You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

757 lines
28KB

  1. /*
  2. * DV encoder
  3. * Copyright (c) 2003 Roman Shaposhnik
  4. *
  5. * This file is part of Libav.
  6. *
  7. * Libav 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. * Libav 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 Libav; if not, write to the Free Software
  19. * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20. */
  21. /**
  22. * @file
  23. * DV encoder
  24. */
  25. #include "config.h"
  26. #include "libavutil/attributes.h"
  27. #include "libavutil/internal.h"
  28. #include "libavutil/pixdesc.h"
  29. #include "avcodec.h"
  30. #include "dv.h"
  31. #include "dv_profile_internal.h"
  32. #include "dv_tablegen.h"
  33. #include "fdctdsp.h"
  34. #include "internal.h"
  35. #include "mathops.h"
  36. #include "me_cmp.h"
  37. #include "pixblockdsp.h"
  38. #include "put_bits.h"
  39. static av_cold int dvvideo_encode_init(AVCodecContext *avctx)
  40. {
  41. DVVideoContext *s = avctx->priv_data;
  42. FDCTDSPContext fdsp;
  43. MECmpContext mecc;
  44. PixblockDSPContext pdsp;
  45. int ret;
  46. s->sys = av_dv_codec_profile(avctx->width, avctx->height, avctx->pix_fmt);
  47. if (!s->sys) {
  48. av_log(avctx, AV_LOG_ERROR, "Found no DV profile for %ix%i %s video. "
  49. "Valid DV profiles are:\n",
  50. avctx->width, avctx->height, av_get_pix_fmt_name(avctx->pix_fmt));
  51. ff_dv_print_profiles(avctx, AV_LOG_ERROR);
  52. return AVERROR(EINVAL);
  53. }
  54. ret = ff_dv_init_dynamic_tables(s, s->sys);
  55. if (ret < 0) {
  56. av_log(avctx, AV_LOG_ERROR, "Error initializing work tables.\n");
  57. return ret;
  58. }
  59. dv_vlc_map_tableinit();
  60. ff_fdctdsp_init(&fdsp, avctx);
  61. ff_me_cmp_init(&mecc, avctx);
  62. ff_pixblockdsp_init(&pdsp, avctx);
  63. ff_set_cmp(&mecc, mecc.ildct_cmp, avctx->ildct_cmp);
  64. s->get_pixels = pdsp.get_pixels;
  65. s->ildct_cmp = mecc.ildct_cmp[5];
  66. s->fdct[0] = fdsp.fdct;
  67. s->fdct[1] = fdsp.fdct248;
  68. return ff_dvvideo_init(avctx);
  69. }
  70. /* bit budget for AC only in 5 MBs */
  71. static const int vs_total_ac_bits = (100 * 4 + 68 * 2) * 5;
  72. static const int mb_area_start[5] = { 1, 6, 21, 43, 64 };
  73. #if CONFIG_SMALL
  74. /* Convert run and level (where level != 0) pair into VLC, returning bit size */
  75. static av_always_inline int dv_rl2vlc(int run, int level, int sign,
  76. uint32_t *vlc)
  77. {
  78. int size;
  79. if (run < DV_VLC_MAP_RUN_SIZE && level < DV_VLC_MAP_LEV_SIZE) {
  80. *vlc = dv_vlc_map[run][level].vlc | sign;
  81. size = dv_vlc_map[run][level].size;
  82. } else {
  83. if (level < DV_VLC_MAP_LEV_SIZE) {
  84. *vlc = dv_vlc_map[0][level].vlc | sign;
  85. size = dv_vlc_map[0][level].size;
  86. } else {
  87. *vlc = 0xfe00 | (level << 1) | sign;
  88. size = 16;
  89. }
  90. if (run) {
  91. *vlc |= ((run < 16) ? dv_vlc_map[run - 1][0].vlc :
  92. (0x1f80 | (run - 1))) << size;
  93. size += (run < 16) ? dv_vlc_map[run - 1][0].size : 13;
  94. }
  95. }
  96. return size;
  97. }
  98. static av_always_inline int dv_rl2vlc_size(int run, int level)
  99. {
  100. int size;
