2 * Copyright (c) 2010, Google, Inc.
4 * This file is part of FFmpeg.
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.
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.
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
23 * VP8 encoder support via libvpx
26 #define VPX_DISABLE_CTRL_TYPECHECKS 1
27 #define VPX_CODEC_DISABLE_COMPAT 1
28 #include <vpx/vpx_encoder.h>
29 #include <vpx/vp8cx.h>
33 #include "libavutil/avassert.h"
35 #include "libavutil/base64.h"
36 #include "libavutil/common.h"
37 #include "libavutil/intreadwrite.h"
38 #include "libavutil/mathematics.h"
39 #include "libavutil/opt.h"
42 * Portion of struct vpx_codec_cx_pkt from vpx_encoder.h.
43 * One encoded frame returned from the library.
45 struct FrameListData {
46 void *buf; /**< compressed data buffer */
47 size_t sz; /**< length of compressed data */
50 int64_t pts; /**< time stamp to show frame
51 (in timebase units) */
52 unsigned long duration; /**< duration to show frame
53 (in timebase units) */
54 uint32_t flags; /**< flags for this frame */
56 int have_sse; /**< true if we have pending sse[] */
57 uint64_t frame_number;
58 struct FrameListData *next;
61 typedef struct VP8EncoderContext {
63 struct vpx_codec_ctx encoder;
64 struct vpx_image rawimg;
65 struct vpx_codec_ctx encoder_alpha;
66 struct vpx_image rawimg_alpha;
68 struct vpx_fixed_buf twopass_stats;
69 int deadline; //i.e., RT/GOOD/BEST
71 int have_sse; /**< true if we have pending sse[] */
72 uint64_t frame_number;
73 struct FrameListData *coded_frame_list;
77 * VP8 specific flags, see VP8F_* below.
80 #define VP8F_ERROR_RESILIENT 0x00000001 ///< Enable measures appropriate for streaming over lossy links
81 #define VP8F_AUTO_ALT_REF 0x00000002 ///< Enable automatic alternate reference frame generation
94 int rc_undershoot_pct;
105 /** String mappings for enum vp8e_enc_control_id */
106 static const char *const ctlidstr[] = {
107 [VP8E_UPD_ENTROPY] = "VP8E_UPD_ENTROPY",
108 [VP8E_UPD_REFERENCE] = "VP8E_UPD_REFERENCE",
109 [VP8E_USE_REFERENCE] = "VP8E_USE_REFERENCE",
110 [VP8E_SET_ROI_MAP] = "VP8E_SET_ROI_MAP",
111 [VP8E_SET_ACTIVEMAP] = "VP8E_SET_ACTIVEMAP",
112 [VP8E_SET_SCALEMODE] = "VP8E_SET_SCALEMODE",
113 [VP8E_SET_CPUUSED] = "VP8E_SET_CPUUSED",
114 [VP8E_SET_ENABLEAUTOALTREF] = "VP8E_SET_ENABLEAUTOALTREF",
115 [VP8E_SET_NOISE_SENSITIVITY] = "VP8E_SET_NOISE_SENSITIVITY",
116 [VP8E_SET_SHARPNESS] = "VP8E_SET_SHARPNESS",
117 [VP8E_SET_STATIC_THRESHOLD] = "VP8E_SET_STATIC_THRESHOLD",
118 [VP8E_SET_TOKEN_PARTITIONS] = "VP8E_SET_TOKEN_PARTITIONS",
119 [VP8E_GET_LAST_QUANTIZER] = "VP8E_GET_LAST_QUANTIZER",
120 [VP8E_SET_ARNR_MAXFRAMES] = "VP8E_SET_ARNR_MAXFRAMES",
121 [VP8E_SET_ARNR_STRENGTH] = "VP8E_SET_ARNR_STRENGTH",
122 [VP8E_SET_ARNR_TYPE] = "VP8E_SET_ARNR_TYPE",
123 [VP8E_SET_CQ_LEVEL] = "VP8E_SET_CQ_LEVEL",
124 [VP8E_SET_MAX_INTRA_BITRATE_PCT] = "VP8E_SET_MAX_INTRA_BITRATE_PCT",
125 #if CONFIG_LIBVPX_VP9_ENCODER
126 [VP9E_SET_LOSSLESS] = "VP9E_SET_LOSSLESS",
127 [VP9E_SET_TILE_COLUMNS] = "VP9E_SET_TILE_COLUMNS",
128 [VP9E_SET_TILE_ROWS] = "VP9E_SET_TILE_ROWS",
129 [VP9E_SET_FRAME_PARALLEL_DECODING] = "VP9E_SET_FRAME_PARALLEL_DECODING",
130 [VP9E_SET_AQ_MODE] = "VP9E_SET_AQ_MODE",
134 static av_cold void log_encoder_error(AVCodecContext *avctx, const char *desc)
136 VP8Context *ctx = avctx->priv_data;
137 const char *error = vpx_codec_error(&ctx->encoder);
138 const char *detail = vpx_codec_error_detail(&ctx->encoder);
140 av_log(avctx, AV_LOG_ERROR, "%s: %s\n", desc, error);
142 av_log(avctx, AV_LOG_ERROR, " Additional information: %s\n", detail);
145 static av_cold void dump_enc_cfg(AVCodecContext *avctx,
146 const struct vpx_codec_enc_cfg *cfg)
149 int level = AV_LOG_DEBUG;
151 av_log(avctx, level, "vpx_codec_enc_cfg\n");
152 av_log(avctx, level, "generic settings\n"
153 " %*s%u\n %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
154 #if CONFIG_LIBVPX_VP9_ENCODER && defined(VPX_IMG_FMT_HIGHBITDEPTH)
