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.

682 lines
26KB

  1. /*
  2. * Copyright (c) 2010, Google, Inc.
  3. *
  4. * This file is part of Libav.
  5. *
  6. * Libav 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. * Libav 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 Libav; 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 "libvpx.h"
  31. #include "libavutil/base64.h"
  32. #include "libavutil/common.h"
  33. #include "libavutil/mathematics.h"
  34. #include "libavutil/opt.h"
  35. /**
  36. * Portion of struct vpx_codec_cx_pkt from vpx_encoder.h.
  37. * One encoded frame returned from the library.
  38. */
  39. struct FrameListData {
  40. void *buf; /**< compressed data buffer */
  41. size_t sz; /**< length of compressed data */
  42. int64_t pts; /**< time stamp to show frame
  43. (in timebase units) */
  44. unsigned long duration; /**< duration to show frame
  45. (in timebase units) */
  46. uint32_t flags; /**< flags for this frame */
  47. struct FrameListData *next;
  48. };
  49. typedef struct VP8EncoderContext {
  50. AVClass *class;
  51. struct vpx_codec_ctx encoder;
  52. struct vpx_image rawimg;
  53. struct vpx_fixed_buf twopass_stats;
  54. unsigned long deadline; //i.e., RT/GOOD/BEST
  55. struct FrameListData *coded_frame_list;
  56. int cpu_used;
  57. int auto_alt_ref;
  58. int arnr_max_frames;
  59. int arnr_strength;
  60. int arnr_type;
  61. int lag_in_frames;
  62. int error_resilient;
  63. int crf;
  64. int static_thresh;
  65. } VP8Context;
  66. /** String mappings for enum vp8e_enc_control_id */
  67. static const char *const ctlidstr[] = {
  68. [VP8E_SET_ARNR_MAXFRAMES] = "VP8E_SET_ARNR_MAXFRAMES",
  69. [VP8E_SET_ARNR_STRENGTH] = "VP8E_SET_ARNR_STRENGTH",
  70. [VP8E_SET_ARNR_TYPE] = "VP8E_SET_ARNR_TYPE",
  71. [VP8E_SET_CPUUSED] = "VP8E_SET_CPUUSED",
  72. [VP8E_SET_CQ_LEVEL] = "VP8E_SET_CQ_LEVEL",
  73. [VP8E_SET_ENABLEAUTOALTREF] = "VP8E_SET_ENABLEAUTOALTREF",
  74. [VP8E_SET_NOISE_SENSITIVITY] = "VP8E_SET_NOISE_SENSITIVITY",
  75. [VP8E_SET_STATIC_THRESHOLD] = "VP8E_SET_STATIC_THRESHOLD",
  76. [VP8E_SET_TOKEN_PARTITIONS] = "VP8E_SET_TOKEN_PARTITIONS",
  77. };
  78. static av_cold void log_encoder_error(AVCodecContext *avctx, const char *desc)
  79. {
  80. VP8Context *ctx = avctx->priv_data;
  81. const char *error = vpx_codec_error(&ctx->encoder);
  82. const char *detail = vpx_codec_error_detail(&ctx->encoder);
  83. av_log(avctx, AV_LOG_ERROR, "%s: %s\n", desc, error);
  84. if (detail)
  85. av_log(avctx, AV_LOG_ERROR, " Additional information: %s\n", detail);
  86. }
  87. static av_cold void dump_enc_cfg(AVCodecContext *avctx,
  88. const struct vpx_codec_enc_cfg *cfg)
  89. {
  90. int width = -30;
  91. int level = AV_LOG_DEBUG;
  92. av_log(avctx, level, "vpx_codec_enc_cfg\n");
  93. av_log(avctx, level, "generic settings\n"
  94. " %*s%u\n %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
  95. " %*s{%u/%u}\n %*s%u\n %*s%d\n %*s%u\n",
  96. width, "g_usage:", cfg->g_usage,
  97. width, "g_threads:", cfg->g_threads,
