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.

615 lines
25KB

  1. /*
  2. * Copyright (c) 2010, Google, Inc.
  3. *
  4. * This file is part of FFmpeg.
  5. *
  6. * FFmpeg is free software; you can redistribute it and/or
  7. * modify it under the terms of the GNU Lesser General Public
  8. * License as published by the Free Software Foundation; either
  9. * version 2.1 of the License, or (at your option) any later version.
  10. *
  11. * FFmpeg is distributed in the hope that it will be useful,
  12. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  14. * Lesser General Public License for more details.
  15. *
  16. * You should have received a copy of the GNU Lesser General Public
  17. * License along with FFmpeg; if not, write to the Free Software
  18. * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  19. */
  20. /**
  21. * @file
  22. * VP8 encoder support via libvpx
  23. */
  24. #define VPX_DISABLE_CTRL_TYPECHECKS 1
  25. #define VPX_CODEC_DISABLE_COMPAT 1
  26. #include <vpx/vpx_encoder.h>
  27. #include <vpx/vp8cx.h>
  28. #include "avcodec.h"
  29. #include "internal.h"
  30. #include "libavutil/base64.h"
  31. #include "libavutil/mathematics.h"
  32. #include "libavutil/opt.h"
  33. /**
  34. * Portion of struct vpx_codec_cx_pkt from vpx_encoder.h.
  35. * One encoded frame returned from the library.
  36. */
  37. struct FrameListData {
  38. void *buf; /**< compressed data buffer */
  39. size_t sz; /**< length of compressed data */
  40. int64_t pts; /**< time stamp to show frame
  41. (in timebase units) */
  42. unsigned long duration; /**< duration to show frame
  43. (in timebase units) */
  44. uint32_t flags; /**< flags for this frame */
  45. struct FrameListData *next;
  46. };
  47. typedef struct VP8EncoderContext {
  48. AVClass *class;
  49. struct vpx_codec_ctx encoder;
  50. struct vpx_image rawimg;
  51. struct vpx_fixed_buf twopass_stats;
  52. int deadline; //i.e., RT/GOOD/BEST
  53. struct FrameListData *coded_frame_list;
  54. int cpu_used;
  55. /**
  56. * VP8 specific flags, see VP8F_* below.
  57. */
  58. int flags;
  59. #define VP8F_ERROR_RESILIENT 0x00000001 ///< Enable measures appropriate for streaming over lossy links
  60. #define VP8F_AUTO_ALT_REF 0x00000002 ///< Enable automatic alternate reference frame generation
  61. int auto_alt_ref;
  62. int arnr_max_frames;
  63. int arnr_strength;
  64. int arnr_type;
  65. int lag_in_frames;
  66. int error_resilient;
  67. int crf;
  68. int max_intra_rate;
  69. } VP8Context;
  70. /** String mappings for enum vp8e_enc_control_id */
  71. static const char *const ctlidstr[] = {
  72. [VP8E_UPD_ENTROPY] = "VP8E_UPD_ENTROPY",
  73. [VP8E_UPD_REFERENCE] = "VP8E_UPD_REFERENCE",
  74. [VP8E_USE_REFERENCE] = "VP8E_USE_REFERENCE",
  75. [VP8E_SET_ROI_MAP] = "VP8E_SET_ROI_MAP",
  76. [VP8E_SET_ACTIVEMAP] = "VP8E_SET_ACTIVEMAP",
  77. [VP8E_SET_SCALEMODE] = "VP8E_SET_SCALEMODE",
  78. [VP8E_SET_CPUUSED] = "VP8E_SET_CPUUSED",
  79. [VP8E_SET_ENABLEAUTOALTREF] = "VP8E_SET_ENABLEAUTOALTREF",
  80. [VP8E_SET_NOISE_SENSITIVITY] = "VP8E_SET_NOISE_SENSITIVITY",
