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
103 /** String mappings for enum vp8e_enc_control_id */
104 static const char *const ctlidstr[] = {
105 [VP8E_UPD_ENTROPY] = "VP8E_UPD_ENTROPY",
106 [VP8E_UPD_REFERENCE] = "VP8E_UPD_REFERENCE",
107 [VP8E_USE_REFERENCE] = "VP8E_USE_REFERENCE",
108 [VP8E_SET_ROI_MAP] = "VP8E_SET_ROI_MAP",
109 [VP8E_SET_ACTIVEMAP] = "VP8E_SET_ACTIVEMAP",
110 [VP8E_SET_SCALEMODE] = "VP8E_SET_SCALEMODE",
111 [VP8E_SET_CPUUSED] = "VP8E_SET_CPUUSED",
112 [VP8E_SET_ENABLEAUTOALTREF] = "VP8E_SET_ENABLEAUTOALTREF",
113 [VP8E_SET_NOISE_SENSITIVITY] = "VP8E_SET_NOISE_SENSITIVITY",
114 [VP8E_SET_SHARPNESS] = "VP8E_SET_SHARPNESS",
115 [VP8E_SET_STATIC_THRESHOLD] = "VP8E_SET_STATIC_THRESHOLD",
116 [VP8E_SET_TOKEN_PARTITIONS] = "VP8E_SET_TOKEN_PARTITIONS",
117 [VP8E_GET_LAST_QUANTIZER] = "VP8E_GET_LAST_QUANTIZER",
118 [VP8E_SET_ARNR_MAXFRAMES] = "VP8E_SET_ARNR_MAXFRAMES",
119 [VP8E_SET_ARNR_STRENGTH] = "VP8E_SET_ARNR_STRENGTH",
120 [VP8E_SET_ARNR_TYPE] = "VP8E_SET_ARNR_TYPE",
121 [VP8E_SET_CQ_LEVEL] = "VP8E_SET_CQ_LEVEL",
122 [VP8E_SET_MAX_INTRA_BITRATE_PCT] = "VP8E_SET_MAX_INTRA_BITRATE_PCT",
123 #if CONFIG_LIBVPX_VP9_ENCODER
124 [VP9E_SET_LOSSLESS] = "VP9E_SET_LOSSLESS",
125 [VP9E_SET_TILE_COLUMNS] = "VP9E_SET_TILE_COLUMNS",
126 [VP9E_SET_TILE_ROWS] = "VP9E_SET_TILE_ROWS",
127 [VP9E_SET_FRAME_PARALLEL_DECODING] = "VP9E_SET_FRAME_PARALLEL_DECODING",
128 [VP9E_SET_AQ_MODE] = "VP9E_SET_AQ_MODE",
132 static av_cold void log_encoder_error(AVCodecContext *avctx, const char *desc)
134 VP8Context *ctx = avctx->priv_data;
135 const char *error = vpx_codec_error(&ctx->encoder);
136 const char *detail = vpx_codec_error_detail(&ctx->encoder);
138 av_log(avctx, AV_LOG_ERROR, "%s: %s\n", desc, error);
140 av_log(avctx, AV_LOG_ERROR, " Additional information: %s\n", detail);
143 static av_cold void dump_enc_cfg(AVCodecContext *avctx,
144 const struct vpx_codec_enc_cfg *cfg)
147 int level = AV_LOG_DEBUG;
149 av_log(avctx, level, "vpx_codec_enc_cfg\n");
150 av_log(avctx, level, "generic settings\n"
151 " %*s%u\n %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
152 #if CONFIG_LIBVPX_VP9_ENCODER && defined(VPX_IMG_FMT_HIGHBITDEPTH)
155 " %*s{%u/%u}\n %*s%u\n %*s%d\n %*s%u\n",
156 width, "g_usage:", cfg->g_usage,
157 width, "g_threads:", cfg->g_threads,
158 width, "g_profile:", cfg->g_profile,
159 width, "g_w:", cfg->g_w,
160 width, "g_h:", cfg->g_h,
161 #if CONFIG_LIBVPX_VP9_ENCODER && defined(VPX_IMG_FMT_HIGHBITDEPTH)
162 width, "g_bit_depth:", cfg->g_bit_depth,
163 width, "g_input_bit_depth:", cfg->g_input_bit_depth,
165 width, "g_timebase:", cfg->g_timebase.num, cfg->g_timebase.den,
166 width, "g_error_resilient:", cfg->g_error_resilient,
167 width, "g_pass:", cfg->g_pass,
168 width, "g_lag_in_frames:", cfg->g_lag_in_frames);
169 av_log(avctx, level, "rate control settings\n"
170 " %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
171 " %*s%d\n %*s%p(%"SIZE_SPECIFIER")\n %*s%u\n",
172 width, "rc_dropframe_thresh:", cfg->rc_dropframe_thresh,
173 width, "rc_resize_allowed:", cfg->rc_resize_allowed,
174 width, "rc_resize_up_thresh:", cfg->rc_resize_up_thresh,
175 width, "rc_resize_down_thresh:", cfg->rc_resize_down_thresh,
176 width, "rc_end_usage:", cfg->rc_end_usage,
177 width, "rc_twopass_stats_in:", cfg->rc_twopass_stats_in.buf, cfg->rc_twopass_stats_in.sz,
178 width, "rc_target_bitrate:", cfg->rc_target_bitrate);
179 av_log(avctx, level, "quantizer settings\n"
181 width, "rc_min_quantizer:", cfg->rc_min_quantizer,
182 width, "rc_max_quantizer:", cfg->rc_max_quantizer);
183 av_log(avctx, level, "bitrate tolerance\n"
185 width, "rc_undershoot_pct:", cfg->rc_undershoot_pct,