  98. width, "g_profile:", cfg->g_profile,
  99. width, "g_w:", cfg->g_w,
  100. width, "g_h:", cfg->g_h,
  101. width, "g_timebase:", cfg->g_timebase.num, cfg->g_timebase.den,
  102. width, "g_error_resilient:", cfg->g_error_resilient,
  103. width, "g_pass:", cfg->g_pass,
  104. width, "g_lag_in_frames:", cfg->g_lag_in_frames);
  105. av_log(avctx, level, "rate control settings\n"
  106. " %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
  107. " %*s%d\n %*s%p(%zu)\n %*s%u\n",
  108. width, "rc_dropframe_thresh:", cfg->rc_dropframe_thresh,
  109. width, "rc_resize_allowed:", cfg->rc_resize_allowed,
  110. width, "rc_resize_up_thresh:", cfg->rc_resize_up_thresh,
  111. width, "rc_resize_down_thresh:", cfg->rc_resize_down_thresh,
  112. width, "rc_end_usage:", cfg->rc_end_usage,
  113. width, "rc_twopass_stats_in:", cfg->rc_twopass_stats_in.buf, cfg->rc_twopass_stats_in.sz,
  114. width, "rc_target_bitrate:", cfg->rc_target_bitrate);
  115. av_log(avctx, level, "quantizer settings\n"
  116. " %*s%u\n %*s%u\n",
  117. width, "rc_min_quantizer:", cfg->rc_min_quantizer,
  118. width, "rc_max_quantizer:", cfg->rc_max_quantizer);
  119. av_log(avctx, level, "bitrate tolerance\n"
  120. " %*s%u\n %*s%u\n",
  121. width, "rc_undershoot_pct:", cfg->rc_undershoot_pct,
  122. width, "rc_overshoot_pct:", cfg->rc_overshoot_pct);
  123. av_log(avctx, level, "decoder buffer model\n"
  124. " %*s%u\n %*s%u\n %*s%u\n",
  125. width, "rc_buf_sz:", cfg->rc_buf_sz,
  126. width, "rc_buf_initial_sz:", cfg->rc_buf_initial_sz,
  127. width, "rc_buf_optimal_sz:", cfg->rc_buf_optimal_sz);
  128. av_log(avctx, level, "2 pass rate control settings\n"
  129. " %*s%u\n %*s%u\n %*s%u\n",
  130. width, "rc_2pass_vbr_bias_pct:", cfg->rc_2pass_vbr_bias_pct,
  131. width, "rc_2pass_vbr_minsection_pct:", cfg->rc_2pass_vbr_minsection_pct,
  132. width, "rc_2pass_vbr_maxsection_pct:", cfg->rc_2pass_vbr_maxsection_pct);
  133. av_log(avctx, level, "keyframing settings\n"
  134. " %*s%d\n %*s%u\n %*s%u\n",
  135. width, "kf_mode:", cfg->kf_mode,
  136. width, "kf_min_dist:", cfg->kf_min_dist,
  137. width, "kf_max_dist:", cfg->kf_max_dist);
  138. av_log(avctx, level, "\n");
  139. }
  140. static void coded_frame_add(void *list, struct FrameListData *cx_frame)
  141. {
  142. struct FrameListData **p = list;
  143. while (*p)
  144. p = &(*p)->next;
  145. *p = cx_frame;
  146. cx_frame->next = NULL;
  147. }
  148. static av_cold void free_coded_frame(struct FrameListData *cx_frame)
  149. {
  150. av_freep(&cx_frame->buf);
  151. av_freep(&cx_frame);
  152. }
  153. static av_cold void free_frame_list(struct FrameListData *list)
  154. {
  155. struct FrameListData *p = list;
  156. while (p) {
  157. list = list->next;
  158. free_coded_frame(p);
  159. p = list;
  160. }
  161. }
  162. static av_cold int codecctl_int(AVCodecContext *avctx,
  163. enum vp8e_enc_control_id id, int val)
  164. {
  165. VP8Context *ctx = avctx->priv_data;
  166. char buf[80];
  167. int width = -30;
  168. int res;