  101. if (run < DV_VLC_MAP_RUN_SIZE && level < DV_VLC_MAP_LEV_SIZE) {
  102. size = dv_vlc_map[run][level].size;
  103. } else {
  104. size = (level < DV_VLC_MAP_LEV_SIZE) ? dv_vlc_map[0][level].size : 16;
  105. if (run)
  106. size += (run < 16) ? dv_vlc_map[run - 1][0].size : 13;
  107. }
  108. return size;
  109. }
  110. #else
  111. static av_always_inline int dv_rl2vlc(int run, int l, int sign, uint32_t *vlc)
  112. {
  113. *vlc = dv_vlc_map[run][l].vlc | sign;
  114. return dv_vlc_map[run][l].size;
  115. }
  116. static av_always_inline int dv_rl2vlc_size(int run, int l)
  117. {
  118. return dv_vlc_map[run][l].size;
  119. }
  120. #endif
  121. typedef struct EncBlockInfo {
  122. int area_q[4];
  123. int bit_size[4];
  124. int prev[5];
  125. int cur_ac;
  126. int cno;
  127. int dct_mode;
  128. int16_t mb[64];
  129. uint8_t next[64];
  130. uint8_t sign[64];
  131. uint8_t partial_bit_count;
  132. uint32_t partial_bit_buffer; /* we can't use uint16_t here */
  133. } EncBlockInfo;
  134. static av_always_inline PutBitContext *dv_encode_ac(EncBlockInfo *bi,
  135. PutBitContext *pb_pool,
  136. PutBitContext *pb_end)
  137. {
  138. int prev, bits_left;
  139. PutBitContext *pb = pb_pool;
  140. int size = bi->partial_bit_count;
  141. uint32_t vlc = bi->partial_bit_buffer;
  142. bi->partial_bit_count =
  143. bi->partial_bit_buffer = 0;
  144. for (;;) {
  145. /* Find suitable storage space */
  146. for (; size > (bits_left = put_bits_left(pb)); pb++) {
  147. if (bits_left) {
  148. size -= bits_left;
  149. put_bits(pb, bits_left, vlc >> size);
  150. vlc = vlc & ((1 << size) - 1);
  151. }
  152. if (pb + 1 >= pb_end) {
  153. bi->partial_bit_count = size;
  154. bi->partial_bit_buffer = vlc;
  155. return pb;
  156. }
  157. }
  158. /* Store VLC */
  159. put_bits(pb, size, vlc);
  160. if (bi->cur_ac >= 64)
  161. break;
  162. /* Construct the next VLC */
  163. prev = bi->cur_ac;
  164. bi->cur_ac = bi->next[prev];
  165. if (bi->cur_ac < 64) {
  166. size = dv_rl2vlc(bi->cur_ac - prev - 1, bi->mb[bi->cur_ac],
  167. bi->sign[bi->cur_ac], &vlc);
  168. } else {
  169. size = 4;
  170. vlc = 6; /* End Of Block stamp */
  171. }
  172. }
  173. return pb;
  174. }
  175. static av_always_inline int dv_guess_dct_mode(DVVideoContext *s, uint8_t *data,
  176. ptrdiff_t linesize)
  177. {
  178. if (s->avctx->flags & AV_CODEC_FLAG_INTERLACED_DCT) {
  179. int ps = s->ildct_cmp(NULL, data, NULL, linesize, 8) - 400;
  180. if (ps > 0) {
  181. int is = s->ildct_cmp(NULL, data, NULL, linesize << 1, 4) +
  182. s->ildct_cmp(NULL, data + linesize, NULL, linesize << 1, 4);
  183. return ps > is;
  184. }
  185. }
  186. return 0;
  187. }
  188. static const int dv_weight_bits = 18;
  189. static const int dv_weight_88[64] = {
  190. 131072, 257107, 257107, 242189, 252167, 242189, 235923, 237536,
  191. 237536, 235923, 229376, 231390, 223754, 231390, 229376, 222935,
  192. 224969, 217965, 217965, 224969, 222935, 200636, 218652, 211916,
  193. 212325, 211916, 218652, 200636, 188995, 196781, 205965, 206433,
  194. 206433, 205965, 196781, 188995, 185364, 185364, 200636, 200704,
  195. 200636, 185364, 185364, 174609, 180568, 195068, 195068, 180568,