186 width, "rc_overshoot_pct:", cfg->rc_overshoot_pct);
187 av_log(avctx, level, "decoder buffer model\n"
188 " %*s%u\n %*s%u\n %*s%u\n",
189 width, "rc_buf_sz:", cfg->rc_buf_sz,
190 width, "rc_buf_initial_sz:", cfg->rc_buf_initial_sz,
191 width, "rc_buf_optimal_sz:", cfg->rc_buf_optimal_sz);
192 av_log(avctx, level, "2 pass rate control settings\n"
193 " %*s%u\n %*s%u\n %*s%u\n",
194 width, "rc_2pass_vbr_bias_pct:", cfg->rc_2pass_vbr_bias_pct,
195 width, "rc_2pass_vbr_minsection_pct:", cfg->rc_2pass_vbr_minsection_pct,
196 width, "rc_2pass_vbr_maxsection_pct:", cfg->rc_2pass_vbr_maxsection_pct);
197 av_log(avctx, level, "keyframing settings\n"
198 " %*s%d\n %*s%u\n %*s%u\n",
199 width, "kf_mode:", cfg->kf_mode,
200 width, "kf_min_dist:", cfg->kf_min_dist,
201 width, "kf_max_dist:", cfg->kf_max_dist);
202 av_log(avctx, level, "\n");
205 static void coded_frame_add(void *list, struct FrameListData *cx_frame)
207 struct FrameListData **p = list;
212 cx_frame->next = NULL;
215 static av_cold void free_coded_frame(struct FrameListData *cx_frame)
217 av_freep(&cx_frame->buf);
218 if (cx_frame->buf_alpha)
219 av_freep(&cx_frame->buf_alpha);
223 static av_cold void free_frame_list(struct FrameListData *list)
225 struct FrameListData *p = list;
234 static av_cold int codecctl_int(AVCodecContext *avctx,
235 enum vp8e_enc_control_id id, int val)
237 VP8Context *ctx = avctx->priv_data;
242 snprintf(buf, sizeof(buf), "%s:", ctlidstr[id]);
243 av_log(avctx, AV_LOG_DEBUG, " %*s%d\n", width, buf, val);
245 res = vpx_codec_control(&ctx->encoder, id, val);
246 if (res != VPX_CODEC_OK) {
247 snprintf(buf, sizeof(buf), "Failed to set %s codec control",
249 log_encoder_error(avctx, buf);
252 return res == VPX_CODEC_OK ? 0 : AVERROR(EINVAL);
255 static av_cold int vp8_free(AVCodecContext *avctx)
257 VP8Context *ctx = avctx->priv_data;
259 vpx_codec_destroy(&ctx->encoder);
261 vpx_codec_destroy(&ctx->encoder_alpha);
262 av_freep(&ctx->twopass_stats.buf);
263 av_frame_free(&avctx->coded_frame);
264 av_freep(&avctx->stats_out);
265 free_frame_list(ctx->coded_frame_list);
269 #if CONFIG_LIBVPX_VP9_ENCODER
270 static int set_pix_fmt(AVCodecContext *avctx, vpx_codec_caps_t codec_caps,
271 struct vpx_codec_enc_cfg *enccfg, vpx_codec_flags_t *flags,
272 vpx_img_fmt_t *img_fmt)
274 #ifdef VPX_IMG_FMT_HIGHBITDEPTH
275 enccfg->g_bit_depth = enccfg->g_input_bit_depth = 8;
277 switch (avctx->pix_fmt) {
278 case AV_PIX_FMT_YUV420P:
279 enccfg->g_profile = 0;
280 *img_fmt = VPX_IMG_FMT_I420;
282 case AV_PIX_FMT_YUV422P:
283 case AV_PIX_FMT_YUV444P:
284 enccfg->g_profile = 1;
285 *img_fmt = avctx->pix_fmt == AV_PIX_FMT_YUV422P ? VPX_IMG_FMT_I422 : VPX_IMG_FMT_I444;
287 #ifdef VPX_IMG_FMT_HIGHBITDEPTH
288 case AV_PIX_FMT_YUV420P10LE:
289 case AV_PIX_FMT_YUV420P12LE:
290 if (codec_caps & VPX_CODEC_CAP_HIGHBITDEPTH) {
291 enccfg->g_bit_depth = enccfg->g_input_bit_depth =
292 avctx->pix_fmt == AV_PIX_FMT_YUV420P10LE ? 10 : 12;
293 enccfg->g_profile = 2;
294 *img_fmt = VPX_IMG_FMT_I42016;
295 *flags |= VPX_CODEC_USE_HIGHBITDEPTH;
299 case AV_PIX_FMT_YUV422P10LE:
300 case AV_PIX_FMT_YUV422P12LE:
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_YUV422P10LE ? 10 : 12;
304 enccfg->g_profile = 3;
305 *img_fmt = VPX_IMG_FMT_I42216;
306 *flags |= VPX_CODEC_USE_HIGHBITDEPTH;
310 case AV_PIX_FMT_YUV444P10LE:
311 case AV_PIX_FMT_YUV444P12LE:
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_YUV444P10LE ? 10 : 12;
315 enccfg->g_profile = 3;
316 *img_fmt = VPX_IMG_FMT_I44416;
317 *flags |= VPX_CODEC_USE_HIGHBITDEPTH;
325 av_log(avctx, AV_LOG_ERROR, "Unsupported pixel format.\n");
326 return AVERROR_INVALIDDATA;