  169. snprintf(buf, sizeof(buf), "%s:", ctlidstr[id]);
  170. av_log(avctx, AV_LOG_DEBUG, " %*s%d\n", width, buf, val);
  171. res = vpx_codec_control(&ctx->encoder, id, val);
  172. if (res != VPX_CODEC_OK) {
  173. snprintf(buf, sizeof(buf), "Failed to set %s codec control",
  174. ctlidstr[id]);
  175. log_encoder_error(avctx, buf);
  176. }
  177. return res == VPX_CODEC_OK ? 0 : AVERROR(EINVAL);
  178. }
  179. static av_cold int vp8_free(AVCodecContext *avctx)
  180. {
  181. VP8Context *ctx = avctx->priv_data;
  182. vpx_codec_destroy(&ctx->encoder);
  183. av_freep(&ctx->twopass_stats.buf);
  184. av_freep(&avctx->stats_out);
  185. free_frame_list(ctx->coded_frame_list);
  186. return 0;
  187. }
  188. static av_cold int vpx_init(AVCodecContext *avctx,
  189. const struct vpx_codec_iface *iface)
  190. {
  191. VP8Context *ctx = avctx->priv_data;
  192. struct vpx_codec_enc_cfg enccfg = { 0 };
  193. AVCPBProperties *cpb_props;
  194. int res;
  195. av_log(avctx, AV_LOG_INFO, "%s\n", vpx_codec_version_str());
  196. av_log(avctx, AV_LOG_VERBOSE, "%s\n", vpx_codec_build_config());
  197. if ((res = vpx_codec_enc_config_default(iface, &enccfg, 0)) != VPX_CODEC_OK) {
  198. av_log(avctx, AV_LOG_ERROR, "Failed to get config: %s\n",
  199. vpx_codec_err_to_string(res));
  200. return AVERROR(EINVAL);
  201. }
  202. dump_enc_cfg(avctx, &enccfg);
  203. enccfg.g_w = avctx->width;
  204. enccfg.g_h = avctx->height;
  205. enccfg.g_timebase.num = avctx->time_base.num;
  206. enccfg.g_timebase.den = avctx->time_base.den;
  207. enccfg.g_threads = avctx->thread_count;
  208. if (ctx->lag_in_frames >= 0)
  209. enccfg.g_lag_in_frames = ctx->lag_in_frames;
  210. if (avctx->flags & AV_CODEC_FLAG_PASS1)
  211. enccfg.g_pass = VPX_RC_FIRST_PASS;
  212. else if (avctx->flags & AV_CODEC_FLAG_PASS2)
  213. enccfg.g_pass = VPX_RC_LAST_PASS;
  214. else
  215. enccfg.g_pass = VPX_RC_ONE_PASS;
  216. if (!avctx->bit_rate)
  217. avctx->bit_rate = enccfg.rc_target_bitrate * 1000;
  218. else
  219. enccfg.rc_target_bitrate = av_rescale_rnd(avctx->bit_rate, 1, 1000,
  220. AV_ROUND_NEAR_INF);
  221. if (ctx->crf)
  222. enccfg.rc_end_usage = VPX_CQ;
  223. else if (avctx->rc_min_rate == avctx->rc_max_rate &&
  224. avctx->rc_min_rate == avctx->bit_rate)
  225. enccfg.rc_end_usage = VPX_CBR;
  226. if (avctx->qmin > 0)
  227. enccfg.rc_min_quantizer = avctx->qmin;
  228. if (avctx->qmax > 0)
  229. enccfg.rc_max_quantizer = avctx->qmax;
  230. enccfg.rc_dropframe_thresh = avctx->frame_skip_threshold;
  231. //0-100 (0 => CBR, 100 => VBR)
  232. enccfg.rc_2pass_vbr_bias_pct = round(avctx->qcompress * 100);
  233. enccfg.rc_2pass_vbr_minsection_pct =
  234. avctx->rc_min_rate * 100LL / avctx->bit_rate;
  235. if (avctx->rc_max_rate)
  236. enccfg.rc_2pass_vbr_maxsection_pct =
  237. avctx->rc_max_rate * 100LL / avctx->bit_rate;
  238. if (avctx->rc_buffer_size)
  239. enccfg.rc_buf_sz =
  240. avctx->rc_buffer_size * 1000LL / avctx->bit_rate;
  241. if (avctx->rc_initial_buffer_occupancy)
  242. enccfg.rc_buf_initial_sz =