  196. 174609, 170091, 175557, 189591, 175557, 170091, 165371, 170627,
  197. 170627, 165371, 160727, 153560, 160727, 144651, 144651, 136258,
  198. };
  199. static const int dv_weight_248[64] = {
  200. 131072, 242189, 257107, 237536, 229376, 200636, 242189, 223754,
  201. 224969, 196781, 262144, 242189, 229376, 200636, 257107, 237536,
  202. 211916, 185364, 235923, 217965, 229376, 211916, 206433, 180568,
  203. 242189, 223754, 224969, 196781, 211916, 185364, 235923, 217965,
  204. 200704, 175557, 222935, 205965, 200636, 185364, 195068, 170627,
  205. 229376, 211916, 206433, 180568, 200704, 175557, 222935, 205965,
  206. 175557, 153560, 188995, 174609, 165371, 144651, 200636, 185364,
  207. 195068, 170627, 175557, 153560, 188995, 174609, 165371, 144651,
  208. };
  209. static av_always_inline int dv_init_enc_block(EncBlockInfo *bi, uint8_t *data,
  210. ptrdiff_t linesize,
  211. DVVideoContext *s, int bias)
  212. {
  213. const int *weight;
  214. const uint8_t *zigzag_scan;
  215. LOCAL_ALIGNED_16(int16_t, blk, [64]);
  216. int i, area;
  217. /* We offer two different methods for class number assignment: the
  218. * method suggested in SMPTE 314M Table 22, and an improved
  219. * method. The SMPTE method is very conservative; it assigns class
  220. * 3 (i.e. severe quantization) to any block where the largest AC
  221. * component is greater than 36. Libav's DV encoder tracks AC bit
  222. * consumption precisely, so there is no need to bias most blocks
  223. * towards strongly lossy compression. Instead, we assign class 2
  224. * to most blocks, and use class 3 only when strictly necessary
  225. * (for blocks whose largest AC component exceeds 255). */
  226. #if 0 /* SMPTE spec method */
  227. static const int classes[] = { 12, 24, 36, 0xffff };
  228. #else /* improved Libav method */
  229. static const int classes[] = { -1, -1, 255, 0xffff };
  230. #endif
  231. int max = classes[0];
  232. int prev = 0;
  233. assert((((int) blk) & 15) == 0);
  234. bi->area_q[0] =
  235. bi->area_q[1] =
  236. bi->area_q[2] =
  237. bi->area_q[3] = 0;
  238. bi->partial_bit_count = 0;
  239. bi->partial_bit_buffer = 0;
  240. bi->cur_ac = 0;
  241. if (data) {
  242. bi->dct_mode = dv_guess_dct_mode(s, data, linesize);
  243. s->get_pixels(blk, data, linesize);
  244. s->fdct[bi->dct_mode](blk);
  245. } else {
  246. /* We rely on the fact that encoding all zeros leads to an immediate
  247. * EOB, which is precisely what the spec calls for in the "dummy"
  248. * blocks. */
  249. memset(blk, 0, 64 * sizeof(*blk));
  250. bi->dct_mode = 0;
  251. }
  252. bi->mb[0] = blk[0];
  253. zigzag_scan = bi->dct_mode ? ff_dv_zigzag248_direct : ff_zigzag_direct;
  254. weight = bi->dct_mode ? dv_weight_248 : dv_weight_88;
  255. for (area = 0; area < 4; area++) {
  256. bi->prev[area] = prev;
  257. bi->bit_size[area] = 1; // 4 areas 4 bits for EOB :)
  258. for (i = mb_area_start[area]; i < mb_area_start[area + 1]; i++) {
  259. int level = blk[zigzag_scan[i]];
  260. if (level + 15 > 30U) {
  261. bi->sign[i] = (level >> 31) & 1;
  262. /* Weight it and and shift down into range, adding for rounding.