  81. [VP8E_SET_SHARPNESS] = "VP8E_SET_SHARPNESS",
  82. [VP8E_SET_STATIC_THRESHOLD] = "VP8E_SET_STATIC_THRESHOLD",
  83. [VP8E_SET_TOKEN_PARTITIONS] = "VP8E_SET_TOKEN_PARTITIONS",
  84. [VP8E_GET_LAST_QUANTIZER] = "VP8E_GET_LAST_QUANTIZER",
  85. [VP8E_SET_ARNR_MAXFRAMES] = "VP8E_SET_ARNR_MAXFRAMES",
  86. [VP8E_SET_ARNR_STRENGTH] = "VP8E_SET_ARNR_STRENGTH",
  87. [VP8E_SET_ARNR_TYPE] = "VP8E_SET_ARNR_TYPE",
  88. [VP8E_SET_CQ_LEVEL] = "VP8E_SET_CQ_LEVEL",
  89. [VP8E_SET_MAX_INTRA_BITRATE_PCT] = "VP8E_SET_MAX_INTRA_BITRATE_PCT",
  90. };
  91. static av_cold void log_encoder_error(AVCodecContext *avctx, const char *desc)
  92. {
  93. VP8Context *ctx = avctx->priv_data;
  94. const char *error = vpx_codec_error(&ctx->encoder);
  95. const char *detail = vpx_codec_error_detail(&ctx->encoder);
  96. av_log(avctx, AV_LOG_ERROR, "%s: %s\n", desc, error);
  97. if (detail)
  98. av_log(avctx, AV_LOG_ERROR, " Additional information: %s\n", detail);
  99. }
  100. static av_cold void dump_enc_cfg(AVCodecContext *avctx,
  101. const struct vpx_codec_enc_cfg *cfg)
  102. {
  103. int width = -30;
  104. int level = AV_LOG_DEBUG;
  105. av_log(avctx, level, "vpx_codec_enc_cfg\n");
  106. av_log(avctx, level, "generic settings\n"
  107. " %*s%u\n %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
  108. " %*s{%u/%u}\n %*s%u\n %*s%d\n %*s%u\n",
  109. width, "g_usage:", cfg->g_usage,
  110. width, "g_threads:", cfg->g_threads,
  111. width, "g_profile:", cfg->g_profile,
  112. width, "g_w:", cfg->g_w,
  113. width, "g_h:", cfg->g_h,
  114. width, "g_timebase:", cfg->g_timebase.num, cfg->g_timebase.den,
  115. width, "g_error_resilient:", cfg->g_error_resilient,
  116. width, "g_pass:", cfg->g_pass,
  117. width, "g_lag_in_frames:", cfg->g_lag_in_frames);
  118. av_log(avctx, level, "rate control settings\n"
  119. " %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
  120. " %*s%d\n %*s%p(%zu)\n %*s%u\n",
  121. width, "rc_dropframe_thresh:", cfg->rc_dropframe_thresh,
  122. width, "rc_resize_allowed:", cfg->rc_resize_allowed,
  123. width, "rc_resize_up_thresh:", cfg->rc_resize_up_thresh,
  124. width, "rc_resize_down_thresh:", cfg->rc_resize_down_thresh,
  125. width, "rc_end_usage:", cfg->rc_end_usage,
  126. width, "rc_twopass_stats_in:", cfg->rc_twopass_stats_in.buf, cfg->rc_twopass_stats_in.sz,
  127. width, "rc_target_bitrate:", cfg->rc_target_bitrate);
  128. av_log(avctx, level, "quantizer settings\n"
  129. " %*s%u\n %*s%u\n",
  130. width, "rc_min_quantizer:", cfg->rc_min_quantizer,
  131. width, "rc_max_quantizer:", cfg->rc_max_quantizer);
  132. av_log(avctx, level, "bitrate tolerance\n"
  133. " %*s%u\n %*s%u\n",
  134. width, "rc_undershoot_pct:", cfg->rc_undershoot_pct,
  135. width, "rc_overshoot_pct:", cfg->rc_overshoot_pct);
  136. av_log(avctx, level, "decoder buffer model\n"
  137. " %*s%u\n %*s%u\n %*s%u\n",
  138. width, "rc_buf_sz:", cfg->rc_buf_sz,
  139. width, "rc_buf_initial_sz:", cfg->rc_buf_initial_sz,
  140. width, "rc_buf_optimal_sz:", cfg->rc_buf_optimal_sz);
  141. av_log(avctx, level, "2 pass rate control settings\n"
  142. " %*s%u\n %*s%u\n %*s%u\n",