157 " %*s{%u/%u}\n %*s%u\n %*s%d\n %*s%u\n",
158 width, "g_usage:", cfg->g_usage,
159 width, "g_threads:", cfg->g_threads,
160 width, "g_profile:", cfg->g_profile,
161 width, "g_w:", cfg->g_w,
162 width, "g_h:", cfg->g_h,
163 #if CONFIG_LIBVPX_VP9_ENCODER && defined(VPX_IMG_FMT_HIGHBITDEPTH)
164 width, "g_bit_depth:", cfg->g_bit_depth,
165 width, "g_input_bit_depth:", cfg->g_input_bit_depth,
167 width, "g_timebase:", cfg->g_timebase.num, cfg->g_timebase.den,
168 width, "g_error_resilient:", cfg->g_error_resilient,
169 width, "g_pass:", cfg->g_pass,
170 width, "g_lag_in_frames:", cfg->g_lag_in_frames);
171 av_log(avctx, level, "rate control settings\n"
172 " %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
173 " %*s%d\n %*s%p(%"SIZE_SPECIFIER")\n %*s%u\n",
174 width, "rc_dropframe_thresh:", cfg->rc_dropframe_thresh,
175 width, "rc_resize_allowed:", cfg->rc_resize_allowed,
176 width, "rc_resize_up_thresh:", cfg->rc_resize_up_thresh,
177 width, "rc_resize_down_thresh:", cfg->rc_resize_down_thresh,
178 width, "rc_end_usage:", cfg->rc_end_usage,
179 width, "rc_twopass_stats_in:", cfg->rc_twopass_stats_in.buf, cfg->rc_twopass_stats_in.sz,
180 width, "rc_target_bitrate:", cfg->rc_target_bitrate);
181 av_log(avctx, level, "quantizer settings\n"
183 width, "rc_min_quantizer:", cfg->rc_min_quantizer,
184 width, "rc_max_quantizer:", cfg->rc_max_quantizer);
185 av_log(avctx, level, "bitrate tolerance\n"
187 width, "rc_undershoot_pct:", cfg->rc_undershoot_pct,
188 width, "rc_overshoot_pct:", cfg->rc_overshoot_pct);
189 av_log(avctx, level, "decoder buffer model\n"
190 " %*s%u\n %*s%u\n %*s%u\n",
191 width, "rc_buf_sz:", cfg->rc_buf_sz,
192 width, "rc_buf_initial_sz:", cfg->rc_buf_initial_sz,
193 width, "rc_buf_optimal_sz:", cfg->rc_buf_optimal_sz);
194 av_log(avctx, level, "2 pass rate control settings\n"
195 " %*s%u\n %*s%u\n %*s%u\n",
196 width, "rc_2pass_vbr_bias_pct:", cfg->rc_2pass_vbr_bias_pct,
197 width, "rc_2pass_vbr_minsection_pct:", cfg->rc_2pass_vbr_minsection_pct,
198 width, "rc_2pass_vbr_maxsection_pct:", cfg->rc_2pass_vbr_maxsection_pct);
199 av_log(avctx, level, "keyframing settings\n"
200 " %*s%d\n %*s%u\n %*s%u\n",
201 width, "kf_mode:", cfg->kf_mode,
202 width, "kf_min_dist:", cfg->kf_min_dist,
203 width, "kf_max_dist:", cfg->kf_max_dist);
204 av_log(avctx, level, "\n");
207 static void coded_frame_add(void *list, struct FrameListData *cx_frame)
209 struct FrameListData **p = list;
214 cx_frame->next = NULL;
217 static av_cold void free_coded_frame(struct FrameListData *cx_frame)
219 av_freep(&cx_frame->buf);
220 if (cx_frame->buf_alpha)
221 av_freep(&cx_frame->buf_alpha);
225 static av_cold void free_frame_list(struct FrameListData *list)
227 struct FrameListData *p = list;
236 static av_cold int codecctl_int(AVCodecContext *avctx,
237 enum vp8e_enc_control_id id, int val)
239 VP8Context *ctx = avctx->priv_data;
244 snprintf(buf, sizeof(buf), "%s:", ctlidstr[id]);
245 av_log(avctx, AV_LOG_DEBUG, " %*s%d\n", width, buf, val);
247 res = vpx_codec_control(&ctx->encoder, id, val);
248 if (res != VPX_CODEC_OK) {
249 snprintf(buf, sizeof(buf), "Failed to set %s codec control",
251 log_encoder_error(avctx, buf);
254 return res == VPX_CODEC_OK ? 0 : AVERROR(EINVAL);
257 static av_cold int vp8_free(AVCodecContext *avctx)
259 VP8Context *ctx = avctx->priv_data;
261 vpx_codec_destroy(&ctx->encoder);
263 vpx_codec_destroy(&ctx->encoder_alpha);
264 av_freep(&ctx->twopass_stats.buf);
265 av_frame_free(&avctx->coded_frame);
266 av_freep(&avctx->stats_out);
267 free_frame_list(ctx->coded_frame_list);
271 #if CONFIG_LIBVPX_VP9_ENCODER
272 static int set_pix_fmt(AVCodecContext *avctx, vpx_codec_caps_t codec_caps,
273 struct vpx_codec_enc_cfg *enccfg, vpx_codec_flags_t *flags,
274 vpx_img_fmt_t *img_fmt)
276 #ifdef VPX_IMG_FMT_HIGHBITDEPTH
277 enccfg->g_bit_depth = enccfg->g_input_bit_depth = 8;
279 switch (avctx->pix_fmt) {
280 case AV_PIX_FMT_YUV420P:
281 enccfg->g_profile = 0;