  263. * The extra division by a factor of 2^4 reverses the 8x
  264. * expansion of the DCT AND the 2x doubling of the weights. */
  265. level = (FFABS(level) * weight[i] + (1 << (dv_weight_bits + 3))) >>
  266. (dv_weight_bits + 4);
  267. bi->mb[i] = level;
  268. if (level > max)
  269. max = level;
  270. bi->bit_size[area] += dv_rl2vlc_size(i - prev - 1, level);
  271. bi->next[prev] = i;
  272. prev = i;
  273. }
  274. }
  275. }
  276. bi->next[prev] = i;
  277. for (bi->cno = 0; max > classes[bi->cno]; bi->cno++)
  278. ;
  279. bi->cno += bias;
  280. if (bi->cno >= 3) {
  281. bi->cno = 3;
  282. prev = 0;
  283. i = bi->next[prev];
  284. for (area = 0; area < 4; area++) {
  285. bi->prev[area] = prev;
  286. bi->bit_size[area] = 1; // 4 areas 4 bits for EOB :)
  287. for (; i < mb_area_start[area + 1]; i = bi->next[i]) {
  288. bi->mb[i] >>= 1;
  289. if (bi->mb[i]) {
  290. bi->bit_size[area] += dv_rl2vlc_size(i - prev - 1, bi->mb[i]);
  291. bi->next[prev] = i;
  292. prev = i;
  293. }
  294. }
  295. }
  296. bi->next[prev] = i;
  297. }
  298. return bi->bit_size[0] + bi->bit_size[1] +
  299. bi->bit_size[2] + bi->bit_size[3];
  300. }
  301. static inline void dv_guess_qnos(EncBlockInfo *blks, int *qnos)
  302. {
  303. int size[5];
  304. int i, j, k, a, prev, a2;
  305. EncBlockInfo *b;
  306. size[0] =
  307. size[1] =
  308. size[2] =
  309. size[3] =
  310. size[4] = 1 << 24;
  311. do {
  312. b = blks;
  313. for (i = 0; i < 5; i++) {
  314. if (!qnos[i])
  315. continue;
  316. qnos[i]--;
  317. size[i] = 0;
  318. for (j = 0; j < 6; j++, b++) {
  319. for (a = 0; a < 4; a++) {
  320. if (b->area_q[a] != ff_dv_quant_shifts[qnos[i] + ff_dv_quant_offset[b->cno]][a]) {
  321. b->bit_size[a] = 1; // 4 areas 4 bits for EOB :)
  322. b->area_q[a]++;
  323. prev = b->prev[a];
  324. assert(b->next[prev] >= mb_area_start[a + 1] || b->mb[prev]);
  325. for (k = b->next[prev]; k < mb_area_start[a + 1]; k = b->next[k]) {
  326. b->mb[k] >>= 1;
  327. if (b->mb[k]) {
  328. b->bit_size[a] += dv_rl2vlc_size(k - prev - 1, b->mb[k]);
  329. prev = k;
  330. } else {
  331. if (b->next[k] >= mb_area_start[a + 1] && b->next[k] < 64) {
  332. for (a2 = a + 1; b->next[k] >= mb_area_start[a2 + 1]; a2++)
  333. b->prev[a2] = prev;
  334. assert(a2 < 4);
  335. assert(b->mb[b->next[k]]);
  336. b->bit_size[a2] += dv_rl2vlc_size(b->next[k] - prev - 1, b->mb[b->next[k]]) -
  337. dv_rl2vlc_size(b->next[k] - k - 1, b->mb[b->next[k]]);
  338. assert(b->prev[a2] == k && (a2 + 1 >= 4 || b->prev[a2 + 1] != k));
  339. b->prev[a2] = prev;
  340. }
  341. b->next[prev] = b->next[k];
  342. }
  343. }
  344. b->prev[a + 1] = prev;
  345. }
  346. size[i] += b->bit_size[a];
  347. }
  348. }
  349. if (vs_total_ac_bits >= size[0] + size[1] + size[2] + size[3] + size[4])
  350. return;
  351. }
  352. } while (qnos[0] | qnos[1] | qnos[2] | qnos[3] | qnos[4]);
  353. for (a = 2; a == 2 || vs_total_ac_bits < size[0]; a += a) {
  354. b = blks;
  355. size[0] = 5 * 6 * 4; // EOB