  143. width, "rc_2pass_vbr_bias_pct:", cfg->rc_2pass_vbr_bias_pct,
  144. width, "rc_2pass_vbr_minsection_pct:", cfg->rc_2pass_vbr_minsection_pct,
  145. width, "rc_2pass_vbr_maxsection_pct:", cfg->rc_2pass_vbr_maxsection_pct);
  146. av_log(avctx, level, "keyframing settings\n"
  147. " %*s%d\n %*s%u\n %*s%u\n",
  148. width, "kf_mode:", cfg->kf_mode,
  149. width, "kf_min_dist:", cfg->kf_min_dist,
  150. width, "kf_max_dist:", cfg->kf_max_dist);
  151. av_log(avctx, level, "\n");
  152. }
  153. static void coded_frame_add(void *list, struct FrameListData *cx_frame)
  154. {
  155. struct FrameListData **p = list;
  156. while (*p != NULL)
  157. p = &(*p)->next;
  158. *p = cx_frame;
  159. cx_frame->next = NULL;
  160. }
  161. static av_cold void free_coded_frame(struct FrameListData *cx_frame)
  162. {
  163. av_freep(&cx_frame->buf);
  164. av_freep(&cx_frame);
  165. }
  166. static av_cold void free_frame_list(struct FrameListData *list)
  167. {
  168. struct FrameListData *p = list;
  169. while (p) {
  170. list = list->next;
  171. free_coded_frame(p);
  172. p = list;
  173. }
  174. }
  175. static av_cold int codecctl_int(AVCodecContext *avctx,
  176. enum vp8e_enc_control_id id, int val)
  177. {
  178. VP8Context *ctx = avctx->priv_data;
  179. char buf[80];
  180. int width = -30;
  181. int res;
  182. snprintf(buf, sizeof(buf), "%s:", ctlidstr[id]);
  183. av_log(avctx, AV_LOG_DEBUG, " %*s%d\n", width, buf, val);
  184. res = vpx_codec_control(&ctx->encoder, id, val);
  185. if (res != VPX_CODEC_OK) {
  186. snprintf(buf, sizeof(buf), "Failed to set %s codec control",
  187. ctlidstr[id]);
  188. log_encoder_error(avctx, buf);
  189. }
  190. return res == VPX_CODEC_OK ? 0 : AVERROR(EINVAL);
  191. }
  192. static av_cold int vp8_free(AVCodecContext *avctx)
  193. {
  194. VP8Context *ctx = avctx->priv_data;
  195. vpx_codec_destroy(&ctx->encoder);
  196. av_freep(&ctx->twopass_stats.buf);
  197. av_freep(&avctx->coded_frame);
  198. av_freep(&avctx->stats_out);
  199. free_frame_list(ctx->coded_frame_list);
  200. return 0;
  201. }
  202. static av_cold int vp8_init(AVCodecContext *avctx)
  203. {
  204. VP8Context *ctx = avctx->priv_data;
  205. const struct vpx_codec_iface *iface = &vpx_codec_vp8_cx_algo;
  206. struct vpx_codec_enc_cfg enccfg;
  207. int res;
  208. av_log(avctx, AV_LOG_INFO, "%s\n", vpx_codec_version_str());
  209. av_log(avctx, AV_LOG_VERBOSE, "%s\n", vpx_codec_build_config());
  210. if ((res = vpx_codec_enc_config_default(iface, &enccfg, 0)) != VPX_CODEC_OK) {
  211. av_log(avctx, AV_LOG_ERROR, "Failed to get config: %s\n",
  212. vpx_codec_err_to_string(res));
  213. return AVERROR(EINVAL);
  214. }
  215. dump_enc_cfg(avctx, &enccfg);
  216. enccfg.g_w = avctx->width;
  217. enccfg.g_h = avctx->height;
  218. enccfg.g_timebase.num = avctx->time_base.num;
  219. enccfg.g_timebase.den = avctx->time_base.den;
  220. enccfg.g_threads = avctx->thread_count;
  221. enccfg.g_lag_in_frames= ctx->lag_in_frames;
  222. if (avctx->flags & CODEC_FLAG_PASS1)
  223. enccfg.g_pass = VPX_RC_FIRST_PASS;
  224. else if (avctx->flags & CODEC_FLAG_PASS2)