330 static av_cold int vpx_init(AVCodecContext *avctx,
331 const struct vpx_codec_iface *iface)
333 VP8Context *ctx = avctx->priv_data;
334 struct vpx_codec_enc_cfg enccfg;
335 struct vpx_codec_enc_cfg enccfg_alpha;
336 vpx_codec_flags_t flags = (avctx->flags & CODEC_FLAG_PSNR) ? VPX_CODEC_USE_PSNR : 0;
338 vpx_img_fmt_t img_fmt = VPX_IMG_FMT_I420;
339 #if CONFIG_LIBVPX_VP9_ENCODER
340 vpx_codec_caps_t codec_caps = vpx_codec_get_caps(iface);
343 av_log(avctx, AV_LOG_INFO, "%s\n", vpx_codec_version_str());
344 av_log(avctx, AV_LOG_VERBOSE, "%s\n", vpx_codec_build_config());
346 if (avctx->pix_fmt == AV_PIX_FMT_YUVA420P)
349 if ((res = vpx_codec_enc_config_default(iface, &enccfg, 0)) != VPX_CODEC_OK) {
350 av_log(avctx, AV_LOG_ERROR, "Failed to get config: %s\n",
351 vpx_codec_err_to_string(res));
352 return AVERROR(EINVAL);
355 #if CONFIG_LIBVPX_VP9_ENCODER
356 if (avctx->codec_id == AV_CODEC_ID_VP9) {
357 if (set_pix_fmt(avctx, codec_caps, &enccfg, &flags, &img_fmt))
358 return AVERROR(EINVAL);
363 if(avctx->rc_max_rate || avctx->rc_buffer_size || avctx->rc_initial_buffer_occupancy) {
364 av_log( avctx, AV_LOG_ERROR, "Rate control parameters set without a bitrate\n");
365 return AVERROR(EINVAL);
368 dump_enc_cfg(avctx, &enccfg);
370 enccfg.g_w = avctx->width;
371 enccfg.g_h = avctx->height;
372 enccfg.g_timebase.num = avctx->time_base.num;
373 enccfg.g_timebase.den = avctx->time_base.den;
374 enccfg.g_threads = avctx->thread_count;
375 enccfg.g_lag_in_frames= ctx->lag_in_frames;
377 if (avctx->flags & CODEC_FLAG_PASS1)
378 enccfg.g_pass = VPX_RC_FIRST_PASS;
379 else if (avctx->flags & CODEC_FLAG_PASS2)
380 enccfg.g_pass = VPX_RC_LAST_PASS;
382 enccfg.g_pass = VPX_RC_ONE_PASS;
384 if (avctx->rc_min_rate == avctx->rc_max_rate &&
385 avctx->rc_min_rate == avctx->bit_rate && avctx->bit_rate) {
386 enccfg.rc_end_usage = VPX_CBR;
387 } else if (ctx->crf >= 0) {
388 enccfg.rc_end_usage = VPX_CQ;
389 #if CONFIG_LIBVPX_VP9_ENCODER
390 if (!avctx->bit_rate && avctx->codec_id == AV_CODEC_ID_VP9)
391 enccfg.rc_end_usage = VPX_Q;
395 if (avctx->bit_rate) {
396 enccfg.rc_target_bitrate = av_rescale_rnd(avctx->bit_rate, 1, 1000,
398 #if CONFIG_LIBVPX_VP9_ENCODER
399 } else if (enccfg.rc_end_usage == VPX_Q) {
402 if (enccfg.rc_end_usage == VPX_CQ) {
403 enccfg.rc_target_bitrate = 1000000;
405 avctx->bit_rate = enccfg.rc_target_bitrate * 1000;
406 av_log(avctx, AV_LOG_WARNING,
407 "Neither bitrate nor constrained quality specified, using default bitrate of %dkbit/sec\n",
408 enccfg.rc_target_bitrate);
412 if (avctx->codec_id == AV_CODEC_ID_VP9 && ctx->lossless == 1) {
413 enccfg.rc_min_quantizer =
414 enccfg.rc_max_quantizer = 0;
416 if (avctx->qmin >= 0)
417 enccfg.rc_min_quantizer = avctx->qmin;
418 if (avctx->qmax >= 0)
419 enccfg.rc_max_quantizer = avctx->qmax;
422 if (enccfg.rc_end_usage == VPX_CQ
423 #if CONFIG_LIBVPX_VP9_ENCODER
424 || enccfg.rc_end_usage == VPX_Q
427 if (ctx->crf < enccfg.rc_min_quantizer || ctx->crf > enccfg.rc_max_quantizer) {
428 av_log(avctx, AV_LOG_ERROR,
429 "CQ level %d must be between minimum and maximum quantizer value (%d-%d)\n",
430 ctx->crf, enccfg.rc_min_quantizer, enccfg.rc_max_quantizer);
431 return AVERROR(EINVAL);
435 enccfg.rc_dropframe_thresh = avctx->frame_skip_threshold;
437 //0-100 (0 => CBR, 100 => VBR)
438 enccfg.rc_2pass_vbr_bias_pct = round(avctx->qcompress * 100);
440 enccfg.rc_2pass_vbr_minsection_pct =
441 avctx->rc_min_rate * 100LL / avctx->bit_rate;
442 if (avctx->rc_max_rate)
443 enccfg.rc_2pass_vbr_maxsection_pct =
444 avctx->rc_max_rate * 100LL / avctx->bit_rate;
446 if (avctx->rc_buffer_size)
448 avctx->rc_buffer_size * 1000LL / avctx->bit_rate;