282 *img_fmt = VPX_IMG_FMT_I420;
284 case AV_PIX_FMT_YUV422P:
285 enccfg->g_profile = 1;
286 *img_fmt = VPX_IMG_FMT_I422;
288 #if VPX_IMAGE_ABI_VERSION >= 3
289 case AV_PIX_FMT_YUV440P:
290 enccfg->g_profile = 1;
291 *img_fmt = VPX_IMG_FMT_I440;
294 case AV_PIX_FMT_YUV444P:
295 enccfg->g_profile = 1;
296 *img_fmt = VPX_IMG_FMT_I444;
298 #ifdef VPX_IMG_FMT_HIGHBITDEPTH
299 case AV_PIX_FMT_YUV420P10LE:
300 case AV_PIX_FMT_YUV420P12LE:
301 if (codec_caps & VPX_CODEC_CAP_HIGHBITDEPTH) {
302 enccfg->g_bit_depth = enccfg->g_input_bit_depth =
303 avctx->pix_fmt == AV_PIX_FMT_YUV420P10LE ? 10 : 12;
304 enccfg->g_profile = 2;
305 *img_fmt = VPX_IMG_FMT_I42016;
306 *flags |= VPX_CODEC_USE_HIGHBITDEPTH;
310 case AV_PIX_FMT_YUV422P10LE:
311 case AV_PIX_FMT_YUV422P12LE:
312 if (codec_caps & VPX_CODEC_CAP_HIGHBITDEPTH) {
313 enccfg->g_bit_depth = enccfg->g_input_bit_depth =
314 avctx->pix_fmt == AV_PIX_FMT_YUV422P10LE ? 10 : 12;
315 enccfg->g_profile = 3;
316 *img_fmt = VPX_IMG_FMT_I42216;
317 *flags |= VPX_CODEC_USE_HIGHBITDEPTH;
321 #if VPX_IMAGE_ABI_VERSION >= 3
322 case AV_PIX_FMT_YUV440P10LE:
323 case AV_PIX_FMT_YUV440P12LE:
324 if (codec_caps & VPX_CODEC_CAP_HIGHBITDEPTH) {
325 enccfg->g_bit_depth = enccfg->g_input_bit_depth =
326 avctx->pix_fmt == AV_PIX_FMT_YUV440P10LE ? 10 : 12;
327 enccfg->g_profile = 3;
328 *img_fmt = VPX_IMG_FMT_I44016;
329 *flags |= VPX_CODEC_USE_HIGHBITDEPTH;
334 case AV_PIX_FMT_YUV444P10LE:
335 case AV_PIX_FMT_YUV444P12LE:
336 if (codec_caps & VPX_CODEC_CAP_HIGHBITDEPTH) {
337 enccfg->g_bit_depth = enccfg->g_input_bit_depth =
338 avctx->pix_fmt == AV_PIX_FMT_YUV444P10LE ? 10 : 12;
339 enccfg->g_profile = 3;
340 *img_fmt = VPX_IMG_FMT_I44416;
341 *flags |= VPX_CODEC_USE_HIGHBITDEPTH;
349 av_log(avctx, AV_LOG_ERROR, "Unsupported pixel format.\n");
350 return AVERROR_INVALIDDATA;
354 static av_cold int vpx_init(AVCodecContext *avctx,
355 const struct vpx_codec_iface *iface)
357 VP8Context *ctx = avctx->priv_data;
358 struct vpx_codec_enc_cfg enccfg;
359 struct vpx_codec_enc_cfg enccfg_alpha;
360 vpx_codec_flags_t flags = (avctx->flags & CODEC_FLAG_PSNR) ? VPX_CODEC_USE_PSNR : 0;
362 vpx_img_fmt_t img_fmt = VPX_IMG_FMT_I420;
363 #if CONFIG_LIBVPX_VP9_ENCODER
364 vpx_codec_caps_t codec_caps = vpx_codec_get_caps(iface);
367 av_log(avctx, AV_LOG_INFO, "%s\n", vpx_codec_version_str());
368 av_log(avctx, AV_LOG_VERBOSE, "%s\n", vpx_codec_build_config());
370 if (avctx->pix_fmt == AV_PIX_FMT_YUVA420P)
373 if ((res = vpx_codec_enc_config_default(iface, &enccfg, 0)) != VPX_CODEC_OK) {
374 av_log(avctx, AV_LOG_ERROR, "Failed to get config: %s\n",
375 vpx_codec_err_to_string(res));
376 return AVERROR(EINVAL);
379 #if CONFIG_LIBVPX_VP9_ENCODER
380 if (avctx->codec_id == AV_CODEC_ID_VP9) {
381 if (set_pix_fmt(avctx, codec_caps, &enccfg, &flags, &img_fmt))
382 return AVERROR(EINVAL);
387 if(avctx->rc_max_rate || avctx->rc_buffer_size || avctx->rc_initial_buffer_occupancy) {
388 av_log( avctx, AV_LOG_ERROR, "Rate control parameters set without a bitrate\n");
389 return AVERROR(EINVAL);
392 dump_enc_cfg(avctx, &enccfg);
394 enccfg.g_w = avctx->width;
395 enccfg.g_h = avctx->height;
396 enccfg.g_timebase.num = avctx->time_base.num;
397 enccfg.g_timebase.den = avctx->time_base.den;
398 enccfg.g_threads = avctx->thread_count;
399 enccfg.g_lag_in_frames= ctx->lag_in_frames;
401 if (avctx->flags & CODEC_FLAG_PASS1)
402 enccfg.g_pass = VPX_RC_FIRST_PASS;
403 else if (avctx->flags & CODEC_FLAG_PASS2)
404 enccfg.g_pass = VPX_RC_LAST_PASS;
406 enccfg.g_pass = VPX_RC_ONE_PASS;
408 if (avctx->rc_min_rate == avctx->rc_max_rate &&
409 avctx->rc_min_rate == avctx->bit_rate && avctx->bit_rate) {
410 enccfg.rc_end_usage = VPX_CBR;
411 } else if (ctx->crf >= 0) {
412 enccfg.rc_end_usage = VPX_CQ;
413 #if CONFIG_LIBVPX_VP9_ENCODER
414 if (!avctx->bit_rate && avctx->codec_id == AV_CODEC_ID_VP9)
415 enccfg.rc_end_usage = VPX_Q;
419 if (avctx->bit_rate) {
