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.

472 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 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. #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_packet(pkt, size)) < 0) {
  81. av_log(ctx, AV_LOG_ERROR, "Error getting output packet of size %d.\n", size);
  82. return ret;
  83. }
  84. p = pkt->data;
  85. /* Write the SEI as part of the first frame. */
  86. if (x4->sei_size > 0 && nnal > 0) {
  87. memcpy(p, x4->sei, x4->sei_size);
  88. p += x4->sei_size;
  89. x4->sei_size = 0;
  90. }
  91. for (i = 0; i < nnal; i++) {
  92. s = xavs_nal_encode(p, &size, 1, nals + i);
  93. if (s < 0)
  94. return -1;
  95. p += s;
  96. }
  97. pkt->size = p - pkt->data;
  98. return 1;
  99. }
  100. static int XAVS_frame(AVCodecContext *avctx, AVPacket *pkt,
  101. const AVFrame *frame, int *got_packet)
  102. {
  103. XavsContext *x4 = avctx->priv_data;
  104. xavs_nal_t *nal;
  105. int nnal, i, ret;
  106. xavs_picture_t pic_out;
  107. uint8_t *sd;
  108. x4->pic.img.i_csp = XAVS_CSP_I420;
  109. x4->pic.img.i_plane = 3;
  110. if (frame) {
  111. for (i = 0; i < 3; i++) {
  112. x4->pic.img.plane[i] = frame->data[i];
  113. x4->pic.img.i_stride[i] = frame->linesize[i];
  114. }
  115. x4->pic.i_pts = frame->pts;
  116. x4->pic.i_type = XAVS_TYPE_AUTO;
  117. x4->pts_buffer[avctx->frame_number % (avctx->max_b_frames+1)] = frame->pts;
  118. }
  119. if (xavs_encoder_encode(x4->enc, &nal, &nnal,
  120. frame? &x4->pic: NULL, &pic_out) < 0)
  121. return -1;
  122. ret = encode_nals(avctx, pkt, nal, nnal);
  123. if (ret < 0)
  124. return -1;
  125. if (!ret) {
  126. if (!frame && !(x4->end_of_stream)) {
  127. if ((ret = ff_alloc_packet(pkt, 4)) < 0)
  128. return ret;
  129. pkt->data[0] = 0x0;
  130. pkt->data[1] = 0x0;
  131. pkt->data[2] = 0x01;
  132. pkt->data[3] = 0xb1;
  133. pkt->dts = 2*x4->pts_buffer[(x4->out_frame_count-1)%(avctx->max_b_frames+1)] -
  134. x4->pts_buffer[(x4->out_frame_count-2)%(avctx->max_b_frames+1)];
  135. x4->end_of_stream = END_OF_STREAM;
  136. *got_packet = 1;
  137. }
  138. return 0;
  139. }
  140. #if FF_API_CODED_FRAME
  141. FF_DISABLE_DEPRECATION_WARNINGS
  142. avctx->coded_frame->pts = pic_out.i_pts;
  143. FF_ENABLE_DEPRECATION_WARNINGS
  144. #endif
  145. pkt->pts = pic_out.i_pts;
  146. if (avctx->has_b_frames) {
  147. if (!x4->out_frame_count)
  148. pkt->dts = pkt->pts - (x4->pts_buffer[1] - x4->pts_buffer[0]);
  149. else
  150. pkt->dts = x4->pts_buffer[(x4->out_frame_count-1)%(avctx->max_b_frames+1)];
  151. } else
  152. pkt->dts = pkt->pts;
  153. #if FF_API_CODED_FRAME
  154. FF_DISABLE_DEPRECATION_WARNINGS
  155. switch (pic_out.i_type) {
  156. case XAVS_TYPE_IDR:
  157. case XAVS_TYPE_I:
  158. avctx->coded_frame->pict_type = AV_PICTURE_TYPE_I;
  159. break;
  160. case XAVS_TYPE_P:
  161. avctx->coded_frame->pict_type = AV_PICTURE_TYPE_P;
  162. break;
  163. case XAVS_TYPE_B:
  164. case XAVS_TYPE_BREF:
  165. avctx->coded_frame->pict_type = AV_PICTURE_TYPE_B;
  166. break;
  167. }
  168. FF_ENABLE_DEPRECATION_WARNINGS
  169. #endif
  170. /* There is no IDR frame in AVS JiZhun */
  171. /* Sequence header is used as a flag */
  172. if (pic_out.i_type == XAVS_TYPE_I) {
  173. #if FF_API_CODED_FRAME
  174. FF_DISABLE_DEPRECATION_WARNINGS
  175. avctx->coded_frame->key_frame = 1;
  176. FF_ENABLE_DEPRECATION_WARNINGS
  177. #endif
  178. pkt->flags |= AV_PKT_FLAG_KEY;
  179. }
  180. #if FF_API_CODED_FRAME
  181. FF_DISABLE_DEPRECATION_WARNINGS