  243. avctx->rc_initial_buffer_occupancy * 1000LL / avctx->bit_rate;
  244. enccfg.rc_buf_optimal_sz = enccfg.rc_buf_sz * 5 / 6;
  245. //_enc_init() will balk if kf_min_dist differs from max w/VPX_KF_AUTO
  246. if (avctx->keyint_min >= 0 && avctx->keyint_min == avctx->gop_size)
  247. enccfg.kf_min_dist = avctx->keyint_min;
  248. if (avctx->gop_size >= 0)
  249. enccfg.kf_max_dist = avctx->gop_size;
  250. if (enccfg.g_pass == VPX_RC_FIRST_PASS)
  251. enccfg.g_lag_in_frames = 0;
  252. else if (enccfg.g_pass == VPX_RC_LAST_PASS) {
  253. int decode_size, ret;
  254. if (!avctx->stats_in) {
  255. av_log(avctx, AV_LOG_ERROR, "No stats file for second pass\n");
  256. return AVERROR_INVALIDDATA;
  257. }
  258. ctx->twopass_stats.sz = strlen(avctx->stats_in) * 3 / 4;
  259. ret = av_reallocp(&ctx->twopass_stats.buf, ctx->twopass_stats.sz);
  260. if (ret < 0) {
  261. av_log(avctx, AV_LOG_ERROR,
  262. "Stat buffer alloc (%zu bytes) failed\n",
  263. ctx->twopass_stats.sz);
  264. return ret;
  265. }
  266. decode_size = av_base64_decode(ctx->twopass_stats.buf, avctx->stats_in,
  267. ctx->twopass_stats.sz);
  268. if (decode_size < 0) {
  269. av_log(avctx, AV_LOG_ERROR, "Stat buffer decode failed\n");
  270. return AVERROR_INVALIDDATA;
  271. }
  272. ctx->twopass_stats.sz = decode_size;
  273. enccfg.rc_twopass_stats_in = ctx->twopass_stats;
  274. }
  275. /* 0-3: For non-zero values the encoder increasingly optimizes for reduced
  276. complexity playback on low powered devices at the expense of encode
  277. quality. */
  278. if (avctx->profile != FF_PROFILE_UNKNOWN)
  279. enccfg.g_profile = avctx->profile;
  280. else if (avctx->pix_fmt == AV_PIX_FMT_YUV420P)
  281. avctx->profile = enccfg.g_profile = FF_PROFILE_VP9_0;
  282. else
  283. avctx->profile = enccfg.g_profile = FF_PROFILE_VP9_1;
  284. enccfg.g_error_resilient = ctx->error_resilient;
  285. dump_enc_cfg(avctx, &enccfg);
  286. /* Construct Encoder Context */
  287. res = vpx_codec_enc_init(&ctx->encoder, iface, &enccfg, 0);
  288. if (res != VPX_CODEC_OK) {
  289. log_encoder_error(avctx, "Failed to initialize encoder");
  290. return AVERROR(EINVAL);
  291. }
  292. //codec control failures are currently treated only as warnings
  293. av_log(avctx, AV_LOG_DEBUG, "vpx_codec_control\n");
  294. if (ctx->cpu_used != INT_MIN)
  295. codecctl_int(avctx, VP8E_SET_CPUUSED, ctx->cpu_used);
  296. if (ctx->auto_alt_ref >= 0)
  297. codecctl_int(avctx, VP8E_SET_ENABLEAUTOALTREF, ctx->auto_alt_ref);
  298. if (ctx->arnr_max_frames >= 0)
  299. codecctl_int(avctx, VP8E_SET_ARNR_MAXFRAMES, ctx->arnr_max_frames);
  300. if (ctx->arnr_strength >= 0)
  301. codecctl_int(avctx, VP8E_SET_ARNR_STRENGTH, ctx->arnr_strength);
  302. if (ctx->arnr_type >= 0)
  303. codecctl_int(avctx, VP8E_SET_ARNR_TYPE, ctx->arnr_type);
  304. if (CONFIG_LIBVPX_VP8_ENCODER && iface == &vpx_codec_vp8_cx_algo) {
  305. codecctl_int(avctx, VP8E_SET_NOISE_SENSITIVITY, avctx->noise_reduction);
  306. codecctl_int(avctx, VP8E_SET_TOKEN_PARTITIONS, av_log2(avctx->slices));
