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.

470 lines
17KB

  1. /*
  2. * AVS encoding using the xavs library
  3. * Copyright (C) 2010 Amanda, Y.N. Wu <amanda11192003@gmail.com>
  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. #include <stdio.h>
  22. #include <stdlib.h>
  23. #include <string.h>
  24. #include <math.h>
  25. #include <stdint.h>
  26. #include <float.h>
  27. #include <xavs.h>
  28. #include "avcodec.h"
  29. #include "internal.h"
  30. #include "libavutil/internal.h"
  31. #include "libavutil/mem.h"
  32. #include "libavutil/opt.h"
  33. #define END_OF_STREAM 0x001
  34. #define XAVS_PART_I8X8 0x002 /* Analyze i8x8 (requires 8x8 transform) */
  35. #define XAVS_PART_P8X8 0x010 /* Analyze p16x8, p8x16 and p8x8 */
  36. #define XAVS_PART_B8X8 0x100 /* Analyze b16x8, b*/
  37. typedef struct XavsContext {
  38. AVClass *class;
  39. xavs_param_t params;
  40. xavs_t *enc;
  41. xavs_picture_t pic;
  42. uint8_t *sei;
  43. int sei_size;
  44. int end_of_stream;
  45. float crf;
  46. int cqp;
  47. int b_bias;
  48. float cplxblur;
  49. int direct_pred;
  50. int aud;
  51. int fast_pskip;
  52. int motion_est;
  53. int mbtree;
  54. int mixed_refs;
  55. int64_t *pts_buffer;
  56. int out_frame_count;
  57. } XavsContext;
  58. static void XAVS_log(void *p, int level, const char *fmt, va_list args)
  59. {
  60. static const int level_map[] = {
  61. [XAVS_LOG_ERROR] = AV_LOG_ERROR,
  62. [XAVS_LOG_WARNING] = AV_LOG_WARNING,
  63. [XAVS_LOG_INFO] = AV_LOG_INFO,
  64. [XAVS_LOG_DEBUG] = AV_LOG_DEBUG
  65. };
  66. if (level < 0 || level > XAVS_LOG_DEBUG)
  67. return;
  68. av_vlog(p, level_map[level], fmt, args);
  69. }
  70. static int encode_nals(AVCodecContext *ctx, AVPacket *pkt,
  71. xavs_nal_t *nals, int nnal)
  72. {
  73. XavsContext *x4 = ctx->priv_data;
  74. uint8_t *p;
  75. int i, s, ret, size = x4->sei_size + AV_INPUT_BUFFER_MIN_SIZE;
  76. if (!nnal)
  77. return 0;
  78. for (i = 0; i < nnal; i++)
  79. size += nals[i].i_payload;
  80. if ((ret = ff_alloc_packet2(ctx, pkt, size, 0)) < 0)
  81. return ret;
  82. p = pkt->data;
  83. /* Write the SEI as part of the first frame. */
  84. if (x4->sei_size > 0 && nnal > 0) {
  85. memcpy(p, x4->sei, x4->sei_size);
  86. p += x4->sei_size;
  87. x4->sei_size = 0;
  88. }
  89. for (i = 0; i < nnal; i++) {
  90. s = xavs_nal_encode(p, &size, 1, nals + i);
  91. if (s < 0)
  92. return -1;
  93. p += s;
  94. }
  95. pkt->size = p - pkt->data;
  96. return 1;
  97. }
  98. static int XAVS_frame(AVCodecContext *avctx, AVPacket *pkt,
  99. const AVFrame *frame, int *got_packet)
  100. {
  101. XavsContext *x4 = avctx->priv_data;
  102. xavs_nal_t *nal;
  103. int nnal, i, ret;
  104. xavs_picture_t pic_out;
  105. int pict_type;
  106. x4->pic.img.i_csp = XAVS_CSP_I420;
  107. x4->pic.img.i_plane = 3;
  108. if (frame) {
  109. for (i = 0; i < 3; i++) {
  110. x4->pic.img.plane[i] = frame->data[i];
  111. x4->pic.img.i_stride[i] = frame->linesize[i];
  112. }
  113. x4->pic.i_pts = frame->pts;
  114. x4->pic.i_type = XAVS_TYPE_AUTO;
  115. x4->pts_buffer[avctx->frame_number % (avctx->max_b_frames+1)] = frame->pts;
  116. }
