA_FRAME:
  463. *priority = 3;
  464. break;
  465. case AV1_SWITCH_FRAME:
  466. *priority = 2;
  467. break;
  468. }
  469. }
  470. static void serialize_av1_data(struct serializer *s, const uint8_t *data,
  471. size_t size, bool *is_keyframe, int *priority)
  472. {
  473. uint8_t *buf = (uint8_t *)data;
  474. uint8_t *end = (uint8_t *)data + size;
  475. while (buf < end) {
  476. int64_t obu_size;
  477. int start_pos, type, temporal_id, spatial_id;
  478. assert(end - buf <= INT_MAX);
  479. int len = parse_obu_header(buf, (int)(end - buf), &obu_size,
  480. &start_pos, &type, &temporal_id,
  481. &spatial_id);
  482. if (len < 0)
  483. return;
  484. switch (type) {
  485. case AV1_OBU_TEMPORAL_DELIMITER:
  486. case AV1_OBU_REDUNDANT_FRAME_HEADER:
  487. case AV1_OBU_TILE_LIST:
  488. break;
  489. case AV1_OBU_FRAME:
  490. case AV1_OBU_FRAME_HEADER:
  491. compute_av1_keyframe_priority(buf + start_pos,
  492. is_keyframe, priority);
  493. /* Falls through. */
  494. default:
  495. s_write(s, buf, len);
  496. size += len;
  497. break;
  498. }
  499. buf += len;
  500. }
  501. }
  502. void obs_parse_av1_packet(struct encoder_packet *av1_packet,
  503. const struct encoder_packet *src)
  504. {
  505. struct array_output_data output;
  506. struct serializer s;
  507. long ref = 1;
  508. array_output_serializer_init(&s, &output);
  509. serialize(&s, &ref, sizeof(ref));
  510. *av1_packet = *src;
  511. serialize_av1_data(&s, src->data, src->size, &av1_packet->keyframe,
  512. &av1_packet->priority);
  513. av1_packet->data = output.bytes.array + sizeof(ref);
  514. av1_packet->size = output.bytes.num - sizeof(ref);
  515. av1_packet->drop_priority = av1_packet->priority;
  516. }