  356. for (j = 0; j < 6 * 5; j++, b++) {
  357. prev = b->prev[0];
  358. for (k = b->next[prev]; k < 64; k = b->next[k]) {
  359. if (b->mb[k] < a && b->mb[k] > -a) {
  360. b->next[prev] = b->next[k];
  361. } else {
  362. size[0] += dv_rl2vlc_size(k - prev - 1, b->mb[k]);
  363. prev = k;
  364. }
  365. }
  366. }
  367. }
  368. }
  369. static int dv_encode_video_segment(AVCodecContext *avctx, void *arg)
  370. {
  371. DVVideoContext *s = avctx->priv_data;
  372. DVwork_chunk *work_chunk = arg;
  373. int mb_index, i, j;
  374. int mb_x, mb_y, c_offset;
  375. ptrdiff_t linesize, y_stride;
  376. uint8_t *y_ptr;
  377. uint8_t *dif;
  378. LOCAL_ALIGNED_8(uint8_t, scratch, [128]);
  379. EncBlockInfo enc_blks[5 * DV_MAX_BPM];
  380. PutBitContext pbs[5 * DV_MAX_BPM];
  381. PutBitContext *pb;
  382. EncBlockInfo *enc_blk;
  383. int vs_bit_size = 0;
  384. int qnos[5] = { 15, 15, 15, 15, 15 }; /* No quantization */
  385. int *qnosp = &qnos[0];
  386. dif = &s->buf[work_chunk->buf_offset * 80];
  387. enc_blk = &enc_blks[0];
  388. for (mb_index = 0; mb_index < 5; mb_index++) {
  389. dv_calculate_mb_xy(s, work_chunk, mb_index, &mb_x, &mb_y);
  390. /* initializing luminance blocks */
  391. if ((s->sys->pix_fmt == AV_PIX_FMT_YUV420P) ||
  392. (s->sys->pix_fmt == AV_PIX_FMT_YUV411P && mb_x >= (704 / 8)) ||
  393. (s->sys->height >= 720 && mb_y != 134)) {
  394. y_stride = s->frame->linesize[0] << 3;
  395. } else {
  396. y_stride = 16;
  397. }
  398. y_ptr = s->frame->data[0] +
  399. ((mb_y * s->frame->linesize[0] + mb_x) << 3);
  400. linesize = s->frame->linesize[0];
  401. if (s->sys->video_stype == 4) { /* SD 422 */
  402. vs_bit_size +=
  403. dv_init_enc_block(enc_blk + 0, y_ptr, linesize, s, 0) +
  404. dv_init_enc_block(enc_blk + 1, NULL, linesize, s, 0) +
  405. dv_init_enc_block(enc_blk + 2, y_ptr + 8, linesize, s, 0) +
  406. dv_init_enc_block(enc_blk + 3, NULL, linesize, s, 0);
  407. } else {
  408. vs_bit_size +=
  409. dv_init_enc_block(enc_blk + 0, y_ptr, linesize, s, 0) +
  410. dv_init_enc_block(enc_blk + 1, y_ptr + 8, linesize, s, 0) +
  411. dv_init_enc_block(enc_blk + 2, y_ptr + y_stride, linesize, s, 0) +
  412. dv_init_enc_block(enc_blk + 3, y_ptr + 8 + y_stride, linesize, s, 0);
  413. }
  414. enc_blk += 4;
  415. /* initializing chrominance blocks */
  416. c_offset = (((mb_y >> (s->sys->pix_fmt == AV_PIX_FMT_YUV420P)) * s->frame->linesize[1] +
  417. (mb_x >> ((s->sys->pix_fmt == AV_PIX_FMT_YUV411P) ? 2 : 1))) << 3);
  418. for (j = 2; j; j--) {
  419. uint8_t *c_ptr = s->frame->data[j] + c_offset;
  420. linesize = s->frame->linesize[j];
  421. y_stride = (mb_y == 134) ? 8 : (s->frame->linesize[j] << 3);
  422. if (s->sys->pix_fmt == AV_PIX_FMT_YUV411P && mb_x >= (704 / 8)) {
  423. uint8_t *d;
  424. uint8_t *b = scratch;
  425. for (i = 0; i < 8; i++) {
  426. d = c_ptr + (linesize << 3);
  427. b[0] = c_ptr[0];
  428. b[1] = c_ptr[1];
  429. b[2] = c_ptr[2];
  430. b[3] = c_ptr[3];
  431. b[4] = d[0];
  432. b[5] = d[1];
  433. b[6] = d[2];
  434. b[7] = d[3];