  117. if (xavs_encoder_encode(x4->enc, &nal, &nnal,
  118. frame? &x4->pic: NULL, &pic_out) < 0)
  119. return -1;
  120. ret = encode_nals(avctx, pkt, nal, nnal);
  121. if (ret < 0)
  122. return -1;
  123. if (!ret) {
  124. if (!frame && !(x4->end_of_stream)) {
  125. if ((ret = ff_alloc_packet2(avctx, pkt, 4, 0)) < 0)
  126. return ret;
  127. pkt->data[0] = 0x0;
  128. pkt->data[1] = 0x0;
  129. pkt->data[2] = 0x01;
  130. pkt->data[3] = 0xb1;
  131. pkt->dts = 2*x4->pts_buffer[(x4->out_frame_count-1)%(avctx->max_b_frames+1)] -
  132. x4->pts_buffer[(x4->out_frame_count-2)%(avctx->max_b_frames+1)];
  133. x4->end_of_stream = END_OF_STREAM;
  134. *got_packet = 1;
  135. }
  136. return 0;
  137. }
  138. #if FF_API_CODED_FRAME
  139. FF_DISABLE_DEPRECATION_WARNINGS
  140. avctx->coded_frame->pts = pic_out.i_pts;
  141. FF_ENABLE_DEPRECATION_WARNINGS
  142. #endif
  143. pkt->pts = pic_out.i_pts;
  144. if (avctx->has_b_frames) {
  145. if (!x4->out_frame_count)
  146. pkt->dts = pkt->pts - (x4->pts_buffer[1] - x4->pts_buffer[0]);
  147. else
  148. pkt->dts = x4->pts_buffer[(x4->out_frame_count-1)%(avctx->max_b_frames+1)];
  149. } else
  150. pkt->dts = pkt->pts;
  151. switch (pic_out.i_type) {
  152. case XAVS_TYPE_IDR:
  153. case XAVS_TYPE_I:
  154. pict_type = AV_PICTURE_TYPE_I;
  155. break;
  156. case XAVS_TYPE_P:
  157. pict_type = AV_PICTURE_TYPE_P;
  158. break;
  159. case XAVS_TYPE_B:
  160. case XAVS_TYPE_BREF:
  161. pict_type = AV_PICTURE_TYPE_B;
  162. break;
  163. default:
  164. pict_type = AV_PICTURE_TYPE_NONE;
  165. }
  166. #if FF_API_CODED_FRAME
  167. FF_DISABLE_DEPRECATION_WARNINGS
  168. avctx->coded_frame->pict_type = pict_type;
  169. FF_ENABLE_DEPRECATION_WARNINGS
  170. #endif
  171. /* There is no IDR frame in AVS JiZhun */
  172. /* Sequence header is used as a flag */
  173. if (pic_out.i_type == XAVS_TYPE_I) {
  174. #if FF_API_CODED_FRAME
  175. FF_DISABLE_DEPRECATION_WARNINGS
  176. avctx->coded_frame->key_frame = 1;
  177. FF_ENABLE_DEPRECATION_WARNINGS
  178. #endif
  179. pkt->flags |= AV_PKT_FLAG_KEY;
  180. }
  181. #if FF_API_CODED_FRAME
  182. FF_DISABLE_DEPRECATION_WARNINGS
  183. avctx->coded_frame->quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
  184. FF_ENABLE_DEPRECATION_WARNINGS
  185. #endif
  186. ff_side_data_set_encoder_stats(pkt, (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA, NULL, 0, pict_type);
  187. x4->out_frame_count++;
  188. *got_packet = ret;
  189. return 0;
  190. }
  191. static av_cold int XAVS_close(AVCodecContext *avctx)
  192. {
  193. XavsContext *x4 = avctx->priv_data;
  194. av_freep(&avctx->extradata);
  195. av_freep(&x4->sei);
  196. av_freep(&x4->pts_buffer);
  197. if (x4->enc)
  198. xavs_encoder_close(x4->enc);
  199. return 0;
  200. }
  201. static av_cold int XAVS_init(AVCodecContext *avctx)
  202. {
  203. XavsContext *x4 = avctx->priv_data;
  204. x4->sei_size = 0;
  205. xavs_param_default(&x4->params);
  206. x4->params.pf_log = XAVS_log;
  207. x4->params.p_log_private = avctx;
  208. x4->params.i_keyint_max = avctx->gop_size;
  209. if (avctx->bit_rate) {
  210. x4->params.rc.i_bitrate = avctx->bit_rate / 1000;
  211. x4->params.rc.i_rc_method = XAVS_RC_ABR;