  182. avctx->coded_frame->quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
  183. FF_ENABLE_DEPRECATION_WARNINGS
  184. #endif
  185. sd = av_packet_new_side_data(pkt, AV_PKT_DATA_QUALITY_FACTOR, sizeof(int));
  186. if (!sd)
  187. return AVERROR(ENOMEM);
  188. *(int *)sd = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
  189. x4->out_frame_count++;
  190. *got_packet = ret;
  191. return 0;
  192. }
  193. static av_cold int XAVS_close(AVCodecContext *avctx)
  194. {
  195. XavsContext *x4 = avctx->priv_data;
  196. av_freep(&avctx->extradata);
  197. av_free(x4->sei);
  198. av_freep(&x4->pts_buffer);
  199. if (x4->enc)
  200. xavs_encoder_close(x4->enc);
  201. return 0;
  202. }
  203. static av_cold int XAVS_init(AVCodecContext *avctx)
  204. {
  205. XavsContext *x4 = avctx->priv_data;
  206. x4->sei_size = 0;
  207. xavs_param_default(&x4->params);
  208. x4->params.pf_log = XAVS_log;
  209. x4->params.p_log_private = avctx;
  210. x4->params.i_keyint_max = avctx->gop_size;
  211. if (avctx->bit_rate) {
  212. x4->params.rc.i_bitrate = avctx->bit_rate / 1000;
  213. x4->params.rc.i_rc_method = XAVS_RC_ABR;
  214. }
  215. x4->params.rc.i_vbv_buffer_size = avctx->rc_buffer_size / 1000;
  216. x4->params.rc.i_vbv_max_bitrate = avctx->rc_max_rate / 1000;
  217. x4->params.rc.b_stat_write = avctx->flags & AV_CODEC_FLAG_PASS1;
  218. if (avctx->flags & AV_CODEC_FLAG_PASS2) {
  219. x4->params.rc.b_stat_read = 1;
  220. } else {
  221. if (x4->crf >= 0) {
  222. x4->params.rc.i_rc_method = XAVS_RC_CRF;
  223. x4->params.rc.f_rf_constant = x4->crf;
  224. } else if (x4->cqp >= 0) {
  225. x4->params.rc.i_rc_method = XAVS_RC_CQP;
  226. x4->params.rc.i_qp_constant = x4->cqp;
  227. }
  228. }
  229. if (x4->aud >= 0)
  230. x4->params.b_aud = x4->aud;
  231. if (x4->mbtree >= 0)
  232. x4->params.rc.b_mb_tree = x4->mbtree;
  233. if (x4->direct_pred >= 0)
  234. x4->params.analyse.i_direct_mv_pred = x4->direct_pred;
  235. if (x4->fast_pskip >= 0)
  236. x4->params.analyse.b_fast_pskip = x4->fast_pskip;
  237. if (x4->motion_est >= 0)
  238. x4->params.analyse.i_me_method = x4->motion_est;
  239. if (x4->mixed_refs >= 0)
  240. x4->params.analyse.b_mixed_references = x4->mixed_refs;
  241. if (x4->b_bias != INT_MIN)
  242. x4->params.i_bframe_bias = x4->b_bias;
  243. if (x4->cplxblur >= 0)
  244. x4->params.rc.f_complexity_blur = x4->cplxblur;
  245. #if FF_API_MOTION_EST
  246. FF_DISABLE_DEPRECATION_WARNINGS
  247. if (x4->motion_est < 0) {
  248. switch (avctx->me_method) {
  249. case ME_EPZS:
  250. x4->params.analyse.i_me_method = XAVS_ME_DIA;
  251. break;
  252. case ME_HEX:
  253. x4->params.analyse.i_me_method = XAVS_ME_HEX;
  254. break;
  255. case ME_UMH:
  256. x4->params.analyse.i_me_method = XAVS_ME_UMH;
  257. break;
  258. case ME_FULL:
  259. x4->params.analyse.i_me_method = XAVS_ME_ESA;
  260. break;
  261. case ME_TESA:
  262. x4->params.analyse.i_me_method = XAVS_ME_TESA;
  263. break;
  264. default:
  265. x4->params.analyse.i_me_method = XAVS_ME_HEX;
  266. }
  267. }
  268. FF_ENABLE_DEPRECATION_WARNINGS
  269. #endif
  270. x4->params.i_bframe = avctx->max_b_frames;
  271. /* cabac is not included in AVS JiZhun Profile */
  272. x4->params.b_cabac = 0;
  273. x4->params.i_bframe_adaptive = avctx->b_frame_strategy;
  274. avctx->has_b_frames = !!avctx->max_b_frames;
  275. /* AVS doesn't allow B picture as reference */
  276. /* The max allowed reference frame number of B is 2 */
  277. x4->params.i_keyint_min = avctx->keyint_min;
  278. if (x4->params.i_keyint_min > x4->params.i_keyint_max)