449 if (avctx->rc_initial_buffer_occupancy)
450 enccfg.rc_buf_initial_sz =
451 avctx->rc_initial_buffer_occupancy * 1000LL / avctx->bit_rate;
452 enccfg.rc_buf_optimal_sz = enccfg.rc_buf_sz * 5 / 6;
453 enccfg.rc_undershoot_pct = round(avctx->rc_buffer_aggressivity * 100);
455 //_enc_init() will balk if kf_min_dist differs from max w/VPX_KF_AUTO
456 if (avctx->keyint_min >= 0 && avctx->keyint_min == avctx->gop_size)
457 enccfg.kf_min_dist = avctx->keyint_min;
458 if (avctx->gop_size >= 0)
459 enccfg.kf_max_dist = avctx->gop_size;
461 if (enccfg.g_pass == VPX_RC_FIRST_PASS)
462 enccfg.g_lag_in_frames = 0;
463 else if (enccfg.g_pass == VPX_RC_LAST_PASS) {
464 int decode_size, ret;
466 if (!avctx->stats_in) {
467 av_log(avctx, AV_LOG_ERROR, "No stats file for second pass\n");
468 return AVERROR_INVALIDDATA;
471 ctx->twopass_stats.sz = strlen(avctx->stats_in) * 3 / 4;
472 ret = av_reallocp(&ctx->twopass_stats.buf, ctx->twopass_stats.sz);
474 av_log(avctx, AV_LOG_ERROR,
475 "Stat buffer alloc (%"SIZE_SPECIFIER" bytes) failed\n",
476 ctx->twopass_stats.sz);
477 ctx->twopass_stats.sz = 0;
480 decode_size = av_base64_decode(ctx->twopass_stats.buf, avctx->stats_in,
481 ctx->twopass_stats.sz);
482 if (decode_size < 0) {
483 av_log(avctx, AV_LOG_ERROR, "Stat buffer decode failed\n");
484 return AVERROR_INVALIDDATA;
487 ctx->twopass_stats.sz = decode_size;
488 enccfg.rc_twopass_stats_in = ctx->twopass_stats;
491 /* 0-3: For non-zero values the encoder increasingly optimizes for reduced
492 complexity playback on low powered devices at the expense of encode
494 if (avctx->profile != FF_PROFILE_UNKNOWN)
495 enccfg.g_profile = avctx->profile;
497 enccfg.g_error_resilient = ctx->error_resilient || ctx->flags & VP8F_ERROR_RESILIENT;
499 dump_enc_cfg(avctx, &enccfg);
500 /* Construct Encoder Context */
501 res = vpx_codec_enc_init(&ctx->encoder, iface, &enccfg, flags);
502 if (res != VPX_CODEC_OK) {
503 log_encoder_error(avctx, "Failed to initialize encoder");
504 return AVERROR(EINVAL);
508 enccfg_alpha = enccfg;
509 res = vpx_codec_enc_init(&ctx->encoder_alpha, iface, &enccfg_alpha, flags);
510 if (res != VPX_CODEC_OK) {
511 log_encoder_error(avctx, "Failed to initialize alpha encoder");
512 return AVERROR(EINVAL);
516 //codec control failures are currently treated only as warnings
517 av_log(avctx, AV_LOG_DEBUG, "vpx_codec_control\n");
518 codecctl_int(avctx, VP8E_SET_CPUUSED, ctx->cpu_used);
519 if (ctx->flags & VP8F_AUTO_ALT_REF)
520 ctx->auto_alt_ref = 1;
521 if (ctx->auto_alt_ref >= 0)
522 codecctl_int(avctx, VP8E_SET_ENABLEAUTOALTREF, ctx->auto_alt_ref);
523 if (ctx->arnr_max_frames >= 0)
524 codecctl_int(avctx, VP8E_SET_ARNR_MAXFRAMES, ctx->arnr_max_frames);
525 if (ctx->arnr_strength >= 0)
526 codecctl_int(avctx, VP8E_SET_ARNR_STRENGTH, ctx->arnr_strength);
527 if (ctx->arnr_type >= 0)
528 codecctl_int(avctx, VP8E_SET_ARNR_TYPE, ctx->arnr_type);
529 if (avctx->codec_id == AV_CODEC_ID_VP8) {
530 codecctl_int(avctx, VP8E_SET_NOISE_SENSITIVITY, avctx->noise_reduction);
531 codecctl_int(avctx, VP8E_SET_TOKEN_PARTITIONS, av_log2(avctx->slices));
534 FF_DISABLE_DEPRECATION_WARNINGS
535 if (avctx->mb_threshold) {
536 av_log(avctx, AV_LOG_WARNING, "The mb_threshold option is deprecated, "
537 "use the static-thresh private option instead.\n");
538 ctx->static_thresh = avctx->mb_threshold;
540 FF_ENABLE_DEPRECATION_WARNINGS
542 codecctl_int(avctx, VP8E_SET_STATIC_THRESHOLD, ctx->static_thresh);
544 codecctl_int(avctx, VP8E_SET_CQ_LEVEL, ctx->crf);
545 if (ctx->max_intra_rate >= 0)
546 codecctl_int(avctx, VP8E_SET_MAX_INTRA_BITRATE_PCT, ctx->max_intra_rate);
548 #if CONFIG_LIBVPX_VP9_ENCODER