  307. }
  308. #if FF_API_MPV_OPT
  309. FF_DISABLE_DEPRECATION_WARNINGS
  310. if (avctx->mb_threshold) {
  311. av_log(avctx, AV_LOG_WARNING, "The mb_threshold option is deprecated, "
  312. "use the static-thresh private option instead.\n");
  313. ctx->static_thresh = avctx->mb_threshold;
  314. }
  315. FF_ENABLE_DEPRECATION_WARNINGS
  316. #endif
  317. codecctl_int(avctx, VP8E_SET_STATIC_THRESHOLD, ctx->static_thresh);
  318. codecctl_int(avctx, VP8E_SET_CQ_LEVEL, ctx->crf);
  319. //provide dummy value to initialize wrapper, values will be updated each _encode()
  320. vpx_img_wrap(&ctx->rawimg, ff_vpx_pixfmt_to_imgfmt(avctx->pix_fmt),
  321. avctx->width, avctx->height, 1, (unsigned char *)1);
  322. cpb_props = ff_add_cpb_side_data(avctx);
  323. if (!cpb_props)
  324. return AVERROR(ENOMEM);
  325. if (enccfg.rc_end_usage == VPX_CBR ||
  326. enccfg.g_pass != VPX_RC_ONE_PASS) {
  327. cpb_props->max_bitrate = avctx->rc_max_rate;
  328. cpb_props->min_bitrate = avctx->rc_min_rate;
  329. cpb_props->avg_bitrate = avctx->bit_rate;
  330. }
  331. cpb_props->buffer_size = avctx->rc_buffer_size;
  332. return 0;
  333. }
  334. static inline void cx_pktcpy(struct FrameListData *dst,
  335. const struct vpx_codec_cx_pkt *src)
  336. {
  337. dst->pts = src->data.frame.pts;
  338. dst->duration = src->data.frame.duration;
  339. dst->flags = src->data.frame.flags;
  340. dst->sz = src->data.frame.sz;
  341. dst->buf = src->data.frame.buf;
  342. }
  343. /**
  344. * Store coded frame information in format suitable for return from encode2().
  345. *
  346. * Write information from @a cx_frame to @a pkt
  347. * @return packet data size on success
  348. * @return a negative AVERROR on error
  349. */
  350. static int storeframe(AVCodecContext *avctx, struct FrameListData *cx_frame,
  351. AVPacket *pkt)
  352. {
  353. int ret = ff_alloc_packet(pkt, cx_frame->sz);
  354. if (ret >= 0) {
  355. memcpy(pkt->data, cx_frame->buf, pkt->size);
  356. pkt->pts = pkt->dts = cx_frame->pts;
  357. #if FF_API_CODED_FRAME
  358. FF_DISABLE_DEPRECATION_WARNINGS
  359. avctx->coded_frame->pts = cx_frame->pts;
  360. avctx->coded_frame->key_frame = !!(cx_frame->flags & VPX_FRAME_IS_KEY);
  361. FF_ENABLE_DEPRECATION_WARNINGS
  362. #endif
  363. if (!!(cx_frame->flags & VPX_FRAME_IS_KEY)) {
  364. #if FF_API_CODED_FRAME
  365. FF_DISABLE_DEPRECATION_WARNINGS
  366. avctx->coded_frame->pict_type = AV_PICTURE_TYPE_I;
  367. FF_ENABLE_DEPRECATION_WARNINGS
  368. #endif
  369. pkt->flags |= AV_PKT_FLAG_KEY;
  370. } else {
  371. #if FF_API_CODED_FRAME
  372. FF_DISABLE_DEPRECATION_WARNINGS
  373. avctx->coded_frame->pict_type = AV_PICTURE_TYPE_P;
  374. FF_ENABLE_DEPRECATION_WARNINGS
  375. #endif
  376. }
  377. } else {
  378. av_log(avctx, AV_LOG_ERROR,
  379. "Error getting output packet of size %zu.\n", cx_frame->sz);
  380. return ret;
  381. }
  382. return pkt->size;
  383. }
  384. /**
  385. * Queue multiple output frames from the encoder, returning the front-most.