420 enccfg.rc_target_bitrate = av_rescale_rnd(avctx->bit_rate, 1, 1000,
422 #if CONFIG_LIBVPX_VP9_ENCODER
423 } else if (enccfg.rc_end_usage == VPX_Q) {
426 if (enccfg.rc_end_usage == VPX_CQ) {
427 enccfg.rc_target_bitrate = 1000000;
429 avctx->bit_rate = enccfg.rc_target_bitrate * 1000;
430 av_log(avctx, AV_LOG_WARNING,
431 "Neither bitrate nor constrained quality specified, using default bitrate of %dkbit/sec\n",
432 enccfg.rc_target_bitrate);
436 if (avctx->codec_id == AV_CODEC_ID_VP9 && ctx->lossless == 1) {
437 enccfg.rc_min_quantizer =
438 enccfg.rc_max_quantizer = 0;
440 if (avctx->qmin >= 0)
441 enccfg.rc_min_quantizer = avctx->qmin;
442 if (avctx->qmax >= 0)
443 enccfg.rc_max_quantizer = avctx->qmax;
446 if (enccfg.rc_end_usage == VPX_CQ
447 #if CONFIG_LIBVPX_VP9_ENCODER
448 || enccfg.rc_end_usage == VPX_Q
451 if (ctx->crf < enccfg.rc_min_quantizer || ctx->crf > enccfg.rc_max_quantizer) {
452 av_log(avctx, AV_LOG_ERROR,
453 "CQ level %d must be between minimum and maximum quantizer value (%d-%d)\n",
454 ctx->crf, enccfg.rc_min_quantizer, enccfg.rc_max_quantizer);
455 return AVERROR(EINVAL);
459 enccfg.rc_dropframe_thresh = avctx->frame_skip_threshold;
461 //0-100 (0 => CBR, 100 => VBR)
462 enccfg.rc_2pass_vbr_bias_pct = round(avctx->qcompress * 100);
464 enccfg.rc_2pass_vbr_minsection_pct =
465 avctx->rc_min_rate * 100LL / avctx->bit_rate;
466 if (avctx->rc_max_rate)
467 enccfg.rc_2pass_vbr_maxsection_pct =
468 avctx->rc_max_rate * 100LL / avctx->bit_rate;
470 if (avctx->rc_buffer_size)
472 avctx->rc_buffer_size * 1000LL / avctx->bit_rate;
473 if (avctx->rc_initial_buffer_occupancy)
474 enccfg.rc_buf_initial_sz =
475 avctx->rc_initial_buffer_occupancy * 1000LL / avctx->bit_rate;
476 enccfg.rc_buf_optimal_sz = enccfg.rc_buf_sz * 5 / 6;
478 FF_DISABLE_DEPRECATION_WARNINGS
479 if (avctx->rc_buffer_aggressivity != 1.0) {
480 av_log(avctx, AV_LOG_WARNING, "The rc_buffer_aggressivity option is "
481 "deprecated, use the undershoot-pct private option instead.\n");
482 enccfg.rc_undershoot_pct = round(avctx->rc_buffer_aggressivity * 100);
484 FF_ENABLE_DEPRECATION_WARNINGS
486 if (ctx->rc_undershoot_pct >= 0)
487 enccfg.rc_undershoot_pct = ctx->rc_undershoot_pct;
488 if (ctx->rc_overshoot_pct >= 0)
489 enccfg.rc_overshoot_pct = ctx->rc_overshoot_pct;
491 //_enc_init() will balk if kf_min_dist differs from max w/VPX_KF_AUTO
492 if (avctx->keyint_min >= 0 && avctx->keyint_min == avctx->gop_size)
493 enccfg.kf_min_dist = avctx->keyint_min;
494 if (avctx->gop_size >= 0)
495 enccfg.kf_max_dist = avctx->gop_size;
497 if (enccfg.g_pass == VPX_RC_FIRST_PASS)
498 enccfg.g_lag_in_frames = 0;
499 else if (enccfg.g_pass == VPX_RC_LAST_PASS) {
500 int decode_size, ret;
502 if (!avctx->stats_in) {
503 av_log(avctx, AV_LOG_ERROR, "No stats file for second pass\n");
504 return AVERROR_INVALIDDATA;
507 ctx->twopass_stats.sz = strlen(avctx->stats_in) * 3 / 4;
508 ret = av_reallocp(&ctx->twopass_stats.buf, ctx->twopass_stats.sz);
510 av_log(avctx, AV_LOG_ERROR,
511 "Stat buffer alloc (%"SIZE_SPECIFIER" bytes) failed\n",
512 ctx->twopass_stats.sz);
513 ctx->twopass_stats.sz = 0;
516 decode_size = av_base64_decode(ctx->twopass_stats.buf, avctx->stats_in,
517 ctx->twopass_stats.sz);
518 if (decode_size < 0) {
519 av_log(avctx, AV_LOG_ERROR, "Stat buffer decode failed\n");
520 return AVERROR_INVALIDDATA;
523 ctx->twopass_stats.sz = decode_size;
524 enccfg.rc_twopass_stats_in = ctx->twopass_stats;
527 /* 0-3: For non-zero values the encoder increasingly optimizes for reduced
528 complexity playback on low powered devices at the expense of encode
530 if (avctx->profile != FF_PROFILE_UNKNOWN)
531 enccfg.g_profile = avctx->profile;
533 enccfg.g_error_resilient = ctx->error_resilient || ctx->flags & VP8F_ERROR_RESILIENT;
535 dump_enc_cfg(avctx, &enccfg);
536 /* Construct Encoder Context */
537 res = vpx_codec_enc_init(&ctx->encoder, iface, &enccfg, flags);
538 if (res != VPX_CODEC_OK) {
539 log_encoder_error(avctx, "Failed to initialize encoder");