  279. x4->params.i_keyint_min = x4->params.i_keyint_max;
  280. x4->params.i_scenecut_threshold = avctx->scenechange_threshold;
  281. // x4->params.b_deblocking_filter = avctx->flags & AV_CODEC_FLAG_LOOP_FILTER;
  282. x4->params.rc.i_qp_min = avctx->qmin;
  283. x4->params.rc.i_qp_max = avctx->qmax;
  284. x4->params.rc.i_qp_step = avctx->max_qdiff;
  285. x4->params.rc.f_qcompress = avctx->qcompress; /* 0.0 => cbr, 1.0 => constant qp */
  286. x4->params.rc.f_qblur = avctx->qblur; /* temporally blur quants */
  287. x4->params.i_frame_reference = avctx->refs;
  288. x4->params.i_width = avctx->width;
  289. x4->params.i_height = avctx->height;
  290. x4->params.vui.i_sar_width = avctx->sample_aspect_ratio.num;
  291. x4->params.vui.i_sar_height = avctx->sample_aspect_ratio.den;
  292. /* This is only used for counting the fps */
  293. x4->params.i_fps_num = avctx->time_base.den;
  294. x4->params.i_fps_den = avctx->time_base.num;
  295. x4->params.analyse.inter = XAVS_ANALYSE_I8x8 |XAVS_ANALYSE_PSUB16x16| XAVS_ANALYSE_BSUB16x16;
  296. x4->params.analyse.i_me_range = avctx->me_range;
  297. x4->params.analyse.i_subpel_refine = avctx->me_subpel_quality;
  298. x4->params.analyse.b_chroma_me = avctx->me_cmp & FF_CMP_CHROMA;
  299. /* AVS P2 only enables 8x8 transform */
  300. x4->params.analyse.b_transform_8x8 = 1; //avctx->flags2 & AV_CODEC_FLAG2_8X8DCT;
  301. x4->params.analyse.i_trellis = avctx->trellis;
  302. x4->params.analyse.i_noise_reduction = avctx->noise_reduction;
  303. if (avctx->level > 0)
  304. x4->params.i_level_idc = avctx->level;
  305. if (avctx->bit_rate > 0)
  306. x4->params.rc.f_rate_tolerance =
  307. (float)avctx->bit_rate_tolerance / avctx->bit_rate;
  308. if ((avctx->rc_buffer_size) &&
  309. (avctx->rc_initial_buffer_occupancy <= avctx->rc_buffer_size)) {
  310. x4->params.rc.f_vbv_buffer_init =
  311. (float)avctx->rc_initial_buffer_occupancy / avctx->rc_buffer_size;
  312. } else
  313. x4->params.rc.f_vbv_buffer_init = 0.9;
  314. /* TAG:do we have MB tree RC method */
  315. /* what is the RC method we are now using? Default NO */
  316. x4->params.rc.f_ip_factor = 1 / fabs(avctx->i_quant_factor);
  317. x4->params.rc.f_pb_factor = avctx->b_quant_factor;
  318. x4->params.analyse.i_chroma_qp_offset = avctx->chromaoffset;
  319. x4->params.analyse.b_psnr = avctx->flags & AV_CODEC_FLAG_PSNR;
  320. x4->params.i_log_level = XAVS_LOG_DEBUG;
  321. x4->params.i_threads = avctx->thread_count;
  322. x4->params.b_interlaced = avctx->flags & AV_CODEC_FLAG_INTERLACED_DCT;
  323. if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER)
  324. x4->params.b_repeat_headers = 0;
  325. x4->enc = xavs_encoder_open(&x4->params);
  326. if (!x4->enc)
  327. return -1;
  328. if (!(x4->pts_buffer = av_mallocz((avctx->max_b_frames+1) * sizeof(*x4->pts_buffer))))
  329. return AVERROR(ENOMEM);
  330. /* TAG: Do we have GLOBAL HEADER in AVS */
  331. /* We Have PPS and SPS in AVS */
  332. if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) {
  333. xavs_nal_t *nal;
  334. int nnal, s, i, size;
  335. uint8_t *p;
  336. s = xavs_encoder_headers(x4->enc, &nal, &nnal);
  337. avctx->extradata = p = av_malloc(s);
  338. for (i = 0; i < nnal; i++) {
  339. /* Don't put the SEI in extradata. */
  340. if (nal[i].i_type == NAL_SEI) {
  341. x4->sei = av_malloc( 5 + nal[i].i_payload * 4 / 3 );
  342. if (xavs_nal_encode(x4->sei, &x4->sei_size, 1, nal + i) < 0)
  343. return -1;
  344. continue;
  345. }
  346. size = xavs_nal_encode(p, &s, 1, nal + i);