  386. * In cases where vpx_codec_get_cx_data() returns more than 1 frame append
  387. * the frame queue. Return the head frame if available.
  388. * @return Stored frame size
  389. * @return AVERROR(EINVAL) on output size error
  390. * @return AVERROR(ENOMEM) on coded frame queue data allocation error
  391. */
  392. static int queue_frames(AVCodecContext *avctx, AVPacket *pkt_out)
  393. {
  394. VP8Context *ctx = avctx->priv_data;
  395. const struct vpx_codec_cx_pkt *pkt;
  396. const void *iter = NULL;
  397. int size = 0;
  398. if (ctx->coded_frame_list) {
  399. struct FrameListData *cx_frame = ctx->coded_frame_list;
  400. /* return the leading frame if we've already begun queueing */
  401. size = storeframe(avctx, cx_frame, pkt_out);
  402. if (size < 0)
  403. return size;
  404. ctx->coded_frame_list = cx_frame->next;
  405. free_coded_frame(cx_frame);
  406. }
  407. /* consume all available output from the encoder before returning. buffers
  408. are only good through the next vpx_codec call */
  409. while ((pkt = vpx_codec_get_cx_data(&ctx->encoder, &iter))) {
  410. switch (pkt->kind) {
  411. case VPX_CODEC_CX_FRAME_PKT:
  412. if (!size) {
  413. struct FrameListData cx_frame;
  414. /* avoid storing the frame when the list is empty and we haven't yet
  415. provided a frame for output */
  416. assert(!ctx->coded_frame_list);
  417. cx_pktcpy(&cx_frame, pkt);
  418. size = storeframe(avctx, &cx_frame, pkt_out);
  419. if (size < 0)
  420. return size;
  421. } else {
  422. struct FrameListData *cx_frame =
  423. av_malloc(sizeof(struct FrameListData));
  424. if (!cx_frame) {
  425. av_log(avctx, AV_LOG_ERROR,
  426. "Frame queue element alloc failed\n");
  427. return AVERROR(ENOMEM);
  428. }
  429. cx_pktcpy(cx_frame, pkt);
  430. cx_frame->buf = av_malloc(cx_frame->sz);
  431. if (!cx_frame->buf) {
  432. av_log(avctx, AV_LOG_ERROR,
  433. "Data buffer alloc (%zu bytes) failed\n",
  434. cx_frame->sz);
  435. av_freep(&cx_frame);
  436. return AVERROR(ENOMEM);
  437. }
  438. memcpy(cx_frame->buf, pkt->data.frame.buf, pkt->data.frame.sz);
  439. coded_frame_add(&ctx->coded_frame_list, cx_frame);
  440. }
  441. break;
  442. case VPX_CODEC_STATS_PKT: {
  443. struct vpx_fixed_buf *stats = &ctx->twopass_stats;
  444. int err;
  445. if ((err = av_reallocp(&stats->buf,
  446. stats->sz +
  447. pkt->data.twopass_stats.sz)) < 0) {
  448. stats->sz = 0;
  449. av_log(avctx, AV_LOG_ERROR, "Stat buffer realloc failed\n");
  450. return err;
  451. }
  452. memcpy((uint8_t*)stats->buf + stats->sz,
  453. pkt->data.twopass_stats.buf, pkt->data.twopass_stats.sz);
  454. stats->sz += pkt->data.twopass_stats.sz;
  455. break;
  456. }
  457. case VPX_CODEC_PSNR_PKT: //FIXME add support for AV_CODEC_FLAG_PSNR
  458. case VPX_CODEC_CUSTOM_PKT:
  459. //ignore unsupported/unrecognized packet types
  460. break;
  461. }
  462. }
  463. return size;
  464. }
  465. static int vp8_encode(AVCodecContext *avctx, AVPacket *pkt,
  466. const AVFrame *frame, int *got_packet)
  467. {
  468. VP8Context *ctx = avctx->priv_data;
  469. struct vpx_image *rawimg = NULL;
  470. int64_t timestamp = 0;
  471. int res, coded_size;
  472. vpx_enc_frame_flags_t flags = 0;
  473. if (frame) {
  474. rawimg = &ctx->rawimg;
  475. rawimg->planes[VPX_PLANE_Y] = frame->data[0];