549 if (avctx->codec_id == AV_CODEC_ID_VP9) {
550 if (ctx->lossless >= 0)
551 codecctl_int(avctx, VP9E_SET_LOSSLESS, ctx->lossless);
552 if (ctx->tile_columns >= 0)
553 codecctl_int(avctx, VP9E_SET_TILE_COLUMNS, ctx->tile_columns);
554 if (ctx->tile_rows >= 0)
555 codecctl_int(avctx, VP9E_SET_TILE_ROWS, ctx->tile_rows);
556 if (ctx->frame_parallel >= 0)
557 codecctl_int(avctx, VP9E_SET_FRAME_PARALLEL_DECODING, ctx->frame_parallel);
558 if (ctx->aq_mode >= 0)
559 codecctl_int(avctx, VP9E_SET_AQ_MODE, ctx->aq_mode);
563 av_log(avctx, AV_LOG_DEBUG, "Using deadline: %d\n", ctx->deadline);
565 //provide dummy value to initialize wrapper, values will be updated each _encode()
566 vpx_img_wrap(&ctx->rawimg, img_fmt, avctx->width, avctx->height, 1,
568 #if CONFIG_LIBVPX_VP9_ENCODER && defined(VPX_IMG_FMT_HIGHBITDEPTH)
569 if (avctx->codec_id == AV_CODEC_ID_VP9 && (codec_caps & VPX_CODEC_CAP_HIGHBITDEPTH))
570 ctx->rawimg.bit_depth = enccfg.g_bit_depth;
574 vpx_img_wrap(&ctx->rawimg_alpha, VPX_IMG_FMT_I420, avctx->width, avctx->height, 1,
577 avctx->coded_frame = av_frame_alloc();
578 if (!avctx->coded_frame) {
579 av_log(avctx, AV_LOG_ERROR, "Error allocating coded frame\n");
581 return AVERROR(ENOMEM);
586 static inline void cx_pktcpy(struct FrameListData *dst,
587 const struct vpx_codec_cx_pkt *src,
588 const struct vpx_codec_cx_pkt *src_alpha,
591 dst->pts = src->data.frame.pts;
592 dst->duration = src->data.frame.duration;
593 dst->flags = src->data.frame.flags;
594 dst->sz = src->data.frame.sz;
595 dst->buf = src->data.frame.buf;
597 /* For alt-ref frame, don't store PSNR or increment frame_number */
598 if (!(dst->flags & VPX_FRAME_IS_INVISIBLE)) {
599 dst->frame_number = ++ctx->frame_number;
600 dst->have_sse = ctx->have_sse;
602 /* associate last-seen SSE to the frame. */
603 /* Transfers ownership from ctx to dst. */
604 /* WARNING! This makes the assumption that PSNR_PKT comes
605 just before the frame it refers to! */
606 memcpy(dst->sse, ctx->sse, sizeof(dst->sse));
610 dst->frame_number = -1; /* sanity marker */
613 dst->buf_alpha = src_alpha->data.frame.buf;
614 dst->sz_alpha = src_alpha->data.frame.sz;
616 dst->buf_alpha = NULL;
622 * Store coded frame information in format suitable for return from encode2().
624 * Write information from @a cx_frame to @a pkt
625 * @return packet data size on success
626 * @return a negative AVERROR on error
628 static int storeframe(AVCodecContext *avctx, struct FrameListData *cx_frame,
629 AVPacket *pkt, AVFrame *coded_frame)
631 int ret = ff_alloc_packet2(avctx, pkt, cx_frame->sz);
634 memcpy(pkt->data, cx_frame->buf, pkt->size);
635 pkt->pts = pkt->dts = cx_frame->pts;
636 coded_frame->pts = cx_frame->pts;
637 coded_frame->key_frame = !!(cx_frame->flags & VPX_FRAME_IS_KEY);
639 if (coded_frame->key_frame) {
640 coded_frame->pict_type = AV_PICTURE_TYPE_I;
641 pkt->flags |= AV_PKT_FLAG_KEY;
643 coded_frame->pict_type = AV_PICTURE_TYPE_P;
645 if (cx_frame->have_sse) {
647 /* Beware of the Y/U/V/all order! */
648 coded_frame->error[0] = cx_frame->sse[1];
649 coded_frame->error[1] = cx_frame->sse[2];
650 coded_frame->error[2] = cx_frame->sse[3];
651 coded_frame->error[3] = 0; // alpha
652 for (i = 0; i < 4; ++i) {
653 avctx->error[i] += coded_frame->error[i];
655 cx_frame->have_sse = 0;
657 if (cx_frame->sz_alpha > 0) {
658 side_data = av_packet_new_side_data(pkt,
659 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
660 cx_frame->sz_alpha + 8);
664 return AVERROR(ENOMEM);
666 AV_WB64(side_data, 1);
667 memcpy(side_data + 8, cx_frame->buf_alpha, cx_frame->sz_alpha);
676 * Queue multiple output frames from the encoder, returning the front-most.