  212. }
  213. x4->params.rc.i_vbv_buffer_size = avctx->rc_buffer_size / 1000;
  214. x4->params.rc.i_vbv_max_bitrate = avctx->rc_max_rate / 1000;
  215. x4->params.rc.b_stat_write = avctx->flags & AV_CODEC_FLAG_PASS1;
  216. if (avctx->flags & AV_CODEC_FLAG_PASS2) {
  217. x4->params.rc.b_stat_read = 1;
  218. } else {
  219. if (x4->crf >= 0) {
  220. x4->params.rc.i_rc_method = XAVS_RC_CRF;
  221. x4->params.rc.f_rf_constant = x4->crf;
  222. } else if (x4->cqp >= 0) {
  223. x4->params.rc.i_rc_method = XAVS_RC_CQP;
  224. x4->params.rc.i_qp_constant = x4->cqp;
  225. }
  226. }
  227. if (x4->aud >= 0)
  228. x4->params.b_aud = x4->aud;
  229. if (x4->mbtree >= 0)
  230. x4->params.rc.b_mb_tree = x4->mbtree;
  231. if (x4->direct_pred >= 0)
  232. x4->params.analyse.i_direct_mv_pred = x4->direct_pred;
  233. if (x4->fast_pskip >= 0)
  234. x4->params.analyse.b_fast_pskip = x4->fast_pskip;
  235. if (x4->motion_est >= 0)
  236. x4->params.analyse.i_me_method = x4->motion_est;
  237. if (x4->mixed_refs >= 0)
  238. x4->params.analyse.b_mixed_references = x4->mixed_refs;
  239. if (x4->b_bias != INT_MIN)
  240. x4->params.i_bframe_bias = x4->b_bias;
  241. if (x4->cplxblur >= 0)
  242. x4->params.rc.f_complexity_blur = x4->cplxblur;
  243. #if FF_API_MOTION_EST
  244. FF_DISABLE_DEPRECATION_WARNINGS
  245. if (x4->motion_est < 0) {
  246. switch (avctx->me_method) {
  247. case ME_EPZS:
  248. x4->params.analyse.i_me_method = XAVS_ME_DIA;
  249. break;
  250. case ME_HEX:
  251. x4->params.analyse.i_me_method = XAVS_ME_HEX;
  252. break;
  253. case ME_UMH:
  254. x4->params.analyse.i_me_method = XAVS_ME_UMH;
  255. break;
  256. case ME_FULL:
  257. x4->params.analyse.i_me_method = XAVS_ME_ESA;
  258. break;
  259. case ME_TESA:
  260. x4->params.analyse.i_me_method = XAVS_ME_TESA;
  261. break;
  262. default:
  263. x4->params.analyse.i_me_method = XAVS_ME_HEX;
  264. }
  265. }
  266. FF_ENABLE_DEPRECATION_WARNINGS
  267. #endif
  268. x4->params.i_bframe = avctx->max_b_frames;
  269. /* cabac is not included in AVS JiZhun Profile */
  270. x4->params.b_cabac = 0;
  271. x4->params.i_bframe_adaptive = avctx->b_frame_strategy;
  272. avctx->has_b_frames = !!avctx->max_b_frames;
  273. /* AVS doesn't allow B picture as reference */
  274. /* The max allowed reference frame number of B is 2 */
  275. x4->params.i_keyint_min = avctx->keyint_min;
  276. if (x4->params.i_keyint_min > x4->params.i_keyint_max)
  277. x4->params.i_keyint_min = x4->params.i_keyint_max;
  278. x4->params.i_scenecut_threshold = avctx->scenechange_threshold;
  279. // x4->params.b_deblocking_filter = avctx->flags & AV_CODEC_FLAG_LOOP_FILTER;
  280. x4->params.rc.i_qp_min = avctx->qmin;
  281. x4->params.rc.i_qp_max = avctx->qmax;
  282. x4->params.rc.i_qp_step = avctx->max_qdiff;
  283. x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */
  284. x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */
  285. x4->params.i_frame_reference = avctx->refs;
  286. x4->params.i_width = avctx->width;
  287. x4->params.i_height = avctx->height;
  288. x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num;
  289. x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den;