  225. enccfg.g_pass = VPX_RC_LAST_PASS;
  226. else
  227. enccfg.g_pass = VPX_RC_ONE_PASS;
  228. if (avctx->rc_min_rate == avctx->rc_max_rate &&
  229. avctx->rc_min_rate == avctx->bit_rate)
  230. enccfg.rc_end_usage = VPX_CBR;
  231. else if (ctx->crf)
  232. enccfg.rc_end_usage = VPX_CQ;
  233. enccfg.rc_target_bitrate = av_rescale_rnd(avctx->bit_rate, 1, 1000,
  234. AV_ROUND_NEAR_INF);
  235. if (avctx->qmin > 0)
  236. enccfg.rc_min_quantizer = avctx->qmin;
  237. if (avctx->qmax > 0)
  238. enccfg.rc_max_quantizer = avctx->qmax;
  239. enccfg.rc_dropframe_thresh = avctx->frame_skip_threshold;
  240. //0-100 (0 => CBR, 100 => VBR)
  241. enccfg.rc_2pass_vbr_bias_pct = round(avctx->qcompress * 100);
  242. enccfg.rc_2pass_vbr_minsection_pct =
  243. avctx->rc_min_rate * 100LL / avctx->bit_rate;
  244. if (avctx->rc_max_rate)
  245. enccfg.rc_2pass_vbr_maxsection_pct =
  246. avctx->rc_max_rate * 100LL / avctx->bit_rate;
  247. if (avctx->rc_buffer_size)
  248. enccfg.rc_buf_sz =
  249. avctx->rc_buffer_size * 1000LL / avctx->bit_rate;
  250. if (avctx->rc_initial_buffer_occupancy)
  251. enccfg.rc_buf_initial_sz =
  252. avctx->rc_initial_buffer_occupancy * 1000LL / avctx->bit_rate;
  253. enccfg.rc_buf_optimal_sz = enccfg.rc_buf_sz * 5 / 6;
  254. enccfg.rc_undershoot_pct = round(avctx->rc_buffer_aggressivity * 100);
  255. //_enc_init() will balk if kf_min_dist differs from max w/VPX_KF_AUTO
  256. if (avctx->keyint_min >= 0 && avctx->keyint_min == avctx->gop_size)
  257. enccfg.kf_min_dist = avctx->keyint_min;
  258. if (avctx->gop_size >= 0)
  259. enccfg.kf_max_dist = avctx->gop_size;
  260. if (enccfg.g_pass == VPX_RC_FIRST_PASS)
  261. enccfg.g_lag_in_frames = 0;
  262. else if (enccfg.g_pass == VPX_RC_LAST_PASS) {
  263. int decode_size;
  264. if (!avctx->stats_in) {
  265. av_log(avctx, AV_LOG_ERROR, "No stats file for second pass\n");
  266. return AVERROR_INVALIDDATA;
  267. }
  268. ctx->twopass_stats.sz = strlen(avctx->stats_in) * 3 / 4;
  269. ctx->twopass_stats.buf = av_malloc(ctx->twopass_stats.sz);
  270. if (!ctx->twopass_stats.buf) {
  271. av_log(avctx, AV_LOG_ERROR,
  272. "Stat buffer alloc (%zu bytes) failed\n",
  273. ctx->twopass_stats.sz);
  274. return AVERROR(ENOMEM);
  275. }
  276. decode_size = av_base64_decode(ctx->twopass_stats.buf, avctx->stats_in,
  277. ctx->twopass_stats.sz);
  278. if (decode_size < 0) {
  279. av_log(avctx, AV_LOG_ERROR, "Stat buffer decode failed\n");
  280. return AVERROR_INVALIDDATA;
  281. }
  282. ctx->twopass_stats.sz = decode_size;
  283. enccfg.rc_twopass_stats_in = ctx->twopass_stats;
  284. }
  285. /* 0-3: For non-zero values the encoder increasingly optimizes for reduced
  286. complexity playback on low powered devices at the expense of encode
  287. quality. */
  288. if (avctx->profile != FF_PROFILE_UNKNOWN)
  289. enccfg.g_profile = avctx->profile;
  290. enccfg.g_error_resilient = ctx->error_resilient || ctx->flags & VP8F_ERROR_RESILIENT;
  291. dump_enc_cfg(avctx, &enccfg);
  292. /* Construct Encoder Context */
  293. res = vpx_codec_enc_init(&ctx->encoder, iface, &enccfg, 0);
  294. if (res != VPX_CODEC_OK) {