540 return AVERROR(EINVAL);
544 enccfg_alpha = enccfg;
545 res = vpx_codec_enc_init(&ctx->encoder_alpha, iface, &enccfg_alpha, flags);
546 if (res != VPX_CODEC_OK) {
547 log_encoder_error(avctx, "Failed to initialize alpha encoder");
548 return AVERROR(EINVAL);
552 //codec control failures are currently treated only as warnings
553 av_log(avctx, AV_LOG_DEBUG, "vpx_codec_control\n");
554 codecctl_int(avctx, VP8E_SET_CPUUSED, ctx->cpu_used);
555 if (ctx->flags & VP8F_AUTO_ALT_REF)
556 ctx->auto_alt_ref = 1;
557 if (ctx->auto_alt_ref >= 0)
558 codecctl_int(avctx, VP8E_SET_ENABLEAUTOALTREF, ctx->auto_alt_ref);
559 if (ctx->arnr_max_frames >= 0)
560 codecctl_int(avctx, VP8E_SET_ARNR_MAXFRAMES, ctx->arnr_max_frames);
561 if (ctx->arnr_strength >= 0)
562 codecctl_int(avctx, VP8E_SET_ARNR_STRENGTH, ctx->arnr_strength);
563 if (ctx->arnr_type >= 0)
564 codecctl_int(avctx, VP8E_SET_ARNR_TYPE, ctx->arnr_type);
565 if (avctx->codec_id == AV_CODEC_ID_VP8) {
566 codecctl_int(avctx, VP8E_SET_NOISE_SENSITIVITY, avctx->noise_reduction);
567 codecctl_int(avctx, VP8E_SET_TOKEN_PARTITIONS, av_log2(avctx->slices));
570 FF_DISABLE_DEPRECATION_WARNINGS
571 if (avctx->mb_threshold) {
572 av_log(avctx, AV_LOG_WARNING, "The mb_threshold option is deprecated, "
573 "use the static-thresh private option instead.\n");
574 ctx->static_thresh = avctx->mb_threshold;
576 FF_ENABLE_DEPRECATION_WARNINGS
578 codecctl_int(avctx, VP8E_SET_STATIC_THRESHOLD, ctx->static_thresh);
580 codecctl_int(avctx, VP8E_SET_CQ_LEVEL, ctx->crf);
581 if (ctx->max_intra_rate >= 0)
582 codecctl_int(avctx, VP8E_SET_MAX_INTRA_BITRATE_PCT, ctx->max_intra_rate);
584 #if CONFIG_LIBVPX_VP9_ENCODER
585 if (avctx->codec_id == AV_CODEC_ID_VP9) {
586 if (ctx->lossless >= 0)
587 codecctl_int(avctx, VP9E_SET_LOSSLESS, ctx->lossless);
588 if (ctx->tile_columns >= 0)
589 codecctl_int(avctx, VP9E_SET_TILE_COLUMNS, ctx->tile_columns);
590 if (ctx->tile_rows >= 0)
591 codecctl_int(avctx, VP9E_SET_TILE_ROWS, ctx->tile_rows);
592 if (ctx->frame_parallel >= 0)
593 codecctl_int(avctx, VP9E_SET_FRAME_PARALLEL_DECODING, ctx->frame_parallel);
594 if (ctx->aq_mode >= 0)
595 codecctl_int(avctx, VP9E_SET_AQ_MODE, ctx->aq_mode);
599 av_log(avctx, AV_LOG_DEBUG, "Using deadline: %d\n", ctx->deadline);
601 //provide dummy value to initialize wrapper, values will be updated each _encode()
602 vpx_img_wrap(&ctx->rawimg, img_fmt, avctx->width, avctx->height, 1,
604 #if CONFIG_LIBVPX_VP9_ENCODER && defined(VPX_IMG_FMT_HIGHBITDEPTH)
605 if (avctx->codec_id == AV_CODEC_ID_VP9 && (codec_caps & VPX_CODEC_CAP_HIGHBITDEPTH))
606 ctx->rawimg.bit_depth = enccfg.g_bit_depth;
610 vpx_img_wrap(&ctx->rawimg_alpha, VPX_IMG_FMT_I420, avctx->width, avctx->height, 1,
613 avctx->coded_frame = av_frame_alloc();
614 if (!avctx->coded_frame) {
615 av_log(avctx, AV_LOG_ERROR, "Error allocating coded frame\n");
617 return AVERROR(ENOMEM);
622 static inline void cx_pktcpy(struct FrameListData *dst,
623 const struct vpx_codec_cx_pkt *src,
624 const struct vpx_codec_cx_pkt *src_alpha,
627 dst->pts = src->data.frame.pts;
628 dst->duration = src->data.frame.duration;
629 dst->flags = src->data.frame.flags;
630 dst->sz = src->data.frame.sz;
631 dst->buf = src->data.frame.buf;
633 /* For alt-ref frame, don't store PSNR or increment frame_number */
634 if (!(dst->flags & VPX_FRAME_IS_INVISIBLE)) {
635 dst->frame_number = ++ctx->frame_number;
636 dst->have_sse = ctx->have_sse;
638 /* associate last-seen SSE to the frame. */
639 /* Transfers ownership from ctx to dst. */
640 /* WARNING! This makes the assumption that PSNR_PKT comes
641 just before the frame it refers to! */
642 memcpy(dst->sse, ctx->sse, sizeof(dst->sse));
646 dst->frame_number = -1; /* sanity marker */
649 dst->buf_alpha = src_alpha->data.frame.buf;
650 dst->sz_alpha = src_alpha->data.frame.sz;
652 dst->buf_alpha = NULL;
658 * Store coded frame information in format suitable for return from encode2().