  290. /* This is only used for counting the fps */
  291. x4->params.i_fps_num = avctx->time_base.den;
  292. x4->params.i_fps_den = avctx->time_base.num;
  293. x4->params.analyse.inter = XAVS_ANALYSE_I8x8 |XAVS_ANALYSE_PSUB16x16| XAVS_ANALYSE_BSUB16x16;
  294. x4->params.analyse.i_me_range = avctx->me_range;
  295. x4->params.analyse.i_subpel_refine = avctx->me_subpel_quality;
  296. x4->params.analyse.b_chroma_me = avctx->me_cmp & FF_CMP_CHROMA;
  297. /* AVS P2 only enables 8x8 transform */
  298. x4->params.analyse.b_transform_8x8 = 1; //avctx->flags2 & AV_CODEC_FLAG2_8X8DCT;
  299. x4->params.analyse.i_trellis = avctx->trellis;
  300. x4->params.analyse.i_noise_reduction = avctx->noise_reduction;
  301. if (avctx->level > 0)
  302. x4->params.i_level_idc = avctx->level;
  303. if (avctx->bit_rate > 0)
  304. x4->params.rc.f_rate_tolerance =
  305. (float)avctx->bit_rate_tolerance/avctx->bit_rate;
  306. if ((avctx->rc_buffer_size) &&
  307. (avctx->rc_initial_buffer_occupancy <= avctx->rc_buffer_size)) {
  308. x4->params.rc.f_vbv_buffer_init =
  309. (float)avctx->rc_initial_buffer_occupancy / avctx->rc_buffer_size;
  310. } else
  311. x4->params.rc.f_vbv_buffer_init = 0.9;
  312. /* TAG:do we have MB tree RC method */
  313. /* what is the RC method we are now using? Default NO */
  314. x4->params.rc.f_ip_factor = 1 / fabs(avctx->i_quant_factor);
  315. x4->params.rc.f_pb_factor = avctx->b_quant_factor;
  316. x4->params.analyse.i_chroma_qp_offset = avctx->chromaoffset;
  317. x4->params.analyse.b_psnr = avctx->flags & AV_CODEC_FLAG_PSNR;
  318. x4->params.i_log_level = XAVS_LOG_DEBUG;
  319. x4->params.i_threads = avctx->thread_count;
  320. x4->params.b_interlaced = avctx->flags & AV_CODEC_FLAG_INTERLACED_DCT;
  321. if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)
  322. x4->params.b_repeat_headers = 0;
  323. x4->enc = xavs_encoder_open(&x4->params);
  324. if (!x4->enc)
  325. return -1;
  326. if (!(x4->pts_buffer = av_mallocz_array((avctx->max_b_frames+1), sizeof(*x4->pts_buffer))))
  327. return AVERROR(ENOMEM);
  328. /* TAG: Do we have GLOBAL HEADER in AVS */
  329. /* We Have PPS and SPS in AVS */
  330. if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER && 0) {
  331. xavs_nal_t *nal;
  332. int nnal, s, i, size;
  333. uint8_t *p;
  334. s = xavs_encoder_headers(x4->enc, &nal, &nnal);
  335. avctx->extradata = p = av_malloc(s);
  336. for (i = 0; i < nnal; i++) {
  337. /* Don't put the SEI in extradata. */
  338. if (nal[i].i_type == NAL_SEI) {
  339. x4->sei = av_malloc( 5 + nal[i].i_payload * 4 / 3 );
  340. if (xavs_nal_encode(x4->sei, &x4->sei_size, 1, nal + i) < 0)
  341. return -1;
  342. continue;
  343. }
  344. size = xavs_nal_encode(p, &s, 1, nal + i);
  345. if (size < 0)
  346. return -1;
  347. p += size;
  348. }
  349. avctx->extradata_size = p - avctx->extradata;
  350. }
  351. return 0;
  352. }
  353. #define OFFSET(x) offsetof(XavsContext, x)
  354. #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
  355. static const AVOption options[] = {
  356. { "crf", "Select the quality for constant quality mode", OFFSET(crf), AV_OPT_TYPE_FLOAT, {.dbl = -1 }, -1, FLT_MAX, VE },