  476. rawimg->planes[VPX_PLANE_U] = frame->data[1];
  477. rawimg->planes[VPX_PLANE_V] = frame->data[2];
  478. rawimg->stride[VPX_PLANE_Y] = frame->linesize[0];
  479. rawimg->stride[VPX_PLANE_U] = frame->linesize[1];
  480. rawimg->stride[VPX_PLANE_V] = frame->linesize[2];
  481. timestamp = frame->pts;
  482. if (frame->pict_type == AV_PICTURE_TYPE_I)
  483. flags |= VPX_EFLAG_FORCE_KF;
  484. }
  485. res = vpx_codec_encode(&ctx->encoder, rawimg, timestamp,
  486. avctx->ticks_per_frame, flags, ctx->deadline);
  487. if (res != VPX_CODEC_OK) {
  488. log_encoder_error(avctx, "Error encoding frame");
  489. return AVERROR_INVALIDDATA;
  490. }
  491. coded_size = queue_frames(avctx, pkt);
  492. if (!frame && avctx->flags & AV_CODEC_FLAG_PASS1) {
  493. unsigned int b64_size = AV_BASE64_SIZE(ctx->twopass_stats.sz);
  494. avctx->stats_out = av_malloc(b64_size);
  495. if (!avctx->stats_out) {
  496. av_log(avctx, AV_LOG_ERROR, "Stat buffer alloc (%d bytes) failed\n",
  497. b64_size);
  498. return AVERROR(ENOMEM);
  499. }
  500. av_base64_encode(avctx->stats_out, b64_size, ctx->twopass_stats.buf,
  501. ctx->twopass_stats.sz);
  502. }
  503. *got_packet = !!coded_size;
  504. return 0;
  505. }
  506. #define OFFSET(x) offsetof(VP8Context, x)
  507. #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
  508. static const AVOption options[] = {
  509. { "cpu-used", "Quality/Speed ratio modifier", OFFSET(cpu_used), AV_OPT_TYPE_INT, {.i64 = 1}, INT_MIN, INT_MAX, VE},
  510. { "auto-alt-ref", "Enable use of alternate reference "
  511. "frames (2-pass only)", OFFSET(auto_alt_ref), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE},
  512. { "lag-in-frames", "Number of frames to look ahead for "
  513. "alternate reference frame selection", OFFSET(lag_in_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE},
  514. { "arnr-maxframes", "altref noise reduction max frame count", OFFSET(arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE},
  515. { "arnr-strength", "altref noise reduction filter strength", OFFSET(arnr_strength), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE},
  516. { "arnr-type", "altref noise reduction filter type", OFFSET(arnr_type), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE, "arnr_type"},
  517. { "backward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0, VE, "arnr_type" },
  518. { "forward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0, VE, "arnr_type" },
  519. { "centered", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 3}, 0, 0, VE, "arnr_type" },
  520. { "deadline", "Time to spend encoding, in microseconds.", OFFSET(deadline), AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, "quality"},
  521. { "best", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_BEST_QUALITY}, 0, 0, VE, "quality"},
  522. { "good", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_GOOD_QUALITY}, 0, 0, VE, "quality"},
  523. { "realtime", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_REALTIME}, 0, 0, VE, "quality"},
  524. { "error-resilient", "Error resilience configuration", OFFSET(error_resilient), AV_OPT_TYPE_FLAGS, {.i64 = 0}, INT_MIN, INT_MAX, VE, "er"},
  525. #ifdef VPX_ERROR_RESILIENT_DEFAULT
  526. { "default", "Improve resiliency against losses of whole frames", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_DEFAULT}, 0, 0, VE, "er"},