660 * Write information from @a cx_frame to @a pkt
661 * @return packet data size on success
662 * @return a negative AVERROR on error
664 static int storeframe(AVCodecContext *avctx, struct FrameListData *cx_frame,
665 AVPacket *pkt, AVFrame *coded_frame)
667 int ret = ff_alloc_packet2(avctx, pkt, cx_frame->sz);
670 memcpy(pkt->data, cx_frame->buf, pkt->size);
671 pkt->pts = pkt->dts = cx_frame->pts;
672 coded_frame->pts = cx_frame->pts;
673 coded_frame->key_frame = !!(cx_frame->flags & VPX_FRAME_IS_KEY);
675 if (coded_frame->key_frame) {
676 coded_frame->pict_type = AV_PICTURE_TYPE_I;
677 pkt->flags |= AV_PKT_FLAG_KEY;
679 coded_frame->pict_type = AV_PICTURE_TYPE_P;
681 if (cx_frame->have_sse) {
683 /* Beware of the Y/U/V/all order! */
684 coded_frame->error[0] = cx_frame->sse[1];
685 coded_frame->error[1] = cx_frame->sse[2];
686 coded_frame->error[2] = cx_frame->sse[3];
687 coded_frame->error[3] = 0; // alpha
688 for (i = 0; i < 4; ++i) {
689 avctx->error[i] += coded_frame->error[i];
691 cx_frame->have_sse = 0;
693 if (cx_frame->sz_alpha > 0) {
694 side_data = av_packet_new_side_data(pkt,
695 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
696 cx_frame->sz_alpha + 8);
700 return AVERROR(ENOMEM);
702 AV_WB64(side_data, 1);
703 memcpy(side_data + 8, cx_frame->buf_alpha, cx_frame->sz_alpha);
712 * Queue multiple output frames from the encoder, returning the front-most.
713 * In cases where vpx_codec_get_cx_data() returns more than 1 frame append
714 * the frame queue. Return the head frame if available.
715 * @return Stored frame size
716 * @return AVERROR(EINVAL) on output size error
717 * @return AVERROR(ENOMEM) on coded frame queue data allocation error
719 static int queue_frames(AVCodecContext *avctx, AVPacket *pkt_out,
720 AVFrame *coded_frame)
722 VP8Context *ctx = avctx->priv_data;
723 const struct vpx_codec_cx_pkt *pkt;
724 const struct vpx_codec_cx_pkt *pkt_alpha = NULL;
725 const void *iter = NULL;
726 const void *iter_alpha = NULL;
729 if (ctx->coded_frame_list) {
730 struct FrameListData *cx_frame = ctx->coded_frame_list;
731 /* return the leading frame if we've already begun queueing */
732 size = storeframe(avctx, cx_frame, pkt_out, coded_frame);
735 ctx->coded_frame_list = cx_frame->next;
736 free_coded_frame(cx_frame);
739 /* consume all available output from the encoder before returning. buffers
740 are only good through the next vpx_codec call */
741 while ((pkt = vpx_codec_get_cx_data(&ctx->encoder, &iter)) &&
743 (ctx->is_alpha && (pkt_alpha = vpx_codec_get_cx_data(&ctx->encoder_alpha, &iter_alpha))))) {
745 case VPX_CODEC_CX_FRAME_PKT:
747 struct FrameListData cx_frame;
749 /* avoid storing the frame when the list is empty and we haven't yet
750 provided a frame for output */
751 av_assert0(!ctx->coded_frame_list);
752 cx_pktcpy(&cx_frame, pkt, pkt_alpha, ctx);
753 size = storeframe(avctx, &cx_frame, pkt_out, coded_frame);
757 struct FrameListData *cx_frame =
758 av_malloc(sizeof(struct FrameListData));
761 av_log(avctx, AV_LOG_ERROR,
762 "Frame queue element alloc failed\n");
763 return AVERROR(ENOMEM);
765 cx_pktcpy(cx_frame, pkt, pkt_alpha, ctx);
766 cx_frame->buf = av_malloc(cx_frame->sz);
768 if (!cx_frame->buf) {
769 av_log(avctx, AV_LOG_ERROR,
770 "Data buffer alloc (%"SIZE_SPECIFIER" bytes) failed\n",
773 return AVERROR(ENOMEM);
775 memcpy(cx_frame->buf, pkt->data.frame.buf, pkt->data.frame.sz);
777 cx_frame->buf_alpha = av_malloc(cx_frame->sz_alpha);
778 if (!cx_frame->buf_alpha) {
779 av_log(avctx, AV_LOG_ERROR,
780 "Data buffer alloc (%"SIZE_SPECIFIER" bytes) failed\n",
783 return AVERROR(ENOMEM);
785 memcpy(cx_frame->buf_alpha, pkt_alpha->data.frame.buf, pkt_alpha->data.frame.sz);
787 coded_frame_add(&ctx->coded_frame_list, cx_frame);
790 case VPX_CODEC_STATS_PKT: {
791 struct vpx_fixed_buf *stats = &ctx->twopass_stats;
793 if ((err = av_reallocp(&stats->buf,
795 pkt->data.twopass_stats.sz)) < 0) {
797 av_log(avctx, AV_LOG_ERROR, "Stat buffer realloc failed\n");
800 memcpy((uint8_t*)stats->buf + stats->sz,
801 pkt->data.twopass_stats.buf, pkt->data.twopass_stats.sz);
802 stats->sz += pkt->data.twopass_stats.sz;
805 case VPX_CODEC_PSNR_PKT:
806 av_assert0(!ctx->have_sse);
807 ctx->sse[0] = pkt->data.psnr.sse[0];
808 ctx->sse[1] = pkt->data.psnr.sse[1];
809 ctx->sse[2] = pkt->data.psnr.sse[2];
810 ctx->sse[3] = pkt->data.psnr.sse[3];
813 case VPX_CODEC_CUSTOM_PKT:
814 //ignore unsupported/unrecognized packet types
822 static int vp8_encode(AVCodecContext *avctx, AVPacket *pkt,
823 const AVFrame *frame, int *got_packet)
825 VP8Context *ctx = avctx->priv_data;
826 struct vpx_image *rawimg = NULL;
827 struct vpx_image *rawimg_alpha = NULL;
828 int64_t timestamp = 0;
830 vpx_enc_frame_flags_t flags = 0;
833 rawimg = &ctx->rawimg;