677 * In cases where vpx_codec_get_cx_data() returns more than 1 frame append
678 * the frame queue. Return the head frame if available.
679 * @return Stored frame size
680 * @return AVERROR(EINVAL) on output size error
681 * @return AVERROR(ENOMEM) on coded frame queue data allocation error
683 static int queue_frames(AVCodecContext *avctx, AVPacket *pkt_out,
684 AVFrame *coded_frame)
686 VP8Context *ctx = avctx->priv_data;
687 const struct vpx_codec_cx_pkt *pkt;
688 const struct vpx_codec_cx_pkt *pkt_alpha = NULL;
689 const void *iter = NULL;
690 const void *iter_alpha = NULL;
693 if (ctx->coded_frame_list) {
694 struct FrameListData *cx_frame = ctx->coded_frame_list;
695 /* return the leading frame if we've already begun queueing */
696 size = storeframe(avctx, cx_frame, pkt_out, coded_frame);
699 ctx->coded_frame_list = cx_frame->next;
700 free_coded_frame(cx_frame);
703 /* consume all available output from the encoder before returning. buffers
704 are only good through the next vpx_codec call */
705 while ((pkt = vpx_codec_get_cx_data(&ctx->encoder, &iter)) &&
707 (ctx->is_alpha && (pkt_alpha = vpx_codec_get_cx_data(&ctx->encoder_alpha, &iter_alpha))))) {
709 case VPX_CODEC_CX_FRAME_PKT:
711 struct FrameListData cx_frame;
713 /* avoid storing the frame when the list is empty and we haven't yet
714 provided a frame for output */
715 av_assert0(!ctx->coded_frame_list);
716 cx_pktcpy(&cx_frame, pkt, pkt_alpha, ctx);
717 size = storeframe(avctx, &cx_frame, pkt_out, coded_frame);
721 struct FrameListData *cx_frame =
722 av_malloc(sizeof(struct FrameListData));
725 av_log(avctx, AV_LOG_ERROR,
726 "Frame queue element alloc failed\n");
727 return AVERROR(ENOMEM);
729 cx_pktcpy(cx_frame, pkt, pkt_alpha, ctx);
730 cx_frame->buf = av_malloc(cx_frame->sz);
732 if (!cx_frame->buf) {
733 av_log(avctx, AV_LOG_ERROR,
734 "Data buffer alloc (%"SIZE_SPECIFIER" bytes) failed\n",
737 return AVERROR(ENOMEM);
739 memcpy(cx_frame->buf, pkt->data.frame.buf, pkt->data.frame.sz);
741 cx_frame->buf_alpha = av_malloc(cx_frame->sz_alpha);
742 if (!cx_frame->buf_alpha) {
743 av_log(avctx, AV_LOG_ERROR,
744 "Data buffer alloc (%"SIZE_SPECIFIER" bytes) failed\n",
747 return AVERROR(ENOMEM);
749 memcpy(cx_frame->buf_alpha, pkt_alpha->data.frame.buf, pkt_alpha->data.frame.sz);
751 coded_frame_add(&ctx->coded_frame_list, cx_frame);
754 case VPX_CODEC_STATS_PKT: {
755 struct vpx_fixed_buf *stats = &ctx->twopass_stats;
757 if ((err = av_reallocp(&stats->buf,
759 pkt->data.twopass_stats.sz)) < 0) {
761 av_log(avctx, AV_LOG_ERROR, "Stat buffer realloc failed\n");
764 memcpy((uint8_t*)stats->buf + stats->sz,
765 pkt->data.twopass_stats.buf, pkt->data.twopass_stats.sz);
766 stats->sz += pkt->data.twopass_stats.sz;
769 case VPX_CODEC_PSNR_PKT:
770 av_assert0(!ctx->have_sse);
771 ctx->sse[0] = pkt->data.psnr.sse[0];
772 ctx->sse[1] = pkt->data.psnr.sse[1];
773 ctx->sse[2] = pkt->data.psnr.sse[2];
774 ctx->sse[3] = pkt->data.psnr.sse[3];
777 case VPX_CODEC_CUSTOM_PKT:
778 //ignore unsupported/unrecognized packet types
786 static int vp8_encode(AVCodecContext *avctx, AVPacket *pkt,
787 const AVFrame *frame, int *got_packet)
789 VP8Context *ctx = avctx->priv_data;
790 struct vpx_image *rawimg = NULL;
791 struct vpx_image *rawimg_alpha = NULL;
792 int64_t timestamp = 0;
794 vpx_enc_frame_flags_t flags = 0;
797 rawimg = &ctx->rawimg;
798 rawimg->planes[VPX_PLANE_Y] = frame->data[0];
799 rawimg->planes[VPX_PLANE_U] = frame->data[1];
800 rawimg->planes[VPX_PLANE_V] = frame->data[2];
801 rawimg->stride[VPX_PLANE_Y] = frame->linesize[0];
802 rawimg->stride[VPX_PLANE_U] = frame->linesize[1];
803 rawimg->stride[VPX_PLANE_V] = frame->linesize[2];
805 uint8_t *u_plane, *v_plane;
806 rawimg_alpha = &ctx->rawimg_alpha;
807 rawimg_alpha->planes[VPX_PLANE_Y] = frame->data[3];
808 u_plane = av_malloc(frame->linesize[1] * frame->height);
809 v_plane = av_malloc(frame->linesize[2] * frame->height);
810 if (!u_plane || !v_plane) {
813 return AVERROR(ENOMEM);
815 memset(u_plane, 0x80, frame->linesize[1] * frame->height);
816 rawimg_alpha->planes[VPX_PLANE_U] = u_plane;
817 memset(v_plane, 0x80, frame->linesize[2] * frame->height);