  357. { "qp", "Constant quantization parameter rate control method",OFFSET(cqp), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE },
  358. { "b-bias", "Influences how often B-frames are used", OFFSET(b_bias), AV_OPT_TYPE_INT, {.i64 = INT_MIN}, INT_MIN, INT_MAX, VE },
  359. { "cplxblur", "Reduce fluctuations in QP (before curve compression)", OFFSET(cplxblur), AV_OPT_TYPE_FLOAT, {.dbl = -1 }, -1, FLT_MAX, VE},
  360. { "direct-pred", "Direct MV prediction mode", OFFSET(direct_pred), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE, "direct-pred" },
  361. { "none", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_NONE }, 0, 0, VE, "direct-pred" },
  362. { "spatial", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_SPATIAL }, 0, 0, VE, "direct-pred" },
  363. { "temporal", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_TEMPORAL }, 0, 0, VE, "direct-pred" },
  364. { "auto", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_AUTO }, 0, 0, VE, "direct-pred" },
  365. { "aud", "Use access unit delimiters.", OFFSET(aud), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 1, VE},
  366. { "mbtree", "Use macroblock tree ratecontrol.", OFFSET(mbtree), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 1, VE},
  367. { "mixed-refs", "One reference per partition, as opposed to one reference per macroblock", OFFSET(mixed_refs), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE },
  368. { "fast-pskip", NULL, OFFSET(fast_pskip), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 1, VE},
  369. { "motion-est", "Set motion estimation method", OFFSET(motion_est), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, XAVS_ME_TESA, VE, "motion-est"},
  370. { "dia", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_DIA }, INT_MIN, INT_MAX, VE, "motion-est" },
  371. { "hex", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_HEX }, INT_MIN, INT_MAX, VE, "motion-est" },
  372. { "umh", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_UMH }, INT_MIN, INT_MAX, VE, "motion-est" },
  373. { "esa", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_ESA }, INT_MIN, INT_MAX, VE, "motion-est" },
  374. { "tesa", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_TESA }, INT_MIN, INT_MAX, VE, "motion-est" },
  375. { NULL },
  376. };
  377. static const AVClass xavs_class = {
  378. .class_name = "libxavs",
  379. .item_name = av_default_item_name,
  380. .option = options,
  381. .version = LIBAVUTIL_VERSION_INT,
  382. };
  383. static const AVCodecDefault xavs_defaults[] = {
  384. { "b", "0" },
  385. { NULL },
  386. };
  387. AVCodec ff_libxavs_encoder = {
  388. .name = "libxavs",
  389. .long_name = NULL_IF_CONFIG_SMALL("libxavs Chinese AVS (Audio Video Standard)"),
  390. .type = AVMEDIA_TYPE_VIDEO,
  391. .id = AV_CODEC_ID_CAVS,
  392. .priv_data_size = sizeof(XavsContext),
  393. .init = XAVS_init,
  394. .encode2 = XAVS_frame,
  395. .close = XAVS_close,
  396. .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS,
  397. .pix_fmts = (const enum AVPixelFormat[]) { AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE },
  398. .priv_class = &xavs_class,
  399. .defaults = xavs_defaults,
  400. };