  347. if (size < 0)
  348. return -1;
  349. p += size;
  350. }
  351. avctx->extradata_size = p - avctx->extradata;
  352. }
  353. return 0;
  354. }
  355. #define OFFSET(x) offsetof(XavsContext, x)
  356. #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
  357. static const AVOption options[] = {
  358. { "crf", "Select the quality for constant quality mode", OFFSET(crf), AV_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE },
  359. { "qp", "Constant quantization parameter rate control method",OFFSET(cqp), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE },
  360. { "b-bias", "Influences how often B-frames are used", OFFSET(b_bias), AV_OPT_TYPE_INT, {.i64 = INT_MIN}, INT_MIN, INT_MAX, VE },
  361. { "cplxblur", "Reduce fluctuations in QP (before curve compression)", OFFSET(cplxblur), AV_OPT_TYPE_FLOAT, {-1 }, -1, FLT_MAX, VE},
  362. { "direct-pred", "Direct MV prediction mode", OFFSET(direct_pred), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, INT_MAX, VE, "direct-pred" },
  363. { "none", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_NONE }, 0, 0, VE, "direct-pred" },
  364. { "spatial", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_SPATIAL }, 0, 0, VE, "direct-pred" },
  365. { "temporal", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_TEMPORAL }, 0, 0, VE, "direct-pred" },
  366. { "auto", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_DIRECT_PRED_AUTO }, 0, 0, VE, "direct-pred" },
  367. { "aud", "Use access unit delimiters.", OFFSET(aud), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 1, VE},
  368. { "mbtree", "Use macroblock tree ratecontrol.", OFFSET(mbtree), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 1, VE},
  369. { "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 },
  370. { "fast-pskip", NULL, OFFSET(fast_pskip), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 1, VE},
  371. { "motion-est", "Set motion estimation method", OFFSET(motion_est), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, XAVS_ME_TESA, VE, "motion-est"},
  372. { "dia", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_DIA }, INT_MIN, INT_MAX, VE, "motion-est" },
  373. { "hex", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_HEX }, INT_MIN, INT_MAX, VE, "motion-est" },
  374. { "umh", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_UMH }, INT_MIN, INT_MAX, VE, "motion-est" },
  375. { "esa", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_ESA }, INT_MIN, INT_MAX, VE, "motion-est" },
  376. { "tesa", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = XAVS_ME_TESA }, INT_MIN, INT_MAX, VE, "motion-est" },
  377. { NULL },
  378. };
  379. static const AVClass class = {
  380. .class_name = "libxavs",
  381. .item_name = av_default_item_name,
  382. .option = options,
  383. .version = LIBAVUTIL_VERSION_INT,
  384. };
  385. static const AVCodecDefault xavs_defaults[] = {
  386. { "b", "0" },
  387. { NULL },
  388. };
  389. AVCodec ff_libxavs_encoder = {
  390. .name = "libxavs",
  391. .long_name = NULL_IF_CONFIG_SMALL("libxavs Chinese AVS (Audio Video Standard)"),
  392. .type = AVMEDIA_TYPE_VIDEO,
  393. .id = AV_CODEC_ID_CAVS,
  394. .priv_data_size = sizeof(XavsContext),
  395. .init = XAVS_init,
  396. .encode2 = XAVS_frame,
  397. .close = XAVS_close,
  398. .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS,
  399. .pix_fmts = (const enum AVPixelFormat[]) { AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE },
  400. .priv_class = &class,
  401. .defaults = xavs_defaults,
  402. };