  435. c_ptr += linesize;
  436. b += 16;
  437. }
  438. c_ptr = scratch;
  439. linesize = 16;
  440. }
  441. vs_bit_size += dv_init_enc_block(enc_blk++, c_ptr, linesize, s, 1);
  442. if (s->sys->bpm == 8)
  443. vs_bit_size += dv_init_enc_block(enc_blk++, c_ptr + y_stride,
  444. linesize, s, 1);
  445. }
  446. }
  447. if (vs_total_ac_bits < vs_bit_size)
  448. dv_guess_qnos(&enc_blks[0], qnosp);
  449. /* DIF encoding process */
  450. for (j = 0; j < 5 * s->sys->bpm;) {
  451. int start_mb = j;
  452. dif[3] = *qnosp++;
  453. dif += 4;
  454. /* First pass over individual cells only */
  455. for (i = 0; i < s->sys->bpm; i++, j++) {
  456. int sz = s->sys->block_sizes[i] >> 3;
  457. init_put_bits(&pbs[j], dif, sz);
  458. put_sbits(&pbs[j], 9, ((enc_blks[j].mb[0] >> 3) - 1024 + 2) >> 2);
  459. put_bits(&pbs[j], 1, enc_blks[j].dct_mode);
  460. put_bits(&pbs[j], 2, enc_blks[j].cno);
  461. dv_encode_ac(&enc_blks[j], &pbs[j], &pbs[j + 1]);
  462. dif += sz;
  463. }
  464. /* Second pass over each MB space */
  465. pb = &pbs[start_mb];
  466. for (i = 0; i < s->sys->bpm; i++)
  467. if (enc_blks[start_mb + i].partial_bit_count)
  468. pb = dv_encode_ac(&enc_blks[start_mb + i], pb,
  469. &pbs[start_mb + s->sys->bpm]);
  470. }
  471. /* Third and final pass over the whole video segment space */
  472. pb = &pbs[0];
  473. for (j = 0; j < 5 * s->sys->bpm; j++) {
  474. if (enc_blks[j].partial_bit_count)
  475. pb = dv_encode_ac(&enc_blks[j], pb, &pbs[s->sys->bpm * 5]);
  476. if (enc_blks[j].partial_bit_count)
  477. av_log(avctx, AV_LOG_ERROR, "ac bitstream overflow\n");
  478. }
  479. for (j = 0; j < 5 * s->sys->bpm; j++) {
  480. int pos;
  481. int size = pbs[j].size_in_bits >> 3;
  482. flush_put_bits(&pbs[j]);
  483. pos = put_bits_count(&pbs[j]) >> 3;
  484. if (pos > size) {
  485. av_log(avctx, AV_LOG_ERROR,
  486. "bitstream written beyond buffer size\n");
  487. return -1;
  488. }
  489. memset(pbs[j].buf + pos, 0xff, size - pos);
  490. }
  491. return 0;
  492. }
  493. static inline int dv_write_pack(enum dv_pack_type pack_id, DVVideoContext *c,
  494. uint8_t *buf)
  495. {
  496. /*
  497. * Here's what SMPTE314M says about these two:
  498. * (page 6) APTn, AP1n, AP2n, AP3n: These data shall be identical
  499. * as track application IDs (APTn = 001, AP1n =
  500. * 001, AP2n = 001, AP3n = 001), if the source signal
  501. * comes from a digital VCR. If the signal source is
  502. * unknown, all bits for these data shall be set to 1.
  503. * (page 12) STYPE: STYPE defines a signal type of video signal
  504. * 00000b = 4:1:1 compression
  505. * 00100b = 4:2:2 compression
  506. * XXXXXX = Reserved
  507. * Now, I've got two problems with these statements:
  508. * 1. it looks like APT == 111b should be a safe bet, but it isn't.
  509. * It seems that for PAL as defined in IEC 61834 we have to set
  510. * APT to 000 and for SMPTE314M to 001.
  511. * 2. It is not at all clear what STYPE is used for 4:2:0 PAL
  512. * compression scheme (if any).