818 rawimg_alpha->planes[VPX_PLANE_V] = v_plane;
819 rawimg_alpha->stride[VPX_PLANE_Y] = frame->linesize[0];
820 rawimg_alpha->stride[VPX_PLANE_U] = frame->linesize[1];
821 rawimg_alpha->stride[VPX_PLANE_V] = frame->linesize[2];
823 timestamp = frame->pts;
824 if (frame->pict_type == AV_PICTURE_TYPE_I)
825 flags |= VPX_EFLAG_FORCE_KF;
828 res = vpx_codec_encode(&ctx->encoder, rawimg, timestamp,
829 avctx->ticks_per_frame, flags, ctx->deadline);
830 if (res != VPX_CODEC_OK) {
831 log_encoder_error(avctx, "Error encoding frame");
832 return AVERROR_INVALIDDATA;
836 res = vpx_codec_encode(&ctx->encoder_alpha, rawimg_alpha, timestamp,
837 avctx->ticks_per_frame, flags, ctx->deadline);
838 if (res != VPX_CODEC_OK) {
839 log_encoder_error(avctx, "Error encoding alpha frame");
840 return AVERROR_INVALIDDATA;
844 coded_size = queue_frames(avctx, pkt, avctx->coded_frame);
846 if (!frame && avctx->flags & CODEC_FLAG_PASS1) {
847 unsigned int b64_size = AV_BASE64_SIZE(ctx->twopass_stats.sz);
849 avctx->stats_out = av_malloc(b64_size);
850 if (!avctx->stats_out) {
851 av_log(avctx, AV_LOG_ERROR, "Stat buffer alloc (%d bytes) failed\n",
853 return AVERROR(ENOMEM);
855 av_base64_encode(avctx->stats_out, b64_size, ctx->twopass_stats.buf,
856 ctx->twopass_stats.sz);
860 av_freep(&rawimg_alpha->planes[VPX_PLANE_U]);
861 av_freep(&rawimg_alpha->planes[VPX_PLANE_V]);
864 *got_packet = !!coded_size;
868 #define OFFSET(x) offsetof(VP8Context, x)
869 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
871 #ifndef VPX_ERROR_RESILIENT_DEFAULT
872 #define VPX_ERROR_RESILIENT_DEFAULT 1
873 #define VPX_ERROR_RESILIENT_PARTITIONS 2
876 #define COMMON_OPTIONS \
877 { "cpu-used", "Quality/Speed ratio modifier", OFFSET(cpu_used), AV_OPT_TYPE_INT, {.i64 = 1}, -16, 16, VE}, \
878 { "auto-alt-ref", "Enable use of alternate reference " \
879 "frames (2-pass only)", OFFSET(auto_alt_ref), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE}, \
880 { "lag-in-frames", "Number of frames to look ahead for " \
881 "alternate reference frame selection", OFFSET(lag_in_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
882 { "arnr-maxframes", "altref noise reduction max frame count", OFFSET(arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
883 { "arnr-strength", "altref noise reduction filter strength", OFFSET(arnr_strength), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
884 { "arnr-type", "altref noise reduction filter type", OFFSET(arnr_type), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE, "arnr_type"}, \
885 { "backward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0, VE, "arnr_type" }, \
886 { "forward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0, VE, "arnr_type" }, \
887 { "centered", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 3}, 0, 0, VE, "arnr_type" }, \
888 { "deadline", "Time to spend encoding, in microseconds.", OFFSET(deadline), AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, "quality"}, \
889 { "best", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_BEST_QUALITY}, 0, 0, VE, "quality"}, \
890 { "good", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_GOOD_QUALITY}, 0, 0, VE, "quality"}, \
891 { "realtime", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_REALTIME}, 0, 0, VE, "quality"}, \
892 { "error-resilient", "Error resilience configuration", OFFSET(error_resilient), AV_OPT_TYPE_FLAGS, {.i64 = 0}, INT_MIN, INT_MAX, VE, "er"}, \
893 { "max-intra-rate", "Maximum I-frame bitrate (pct) 0=unlimited", OFFSET(max_intra_rate), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
894 { "default", "Improve resiliency against losses of whole frames", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_DEFAULT}, 0, 0, VE, "er"}, \
895 { "partitions", "The frame partitions are independently decodable " \
896 "by the bool decoder, meaning that partitions can be decoded even " \
897 "though earlier partitions have been lost. Note that intra predicition" \
898 " is still done over the partition boundary.", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_PARTITIONS}, 0, 0, VE, "er"}, \