  295. log_encoder_error(avctx, "Failed to initialize encoder");
  296. return AVERROR(EINVAL);
  297. }
  298. //codec control failures are currently treated only as warnings
  299. av_log(avctx, AV_LOG_DEBUG, "vpx_codec_control\n");
  300. if (ctx->cpu_used != INT_MIN)
  301. codecctl_int(avctx, VP8E_SET_CPUUSED, ctx->cpu_used);
  302. if (ctx->flags & VP8F_AUTO_ALT_REF)
  303. ctx->auto_alt_ref = 1;
  304. if (ctx->auto_alt_ref >= 0)
  305. codecctl_int(avctx, VP8E_SET_ENABLEAUTOALTREF, ctx->auto_alt_ref);
  306. if (ctx->arnr_max_frames >= 0)
  307. codecctl_int(avctx, VP8E_SET_ARNR_MAXFRAMES, ctx->arnr_max_frames);
  308. if (ctx->arnr_strength >= 0)
  309. codecctl_int(avctx, VP8E_SET_ARNR_STRENGTH, ctx->arnr_strength);
  310. if (ctx->arnr_type >= 0)
  311. codecctl_int(avctx, VP8E_SET_ARNR_TYPE, ctx->arnr_type);
  312. codecctl_int(avctx, VP8E_SET_NOISE_SENSITIVITY, avctx->noise_reduction);
  313. codecctl_int(avctx, VP8E_SET_TOKEN_PARTITIONS, av_log2(avctx->slices));
  314. codecctl_int(avctx, VP8E_SET_STATIC_THRESHOLD, avctx->mb_threshold);
  315. codecctl_int(avctx, VP8E_SET_CQ_LEVEL, ctx->crf);
  316. if (ctx->max_intra_rate >= 0)
  317. codecctl_int(avctx, VP8E_SET_MAX_INTRA_BITRATE_PCT, ctx->max_intra_rate);
  318. av_log(avctx, AV_LOG_DEBUG, "Using deadline: %d\n", ctx->deadline);
  319. //provide dummy value to initialize wrapper, values will be updated each _encode()
  320. vpx_img_wrap(&ctx->rawimg, VPX_IMG_FMT_I420, avctx->width, avctx->height, 1,
  321. (unsigned char*)1);
  322. avctx->coded_frame = avcodec_alloc_frame();
  323. if (!avctx->coded_frame) {
  324. av_log(avctx, AV_LOG_ERROR, "Error allocating coded frame\n");
  325. vp8_free(avctx);
  326. return AVERROR(ENOMEM);
  327. }
  328. return 0;
  329. }
  330. static inline void cx_pktcpy(struct FrameListData *dst,
  331. const struct vpx_codec_cx_pkt *src)
  332. {
  333. dst->pts = src->data.frame.pts;
  334. dst->duration = src->data.frame.duration;
  335. dst->flags = src->data.frame.flags;
  336. dst->sz = src->data.frame.sz;
  337. dst->buf = src->data.frame.buf;
  338. }
  339. /**
  340. * Store coded frame information in format suitable for return from encode2().
  341. *
  342. * Write information from @a cx_frame to @a pkt
  343. * @return packet data size on success
  344. * @return a negative AVERROR on error
  345. */
  346. static int storeframe(AVCodecContext *avctx, struct FrameListData *cx_frame,
  347. AVPacket *pkt, AVFrame *coded_frame)
  348. {
  349. int ret = ff_alloc_packet2(avctx, pkt, cx_frame->sz);
  350. if (ret >= 0) {
  351. memcpy(pkt->data, cx_frame->buf, pkt->size);
  352. pkt->pts = pkt->dts = cx_frame->pts;
  353. coded_frame->pts = cx_frame->pts;
  354. coded_frame->key_frame = !!(cx_frame->flags & VPX_FRAME_IS_KEY);
  355. if (coded_frame->key_frame) {
  356. coded_frame->pict_type = AV_PICTURE_TYPE_I;
  357. pkt->flags |= AV_PKT_FLAG_KEY;
  358. } else
  359. coded_frame->pict_type = AV_PICTURE_TYPE_P;
  360. } else {
  361. return ret;
  362. }
  363. return pkt->size;
  364. }
  365. /**
  366. * Queue multiple output frames from the encoder, returning the front-most.