834 rawimg->planes[VPX_PLANE_Y] = frame->data[0];
835 rawimg->planes[VPX_PLANE_U] = frame->data[1];
836 rawimg->planes[VPX_PLANE_V] = frame->data[2];
837 rawimg->stride[VPX_PLANE_Y] = frame->linesize[0];
838 rawimg->stride[VPX_PLANE_U] = frame->linesize[1];
839 rawimg->stride[VPX_PLANE_V] = frame->linesize[2];
841 uint8_t *u_plane, *v_plane;
842 rawimg_alpha = &ctx->rawimg_alpha;
843 rawimg_alpha->planes[VPX_PLANE_Y] = frame->data[3];
844 u_plane = av_malloc(frame->linesize[1] * frame->height);
845 v_plane = av_malloc(frame->linesize[2] * frame->height);
846 if (!u_plane || !v_plane) {
849 return AVERROR(ENOMEM);
851 memset(u_plane, 0x80, frame->linesize[1] * frame->height);
852 rawimg_alpha->planes[VPX_PLANE_U] = u_plane;
853 memset(v_plane, 0x80, frame->linesize[2] * frame->height);
854 rawimg_alpha->planes[VPX_PLANE_V] = v_plane;
855 rawimg_alpha->stride[VPX_PLANE_Y] = frame->linesize[0];
856 rawimg_alpha->stride[VPX_PLANE_U] = frame->linesize[1];
857 rawimg_alpha->stride[VPX_PLANE_V] = frame->linesize[2];
859 timestamp = frame->pts;
860 if (frame->pict_type == AV_PICTURE_TYPE_I)
861 flags |= VPX_EFLAG_FORCE_KF;
864 res = vpx_codec_encode(&ctx->encoder, rawimg, timestamp,
865 avctx->ticks_per_frame, flags, ctx->deadline);
866 if (res != VPX_CODEC_OK) {
867 log_encoder_error(avctx, "Error encoding frame");
868 return AVERROR_INVALIDDATA;
872 res = vpx_codec_encode(&ctx->encoder_alpha, rawimg_alpha, timestamp,
873 avctx->ticks_per_frame, flags, ctx->deadline);
874 if (res != VPX_CODEC_OK) {
875 log_encoder_error(avctx, "Error encoding alpha frame");
876 return AVERROR_INVALIDDATA;
880 coded_size = queue_frames(avctx, pkt, avctx->coded_frame);
882 if (!frame && avctx->flags & CODEC_FLAG_PASS1) {
883 unsigned int b64_size = AV_BASE64_SIZE(ctx->twopass_stats.sz);
885 avctx->stats_out = av_malloc(b64_size);
886 if (!avctx->stats_out) {
887 av_log(avctx, AV_LOG_ERROR, "Stat buffer alloc (%d bytes) failed\n",
889 return AVERROR(ENOMEM);
891 av_base64_encode(avctx->stats_out, b64_size, ctx->twopass_stats.buf,
892 ctx->twopass_stats.sz);
896 av_freep(&rawimg_alpha->planes[VPX_PLANE_U]);
897 av_freep(&rawimg_alpha->planes[VPX_PLANE_V]);
900 *got_packet = !!coded_size;
904 #define OFFSET(x) offsetof(VP8Context, x)
905 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
907 #ifndef VPX_ERROR_RESILIENT_DEFAULT
908 #define VPX_ERROR_RESILIENT_DEFAULT 1
909 #define VPX_ERROR_RESILIENT_PARTITIONS 2
912 #define COMMON_OPTIONS \
913 { "cpu-used", "Quality/Speed ratio modifier", OFFSET(cpu_used), AV_OPT_TYPE_INT, {.i64 = 1}, -16, 16, VE}, \
914 { "auto-alt-ref", "Enable use of alternate reference " \
915 "frames (2-pass only)", OFFSET(auto_alt_ref), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE}, \
916 { "lag-in-frames", "Number of frames to look ahead for " \
917 "alternate reference frame selection", OFFSET(lag_in_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
918 { "arnr-maxframes", "altref noise reduction max frame count", OFFSET(arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
919 { "arnr-strength", "altref noise reduction filter strength", OFFSET(arnr_strength), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
920 { "arnr-type", "altref noise reduction filter type", OFFSET(arnr_type), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE, "arnr_type"}, \
921 { "backward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0, VE, "arnr_type" }, \
922 { "forward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0, VE, "arnr_type" }, \
923 { "centered", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 3}, 0, 0, VE, "arnr_type" }, \
924 { "deadline", "Time to spend encoding, in microseconds.", OFFSET(deadline), AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, "quality"}, \
925 { "best", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_BEST_QUALITY}, 0, 0, VE, "quality"}, \
926 { "good", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_GOOD_QUALITY}, 0, 0, VE, "quality"}, \
927 { "realtime", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_REALTIME}, 0, 0, VE, "quality"}, \
928 { "error-resilient", "Error resilience configuration", OFFSET(error_resilient), AV_OPT_TYPE_FLAGS, {.i64 = 0}, INT_MIN, INT_MAX, VE, "er"}, \
929 { "max-intra-rate", "Maximum I-frame bitrate (pct) 0=unlimited", OFFSET(max_intra_rate), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
930 { "default", "Improve resiliency against losses of whole frames", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_DEFAULT}, 0, 0, VE, "er"}, \
931 { "partitions", "The frame partitions are independently decodable " \
932 "by the bool decoder, meaning that partitions can be decoded even " \
933 "though earlier partitions have been lost. Note that intra predicition" \