  527. { "partitions", "The frame partitions are independently decodable "
  528. "by the bool decoder, meaning that partitions can be decoded even "
  529. "though earlier partitions have been lost. Note that intra predicition"
  530. " is still done over the partition boundary.", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_PARTITIONS}, 0, 0, VE, "er"},
  531. #endif
  532. { "crf", "Select the quality for constant quality mode", offsetof(VP8Context, crf), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 63, VE },
  533. { "static-thresh", "A change threshold on blocks below which they will be skipped by the encoder", OFFSET(static_thresh), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, VE },
  534. { NULL }
  535. };
  536. static const AVCodecDefault defaults[] = {
  537. { "qmin", "-1" },
  538. { "qmax", "-1" },
  539. { "g", "-1" },
  540. { "keyint_min", "-1" },
  541. { NULL },
  542. };
  543. #if CONFIG_LIBVPX_VP8_ENCODER
  544. static av_cold int vp8_init(AVCodecContext *avctx)
  545. {
  546. return vpx_init(avctx, &vpx_codec_vp8_cx_algo);
  547. }
  548. static const AVClass class_vp8 = {
  549. .class_name = "libvpx encoder",
  550. .item_name = av_default_item_name,
  551. .option = options,
  552. .version = LIBAVUTIL_VERSION_INT,
  553. };
  554. AVCodec ff_libvpx_vp8_encoder = {
  555. .name = "libvpx",
  556. .long_name = NULL_IF_CONFIG_SMALL("libvpx VP8"),
  557. .type = AVMEDIA_TYPE_VIDEO,
  558. .id = AV_CODEC_ID_VP8,
  559. .priv_data_size = sizeof(VP8Context),
  560. .init = vp8_init,
  561. .encode2 = vp8_encode,
  562. .close = vp8_free,
  563. .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS,
  564. .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE },
  565. .priv_class = &class_vp8,
  566. .defaults = defaults,
  567. };
  568. #endif /* CONFIG_LIBVPX_VP8_ENCODER */
  569. #if CONFIG_LIBVPX_VP9_ENCODER
  570. static av_cold int vp9_init(AVCodecContext *avctx)
  571. {
  572. return vpx_init(avctx, &vpx_codec_vp9_cx_algo);
  573. }
  574. static const AVClass class_vp9 = {
  575. .class_name = "libvpx encoder",
  576. .item_name = av_default_item_name,
  577. .option = options,
  578. .version = LIBAVUTIL_VERSION_INT,
  579. };
  580. static const AVProfile profiles[] = {
  581. { FF_PROFILE_VP9_0, "Profile 0" },
  582. { FF_PROFILE_VP9_1, "Profile 1" },
  583. { FF_PROFILE_VP9_2, "Profile 2" },
  584. { FF_PROFILE_VP9_3, "Profile 3" },
  585. { FF_PROFILE_UNKNOWN },
  586. };
  587. AVCodec ff_libvpx_vp9_encoder = {
  588. .name = "libvpx-vp9",
  589. .long_name = NULL_IF_CONFIG_SMALL("libvpx VP9"),
  590. .type = AVMEDIA_TYPE_VIDEO,
  591. .id = AV_CODEC_ID_VP9,
  592. .priv_data_size = sizeof(VP8Context),
  593. .init = vp9_init,
  594. .encode2 = vp8_encode,
  595. .close = vp8_free,
  596. .capabilities = AV_CODEC_CAP_DELAY | AV_CODEC_CAP_AUTO_THREADS,
  597. .pix_fmts = (const enum AVPixelFormat[]) {
  598. AV_PIX_FMT_YUV420P,
  599. #if VPX_IMAGE_ABI_VERSION >= 3
  600. AV_PIX_FMT_YUV422P,
  601. AV_PIX_FMT_YUV444P,
  602. AV_PIX_FMT_YUV440P,
  603. #endif
  604. AV_PIX_FMT_NONE,
  605. },
  606. .profiles = NULL_IF_CONFIG_SMALL(profiles),
  607. .priv_class = &class_vp9,
  608. .defaults = defaults,
  609. };
  610. #endif /* CONFIG_LIBVPX_VP9_ENCODER */