  367. * In cases where vpx_codec_get_cx_data() returns more than 1 frame append
  368. * the frame queue. Return the head frame if available.
  369. * @return Stored frame size
  370. * @return AVERROR(EINVAL) on output size error
  371. * @return AVERROR(ENOMEM) on coded frame queue data allocation error
  372. */
  373. static int queue_frames(AVCodecContext *avctx, AVPacket *pkt_out,
  374. AVFrame *coded_frame)
  375. {
  376. VP8Context *ctx = avctx->priv_data;
  377. const struct vpx_codec_cx_pkt *pkt;
  378. const void *iter = NULL;
  379. int size = 0;
  380. if (ctx->coded_frame_list) {
  381. struct FrameListData *cx_frame = ctx->coded_frame_list;
  382. /* return the leading frame if we've already begun queueing */
  383. size = storeframe(avctx, cx_frame, pkt_out, coded_frame);
  384. if (size < 0)
  385. return size;
  386. ctx->coded_frame_list = cx_frame->next;
  387. free_coded_frame(cx_frame);
  388. }
  389. /* consume all available output from the encoder before returning. buffers
  390. are only good through the next vpx_codec call */
  391. while ((pkt = vpx_codec_get_cx_data(&ctx->encoder, &iter))) {
  392. switch (pkt->kind) {
  393. case VPX_CODEC_CX_FRAME_PKT:
  394. if (!size) {
  395. struct FrameListData cx_frame;
  396. /* avoid storing the frame when the list is empty and we haven't yet
  397. provided a frame for output */
  398. assert(!ctx->coded_frame_list);
  399. cx_pktcpy(&cx_frame, pkt);
  400. size = storeframe(avctx, &cx_frame, pkt_out, coded_frame);
  401. if (size < 0)
  402. return size;
  403. } else {
  404. struct FrameListData *cx_frame =
  405. av_malloc(sizeof(struct FrameListData));
  406. if (!cx_frame) {
  407. av_log(avctx, AV_LOG_ERROR,
  408. "Frame queue element alloc failed\n");
  409. return AVERROR(ENOMEM);
  410. }
  411. cx_pktcpy(cx_frame, pkt);
  412. cx_frame->buf = av_malloc(cx_frame->sz);
  413. if (!cx_frame->buf) {
  414. av_log(avctx, AV_LOG_ERROR,
  415. "Data buffer alloc (%zu bytes) failed\n",
  416. cx_frame->sz);
  417. return AVERROR(ENOMEM);
  418. }
  419. memcpy(cx_frame->buf, pkt->data.frame.buf, pkt->data.frame.sz);
  420. coded_frame_add(&ctx->coded_frame_list, cx_frame);
  421. }
  422. break;
  423. case VPX_CODEC_STATS_PKT: {
  424. struct vpx_fixed_buf *stats = &ctx->twopass_stats;
  425. stats->buf = av_realloc_f(stats->buf, 1,
  426. stats->sz + pkt->data.twopass_stats.sz);
  427. if (!stats->buf) {
  428. av_log(avctx, AV_LOG_ERROR, "Stat buffer realloc failed\n");
  429. return AVERROR(ENOMEM);
  430. }
  431. memcpy((uint8_t*)stats->buf + stats->sz,
  432. pkt->data.twopass_stats.buf, pkt->data.twopass_stats.sz);
  433. stats->sz += pkt->data.twopass_stats.sz;
  434. break;
  435. }
  436. case VPX_CODEC_PSNR_PKT: //FIXME add support for CODEC_FLAG_PSNR
  437. case VPX_CODEC_CUSTOM_PKT:
  438. //ignore unsupported/unrecognized packet types
  439. break;
  440. }
  441. }
  442. return size;
  443. }
  444. static int vp8_encode(AVCodecContext *avctx, AVPacket *pkt,
  445. const AVFrame *frame, int *got_packet)
  446. {
  447. VP8Context *ctx = avctx->priv_data;