899 { "crf", "Select the quality for constant quality mode", offsetof(VP8Context, crf), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 63, VE }, \
900 { "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 }, \
902 #define LEGACY_OPTIONS \
903 {"speed", "", offsetof(VP8Context, cpu_used), AV_OPT_TYPE_INT, {.i64 = 1}, -16, 16, VE}, \
904 {"quality", "", offsetof(VP8Context, deadline), AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, "quality"}, \
905 {"vp8flags", "", offsetof(VP8Context, flags), FF_OPT_TYPE_FLAGS, {.i64 = 0}, 0, UINT_MAX, VE, "flags"}, \
906 {"error_resilient", "enable error resilience", 0, FF_OPT_TYPE_CONST, {.dbl = VP8F_ERROR_RESILIENT}, INT_MIN, INT_MAX, VE, "flags"}, \
907 {"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"}, \
908 {"arnr_max_frames", "altref noise reduction max frame count", offsetof(VP8Context, arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 15, VE}, \
909 {"arnr_strength", "altref noise reduction filter strength", offsetof(VP8Context, arnr_strength), AV_OPT_TYPE_INT, {.i64 = 3}, 0, 6, VE}, \
910 {"arnr_type", "altref noise reduction filter type", offsetof(VP8Context, arnr_type), AV_OPT_TYPE_INT, {.i64 = 3}, 1, 3, VE}, \
911 {"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}, \
913 #if CONFIG_LIBVPX_VP8_ENCODER
914 static const AVOption vp8_options[] = {
921 #if CONFIG_LIBVPX_VP9_ENCODER
922 static const AVOption vp9_options[] = {
924 { "lossless", "Lossless mode", OFFSET(lossless), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE},
925 { "tile-columns", "Number of tile columns to use, log2", OFFSET(tile_columns), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 6, VE},
926 { "tile-rows", "Number of tile rows to use, log2", OFFSET(tile_rows), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 2, VE},
927 { "frame-parallel", "Enable frame parallel decodability features", OFFSET(frame_parallel), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE},
928 { "aq-mode", "adaptive quantization mode", OFFSET(aq_mode), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 3, VE, "aq_mode"},
929 { "none", "Aq not used", 0, AV_OPT_TYPE_CONST, {.i64 = 0}, 0, 0, VE, "aq_mode" }, \
930 { "variance", "Variance based Aq", 0, AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0, VE, "aq_mode" }, \
931 { "complexity", "Complexity based Aq", 0, AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0, VE, "aq_mode" }, \
932 { "cyclic", "Cyclic Refresh Aq", 0, AV_OPT_TYPE_CONST, {.i64 = 3}, 0, 0, VE, "aq_mode" }, \
938 #undef COMMON_OPTIONS
939 #undef LEGACY_OPTIONS
941 static const AVCodecDefault defaults[] = {
945 { "keyint_min", "-1" },
949 #if CONFIG_LIBVPX_VP8_ENCODER
950 static av_cold int vp8_init(AVCodecContext *avctx)
952 return vpx_init(avctx, vpx_codec_vp8_cx());
955 static const AVClass class_vp8 = {
956 .class_name = "libvpx-vp8 encoder",
957 .item_name = av_default_item_name,
958 .option = vp8_options,
959 .version = LIBAVUTIL_VERSION_INT,
962 AVCodec ff_libvpx_vp8_encoder = {
964 .long_name = NULL_IF_CONFIG_SMALL("libvpx VP8"),
965 .type = AVMEDIA_TYPE_VIDEO,
966 .id = AV_CODEC_ID_VP8,
967 .priv_data_size = sizeof(VP8Context),
969 .encode2 = vp8_encode,
971 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
972 .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUVA420P, AV_PIX_FMT_NONE },
973 .priv_class = &class_vp8,
974 .defaults = defaults,
976 #endif /* CONFIG_LIBVPX_VP8_ENCODER */
978 #if CONFIG_LIBVPX_VP9_ENCODER
979 static av_cold int vp9_init(AVCodecContext *avctx)
981 return vpx_init(avctx, vpx_codec_vp9_cx());
984 static const AVClass class_vp9 = {
985 .class_name = "libvpx-vp9 encoder",
986 .item_name = av_default_item_name,
987 .option = vp9_options,
988 .version = LIBAVUTIL_VERSION_INT,
991 AVCodec ff_libvpx_vp9_encoder = {
992 .name = "libvpx-vp9",
993 .long_name = NULL_IF_CONFIG_SMALL("libvpx VP9"),
994 .type = AVMEDIA_TYPE_VIDEO,
995 .id = AV_CODEC_ID_VP9,
996 .priv_data_size = sizeof(VP8Context),
998 .encode2 = vp8_encode,
1000 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
1001 .priv_class = &class_vp9,
1002 .defaults = defaults,
1003 .init_static_data = ff_vp9_init_static,
1005 #endif /* CONFIG_LIBVPX_VP9_ENCODER */