934 " is still done over the partition boundary.", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_PARTITIONS}, 0, 0, VE, "er"}, \
935 { "crf", "Select the quality for constant quality mode", offsetof(VP8Context, crf), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 63, VE }, \
936 { "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 }, \
937 { "undershoot-pct", "Datarate undershoot (min) target (%)", OFFSET(rc_undershoot_pct), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 100, VE }, \
938 { "overshoot-pct", "Datarate overshoot (max) target (%)", OFFSET(rc_overshoot_pct), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 1000, VE }, \
940 #define LEGACY_OPTIONS \
941 {"speed", "", offsetof(VP8Context, cpu_used), AV_OPT_TYPE_INT, {.i64 = 1}, -16, 16, VE}, \
942 {"quality", "", offsetof(VP8Context, deadline), AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, "quality"}, \
943 {"vp8flags", "", offsetof(VP8Context, flags), FF_OPT_TYPE_FLAGS, {.i64 = 0}, 0, UINT_MAX, VE, "flags"}, \
944 {"error_resilient", "enable error resilience", 0, FF_OPT_TYPE_CONST, {.dbl = VP8F_ERROR_RESILIENT}, INT_MIN, INT_MAX, VE, "flags"}, \
945 {"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"}, \
946 {"arnr_max_frames", "altref noise reduction max frame count", offsetof(VP8Context, arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 15, VE}, \
947 {"arnr_strength", "altref noise reduction filter strength", offsetof(VP8Context, arnr_strength), AV_OPT_TYPE_INT, {.i64 = 3}, 0, 6, VE}, \
948 {"arnr_type", "altref noise reduction filter type", offsetof(VP8Context, arnr_type), AV_OPT_TYPE_INT, {.i64 = 3}, 1, 3, VE}, \
949 {"rc_lookahead", "Number of frames to look ahead for alternate reference frame selection", offsetof(VP8Context, lag_in_frames), AV_OPT_TYPE_INT, {.i64 = 25}, 0, 25, VE}, \
951 #if CONFIG_LIBVPX_VP8_ENCODER
952 static const AVOption vp8_options[] = {
959 #if CONFIG_LIBVPX_VP9_ENCODER
960 static const AVOption vp9_options[] = {
962 { "lossless", "Lossless mode", OFFSET(lossless), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE},
963 { "tile-columns", "Number of tile columns to use, log2", OFFSET(tile_columns), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 6, VE},
964 { "tile-rows", "Number of tile rows to use, log2", OFFSET(tile_rows), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 2, VE},
965 { "frame-parallel", "Enable frame parallel decodability features", OFFSET(frame_parallel), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE},
966 { "aq-mode", "adaptive quantization mode", OFFSET(aq_mode), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 3, VE, "aq_mode"},
967 { "none", "Aq not used", 0, AV_OPT_TYPE_CONST, {.i64 = 0}, 0, 0, VE, "aq_mode" },
968 { "variance", "Variance based Aq", 0, AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0, VE, "aq_mode" },
969 { "complexity", "Complexity based Aq", 0, AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0, VE, "aq_mode" },
970 { "cyclic", "Cyclic Refresh Aq", 0, AV_OPT_TYPE_CONST, {.i64 = 3}, 0, 0, VE, "aq_mode" },
976 #undef COMMON_OPTIONS
977 #undef LEGACY_OPTIONS
979 static const AVCodecDefault defaults[] = {
983 { "keyint_min", "-1" },
987 #if CONFIG_LIBVPX_VP8_ENCODER
988 static av_cold int vp8_init(AVCodecContext *avctx)
990 return vpx_init(avctx, vpx_codec_vp8_cx());
993 static const AVClass class_vp8 = {
994 .class_name = "libvpx-vp8 encoder",
995 .item_name = av_default_item_name,
996 .option = vp8_options,
997 .version = LIBAVUTIL_VERSION_INT,
1000 AVCodec ff_libvpx_vp8_encoder = {
1002 .long_name = NULL_IF_CONFIG_SMALL("libvpx VP8"),
1003 .type = AVMEDIA_TYPE_VIDEO,
1004 .id = AV_CODEC_ID_VP8,
1005 .priv_data_size = sizeof(VP8Context),
1007 .encode2 = vp8_encode,
1009 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
1010 .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUVA420P, AV_PIX_FMT_NONE },
1011 .priv_class = &class_vp8,
1012 .defaults = defaults,
1014 #endif /* CONFIG_LIBVPX_VP8_ENCODER */
1016 #if CONFIG_LIBVPX_VP9_ENCODER
1017 static av_cold int vp9_init(AVCodecContext *avctx)
1019 return vpx_init(avctx, vpx_codec_vp9_cx());
1022 static const AVClass class_vp9 = {
1023 .class_name = "libvpx-vp9 encoder",
1024 .item_name = av_default_item_name,
1025 .option = vp9_options,
1026 .version = LIBAVUTIL_VERSION_INT,
1029 AVCodec ff_libvpx_vp9_encoder = {
1030 .name = "libvpx-vp9",
1031 .long_name = NULL_IF_CONFIG_SMALL("libvpx VP9"),
1032 .type = AVMEDIA_TYPE_VIDEO,
1033 .id = AV_CODEC_ID_VP9,
1034 .priv_data_size = sizeof(VP8Context),
1036 .encode2 = vp8_encode,
1038 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
1039 .priv_class = &class_vp9,
1040 .defaults = defaults,
1041 .init_static_data = ff_vp9_init_static,
1043 #endif /* CONFIG_LIBVPX_VP9_ENCODER */