  448. struct vpx_image *rawimg = NULL;
  449. int64_t timestamp = 0;
  450. long flags = 0;
  451. int res, coded_size;
  452. if (frame) {
  453. rawimg = &ctx->rawimg;
  454. rawimg->planes[VPX_PLANE_Y] = frame->data[0];
  455. rawimg->planes[VPX_PLANE_U] = frame->data[1];
  456. rawimg->planes[VPX_PLANE_V] = frame->data[2];
  457. rawimg->stride[VPX_PLANE_Y] = frame->linesize[0];
  458. rawimg->stride[VPX_PLANE_U] = frame->linesize[1];
  459. rawimg->stride[VPX_PLANE_V] = frame->linesize[2];
  460. timestamp = frame->pts;
  461. flags = frame->pict_type == AV_PICTURE_TYPE_I ? VPX_EFLAG_FORCE_KF : 0;
  462. }
  463. res = vpx_codec_encode(&ctx->encoder, rawimg, timestamp,
  464. avctx->ticks_per_frame, flags, ctx->deadline);
  465. if (res != VPX_CODEC_OK) {
  466. log_encoder_error(avctx, "Error encoding frame");
  467. return AVERROR_INVALIDDATA;
  468. }
  469. coded_size = queue_frames(avctx, pkt, avctx->coded_frame);
  470. if (!frame && avctx->flags & CODEC_FLAG_PASS1) {
  471. unsigned int b64_size = AV_BASE64_SIZE(ctx->twopass_stats.sz);
  472. avctx->stats_out = av_malloc(b64_size);
  473. if (!avctx->stats_out) {
  474. av_log(avctx, AV_LOG_ERROR, "Stat buffer alloc (%d bytes) failed\n",
  475. b64_size);
  476. return AVERROR(ENOMEM);
  477. }
  478. av_base64_encode(avctx->stats_out, b64_size, ctx->twopass_stats.buf,
  479. ctx->twopass_stats.sz);
  480. }
  481. *got_packet = !!coded_size;
  482. return 0;
  483. }
  484. #define OFFSET(x) offsetof(VP8Context, x)
  485. #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
  486. static const AVOption options[] = {
  487. { "cpu-used", "Quality/Speed ratio modifier", OFFSET(cpu_used), AV_OPT_TYPE_INT, {INT_MIN}, INT_MIN, INT_MAX, VE},
  488. { "auto-alt-ref", "Enable use of alternate reference "
  489. "frames (2-pass only)", OFFSET(auto_alt_ref), AV_OPT_TYPE_INT, {-1}, -1, 1, VE},
  490. { "lag-in-frames", "Number of frames to look ahead for "
  491. "alternate reference frame selection", OFFSET(lag_in_frames), AV_OPT_TYPE_INT, {-1}, -1, INT_MAX, VE},
  492. { "arnr-maxframes", "altref noise reduction max frame count", OFFSET(arnr_max_frames), AV_OPT_TYPE_INT, {-1}, -1, INT_MAX, VE},
  493. { "arnr-strength", "altref noise reduction filter strength", OFFSET(arnr_strength), AV_OPT_TYPE_INT, {-1}, -1, INT_MAX, VE},
  494. { "arnr-type", "altref noise reduction filter type", OFFSET(arnr_type), AV_OPT_TYPE_INT, {-1}, -1, INT_MAX, VE, "arnr_type"},
  495. { "backward", NULL, 0, AV_OPT_TYPE_CONST, {1}, 0, 0, VE, "arnr_type" },
  496. { "forward", NULL, 0, AV_OPT_TYPE_CONST, {2}, 0, 0, VE, "arnr_type" },
  497. { "centered", NULL, 0, AV_OPT_TYPE_CONST, {3}, 0, 0, VE, "arnr_type" },
  498. { "deadline", "Time to spend encoding, in microseconds.", OFFSET(deadline), AV_OPT_TYPE_INT, {VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, "quality"},
  499. { "best", NULL, 0, AV_OPT_TYPE_CONST, {VPX_DL_BEST_QUALITY}, 0, 0, VE, "quality"},
  500. { "good", NULL, 0, AV_OPT_TYPE_CONST, {VPX_DL_GOOD_QUALITY}, 0, 0, VE, "quality"},
  501. { "realtime", NULL, 0, AV_OPT_TYPE_CONST, {VPX_DL_REALTIME}, 0, 0, VE, "quality"},