  513. */
  514. int apt = (c->sys->pix_fmt == AV_PIX_FMT_YUV420P ? 0 : 1);
  515. uint8_t aspect = 0;
  516. if ((int) (av_q2d(c->avctx->sample_aspect_ratio) *
  517. c->avctx->width / c->avctx->height * 10) >= 17) /* 16:9 */
  518. aspect = 0x02;
  519. buf[0] = (uint8_t) pack_id;
  520. switch (pack_id) {
  521. case dv_header525: /* I can't imagine why these two weren't defined as real */
  522. case dv_header625: /* packs in SMPTE314M -- they definitely look like ones */
  523. buf[1] = 0xf8 | /* reserved -- always 1 */
  524. (apt & 0x07); /* APT: Track application ID */
  525. buf[2] = (0 << 7) | /* TF1: audio data is 0 - valid; 1 - invalid */
  526. (0x0f << 3) | /* reserved -- always 1 */
  527. (apt & 0x07); /* AP1: Audio application ID */
  528. buf[3] = (0 << 7) | /* TF2: video data is 0 - valid; 1 - invalid */
  529. (0x0f << 3) | /* reserved -- always 1 */
  530. (apt & 0x07); /* AP2: Video application ID */
  531. buf[4] = (0 << 7) | /* TF3: subcode(SSYB) is 0 - valid; 1 - invalid */
  532. (0x0f << 3) | /* reserved -- always 1 */
  533. (apt & 0x07); /* AP3: Subcode application ID */
  534. break;
  535. case dv_video_source:
  536. buf[1] = 0xff; /* reserved -- always 1 */
  537. buf[2] = (1 << 7) | /* B/W: 0 - b/w, 1 - color */
  538. (1 << 6) | /* following CLF is valid - 0, invalid - 1 */
  539. (3 << 4) | /* CLF: color frames ID (see ITU-R BT.470-4) */
  540. 0xf; /* reserved -- always 1 */
  541. buf[3] = (3 << 6) | /* reserved -- always 1 */
  542. (c->sys->dsf << 5) | /* system: 60fields/50fields */
  543. c->sys->video_stype; /* signal type video compression */
  544. buf[4] = 0xff; /* VISC: 0xff -- no information */
  545. break;
  546. case dv_video_control:
  547. buf[1] = (0 << 6) | /* Copy generation management (CGMS) 0 -- free */
  548. 0x3f; /* reserved -- always 1 */
  549. buf[2] = 0xc8 | /* reserved -- always b11001xxx */
  550. aspect;
  551. buf[3] = (1 << 7) | /* frame/field flag 1 -- frame, 0 -- field */
  552. (1 << 6) | /* first/second field flag 0 -- field 2, 1 -- field 1 */
  553. (1 << 5) | /* frame change flag 0 -- same picture as before, 1 -- different */
  554. (1 << 4) | /* 1 - interlaced, 0 - noninterlaced */
  555. 0xc; /* reserved -- always b1100 */
  556. buf[4] = 0xff; /* reserved -- always 1 */
  557. break;
  558. default:
  559. buf[1] =
  560. buf[2] =
  561. buf[3] =
  562. buf[4] = 0xff;
  563. }
  564. return 5;
  565. }
  566. static inline int dv_write_dif_id(enum dv_section_type t, uint8_t chan_num,
  567. uint8_t seq_num, uint8_t dif_num,
  568. uint8_t *buf)
  569. {
  570. buf[0] = (uint8_t) t; /* Section type */
  571. buf[1] = (seq_num << 4) | /* DIF seq number 0-9 for 525/60; 0-11 for 625/50 */
  572. (chan_num << 3) | /* FSC: for 50Mb/s 0 - first channel; 1 - second */
  573. 7; /* reserved -- always 1 */
  574. buf[2] = dif_num; /* DIF block number Video: 0-134, Audio: 0-8 */
  575. return 3;
  576. }
  577. static inline int dv_write_ssyb_id(uint8_t syb_num, uint8_t fr, uint8_t *buf)
  578. {
  579. if (syb_num == 0 || syb_num == 6) {
  580. buf[0] = (fr << 7) | /* FR ID 1 - first half of each channel; 0 - second */
  581. (0 << 4) | /* AP3 (Subcode application ID) */
  582. 0x0f; /* reserved -- always 1 */
  583. } else if (syb_num == 11) {
  584. buf[0] = (fr << 7) | /* FR ID 1 - first half of each channel; 0 - second */
  585. 0x7f; /* reserved -- always 1 */
  586. } else {
  587. buf[0] = (fr << 7) | /* FR ID 1 - first half of each channel; 0 - second */