  502. { "error-resilient", "Error resilience configuration", OFFSET(error_resilient), AV_OPT_TYPE_FLAGS, {0}, INT_MIN, INT_MAX, VE, "er"},
  503. { "max-intra-rate", "Maximum I-frame bitrate (pct) 0=unlimited", OFFSET(max_intra_rate), AV_OPT_TYPE_INT, {-1}, -1, INT_MAX, VE},
  504. #ifdef VPX_ERROR_RESILIENT_DEFAULT
  505. { "default", "Improve resiliency against losses of whole frames", 0, AV_OPT_TYPE_CONST, {VPX_ERROR_RESILIENT_DEFAULT}, 0, 0, VE, "er"},
  506. { "partitions", "The frame partitions are independently decodable "
  507. "by the bool decoder, meaning that partitions can be decoded even "
  508. "though earlier partitions have been lost. Note that intra predicition"
  509. " is still done over the partition boundary.", 0, AV_OPT_TYPE_CONST, {VPX_ERROR_RESILIENT_PARTITIONS}, 0, 0, VE, "er"},
  510. #endif
  511. {"speed", "", offsetof(VP8Context, cpu_used), AV_OPT_TYPE_INT, {.dbl = 3}, -16, 16, VE},
  512. {"quality", "", offsetof(VP8Context, deadline), AV_OPT_TYPE_INT, {.dbl = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, "quality"},
  513. {"vp8flags", "", offsetof(VP8Context, flags), FF_OPT_TYPE_FLAGS, {.dbl = 0}, 0, UINT_MAX, VE, "flags"},
  514. {"error_resilient", "enable error resilience", 0, FF_OPT_TYPE_CONST, {.dbl = VP8F_ERROR_RESILIENT}, INT_MIN, INT_MAX, VE, "flags"},
  515. {"altref", "enable use of alternate reference frames (VP8/2-pass only)", 0, FF_OPT_TYPE_CONST, {.dbl = VP8F_AUTO_ALT_REF}, INT_MIN, INT_MAX, VE, "flags"},
  516. {"arnr_max_frames", "altref noise reduction max frame count", offsetof(VP8Context, arnr_max_frames), AV_OPT_TYPE_INT, {.dbl = 0}, 0, 15, VE},
  517. {"arnr_strength", "altref noise reduction filter strength", offsetof(VP8Context, arnr_strength), AV_OPT_TYPE_INT, {.dbl = 3}, 0, 6, VE},
  518. {"arnr_type", "altref noise reduction filter type", offsetof(VP8Context, arnr_type), AV_OPT_TYPE_INT, {.dbl = 3}, 1, 3, VE},
  519. {"rc_lookahead", "Number of frames to look ahead for alternate reference frame selection", offsetof(VP8Context, lag_in_frames), AV_OPT_TYPE_INT, {.dbl = 25}, 0, 25, VE},
  520. {"crf", "Select the quality for constant quality mode", offsetof(VP8Context, crf), AV_OPT_TYPE_INT, {.dbl = 0}, 0, 63, VE},
  521. {NULL}
  522. };
  523. static const AVClass class = {
  524. .class_name = "libvpx encoder",
  525. .item_name = av_default_item_name,
  526. .option = options,
  527. .version = LIBAVUTIL_VERSION_INT,
  528. };
  529. static const AVCodecDefault defaults[] = {
  530. { "qmin", "-1" },
  531. { "qmax", "-1" },
  532. { "g", "-1" },
  533. { "keyint_min", "-1" },
  534. { NULL },
  535. };
  536. AVCodec ff_libvpx_encoder = {
  537. .name = "libvpx",
  538. .type = AVMEDIA_TYPE_VIDEO,
  539. .id = CODEC_ID_VP8,
  540. .priv_data_size = sizeof(VP8Context),
  541. .init = vp8_init,
  542. .encode2 = vp8_encode,
  543. .close = vp8_free,
  544. .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
  545. .pix_fmts = (const enum PixelFormat[]){ PIX_FMT_YUV420P, PIX_FMT_NONE },
  546. .long_name = NULL_IF_CONFIG_SMALL("libvpx VP8"),
  547. .priv_class = &class,
  548. .defaults = defaults,
  549. };