  588. (0 << 4) | /* APT (Track application ID) */
  589. 0x0f; /* reserved -- always 1 */
  590. }
  591. buf[1] = 0xf0 | /* reserved -- always 1 */
  592. (syb_num & 0x0f); /* SSYB number 0 - 11 */
  593. buf[2] = 0xff; /* reserved -- always 1 */
  594. return 3;
  595. }
  596. static void dv_format_frame(DVVideoContext *c, uint8_t *buf)
  597. {
  598. int chan, i, j, k;
  599. for (chan = 0; chan < c->sys->n_difchan; chan++) {
  600. for (i = 0; i < c->sys->difseg_size; i++) {
  601. memset(buf, 0xff, 80 * 6); /* first 6 DIF blocks are for control data */
  602. /* DV header: 1DIF */
  603. buf += dv_write_dif_id(dv_sect_header, chan, i, 0, buf);
  604. buf += dv_write_pack((c->sys->dsf ? dv_header625 : dv_header525),
  605. c, buf);
  606. buf += 72; /* unused bytes */
  607. /* DV subcode: 2DIFs */
  608. for (j = 0; j < 2; j++) {
  609. buf += dv_write_dif_id(dv_sect_subcode, chan, i, j, buf);
  610. for (k = 0; k < 6; k++)
  611. buf += dv_write_ssyb_id(k, (i < c->sys->difseg_size / 2), buf) + 5;
  612. buf += 29; /* unused bytes */
  613. }
  614. /* DV VAUX: 3DIFS */
  615. for (j = 0; j < 3; j++) {
  616. buf += dv_write_dif_id(dv_sect_vaux, chan, i, j, buf);
  617. buf += dv_write_pack(dv_video_source, c, buf);
  618. buf += dv_write_pack(dv_video_control, c, buf);
  619. buf += 7 * 5;
  620. buf += dv_write_pack(dv_video_source, c, buf);
  621. buf += dv_write_pack(dv_video_control, c, buf);
  622. buf += 4 * 5 + 2; /* unused bytes */
  623. }
  624. /* DV Audio/Video: 135 Video DIFs + 9 Audio DIFs */
  625. for (j = 0; j < 135; j++) {
  626. if (j % 15 == 0) {
  627. memset(buf, 0xff, 80);
  628. buf += dv_write_dif_id(dv_sect_audio, chan, i, j / 15, buf);
  629. buf += 77; /* audio control & shuffled PCM audio */
  630. }
  631. buf += dv_write_dif_id(dv_sect_video, chan, i, j, buf);
  632. buf += 77; /* 1 video macroblock: 1 bytes control
  633. * 4 * 14 bytes Y 8x8 data
  634. * 10 bytes Cr 8x8 data
  635. * 10 bytes Cb 8x8 data */
  636. }
  637. }
  638. }
  639. }
  640. static int dvvideo_encode_frame(AVCodecContext *c, AVPacket *pkt,
  641. const AVFrame *frame, int *got_packet)
  642. {
  643. DVVideoContext *s = c->priv_data;
  644. int ret;
  645. if ((ret = ff_alloc_packet(pkt, s->sys->frame_size)) < 0) {
  646. av_log(c, AV_LOG_ERROR, "Error getting output packet.\n");
  647. return ret;
  648. }
  649. c->pix_fmt = s->sys->pix_fmt;
  650. s->frame = frame;
  651. #if FF_API_CODED_FRAME
  652. FF_DISABLE_DEPRECATION_WARNINGS
  653. c->coded_frame->key_frame = 1;
  654. c->coded_frame->pict_type = AV_PICTURE_TYPE_I;
  655. FF_ENABLE_DEPRECATION_WARNINGS
  656. #endif
  657. s->buf = pkt->data;
  658. c->execute(c, dv_encode_video_segment, s->work_chunks, NULL,
  659. dv_work_pool_size(s->sys), sizeof(DVwork_chunk));
  660. emms_c();
  661. dv_format_frame(s, pkt->data);
  662. pkt->flags |= AV_PKT_FLAG_KEY;
  663. *got_packet = 1;
  664. return 0;
  665. }
  666. AVCodec ff_dvvideo_encoder = {
  667. .name = "dvvideo",
  668. .long_name = NULL_IF_CONFIG_SMALL("DV (Digital Video)"),
  669. .type = AVMEDIA_TYPE_VIDEO,
  670. .id = AV_CODEC_ID_DVVIDEO,
  671. .priv_data_size = sizeof(DVVideoContext),
  672. .init = dvvideo_encode_init,
  673. .encode2 = dvvideo_encode_frame,
  674. .capabilities = AV_CODEC_CAP_SLICE_THREADS,
  675. .pix_fmts = (const enum AVPixelFormat[]) {
  676. AV_PIX_FMT_YUV411P, AV_PIX_FMT_YUV422P,
  677. AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE
  678. },
  679. };