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 " %*s{%u/%u}\n %*s%u\n %*s%d\n %*s%u\n",
153 width, "g_usage:", cfg->g_usage,
154 width, "g_threads:", cfg->g_threads,
155 width, "g_profile:", cfg->g_profile,
156 width, "g_w:", cfg->g_w,
157 width, "g_h:", cfg->g_h,
158 width, "g_timebase:", cfg->g_timebase.num, cfg->g_timebase.den,
159 width, "g_error_resilient:", cfg->g_error_resilient,
160 width, "g_pass:", cfg->g_pass,
161 width, "g_lag_in_frames:", cfg->g_lag_in_frames);
162 av_log(avctx, level, "rate control settings\n"
163 " %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
164 " %*s%d\n %*s%p(%"SIZE_SPECIFIER")\n %*s%u\n",
165 width, "rc_dropframe_thresh:", cfg->rc_dropframe_thresh,
166 width, "rc_resize_allowed:", cfg->rc_resize_allowed,
167 width, "rc_resize_up_thresh:", cfg->rc_resize_up_thresh,
168 width, "rc_resize_down_thresh:", cfg->rc_resize_down_thresh,
169 width, "rc_end_usage:", cfg->rc_end_usage,
170 width, "rc_twopass_stats_in:", cfg->rc_twopass_stats_in.buf, cfg->rc_twopass_stats_in.sz,
171 width, "rc_target_bitrate:", cfg->rc_target_bitrate);
172 av_log(avctx, level, "quantizer settings\n"
174 width, "rc_min_quantizer:", cfg->rc_min_quantizer,
175 width, "rc_max_quantizer:", cfg->rc_max_quantizer);
176 av_log(avctx, level, "bitrate tolerance\n"
178 width, "rc_undershoot_pct:", cfg->rc_undershoot_pct,
179 width, "rc_overshoot_pct:", cfg->rc_overshoot_pct);
180 av_log(avctx, level, "decoder buffer model\n"
181 " %*s%u\n %*s%u\n %*s%u\n",
182 width, "rc_buf_sz:", cfg->rc_buf_sz,
183 width, "rc_buf_initial_sz:", cfg->rc_buf_initial_sz,
184 width, "rc_buf_optimal_sz:", cfg->rc_buf_optimal_sz);
185 av_log(avctx, level, "2 pass rate control settings\n"
186 " %*s%u\n %*s%u\n %*s%u\n",
187 width, "rc_2pass_vbr_bias_pct:", cfg->rc_2pass_vbr_bias_pct,
188 width, "rc_2pass_vbr_minsection_pct:", cfg->rc_2pass_vbr_minsection_pct,
189 width, "rc_2pass_vbr_maxsection_pct:", cfg->rc_2pass_vbr_maxsection_pct);
190 av_log(avctx, level, "keyframing settings\n"
191 " %*s%d\n %*s%u\n %*s%u\n",
192 width, "kf_mode:", cfg->kf_mode,
193 width, "kf_min_dist:", cfg->kf_min_dist,
194 width, "kf_max_dist:", cfg->kf_max_dist);
195 av_log(avctx, level, "\n");
198 static void coded_frame_add(void *list, struct FrameListData *cx_frame)
200 struct FrameListData **p = list;
205 cx_frame->next = NULL;
208 static av_cold void free_coded_frame(struct FrameListData *cx_frame)
210 av_freep(&cx_frame->buf);
211 if (cx_frame->buf_alpha)
212 av_freep(&cx_frame->buf_alpha);
216 static av_cold void free_frame_list(struct FrameListData *list)
218 struct FrameListData *p = list;
227 static av_cold int codecctl_int(AVCodecContext *avctx,
228 enum vp8e_enc_control_id id, int val)
230 VP8Context *ctx = avctx->priv_data;
235 snprintf(buf, sizeof(buf), "%s:", ctlidstr[id]);
236 av_log(avctx, AV_LOG_DEBUG, " %*s%d\n", width, buf, val);
238 res = vpx_codec_control(&ctx->encoder, id, val);
239 if (res != VPX_CODEC_OK) {
240 snprintf(buf, sizeof(buf), "Failed to set %s codec control",
242 log_encoder_error(avctx, buf);
245 return res == VPX_CODEC_OK ? 0 : AVERROR(EINVAL);
248 static av_cold int vp8_free(AVCodecContext *avctx)
250 VP8Context *ctx = avctx->priv_data;
252 vpx_codec_destroy(&ctx->encoder);
254 vpx_codec_destroy(&ctx->encoder_alpha);
255 av_freep(&ctx->twopass_stats.buf);
256 av_frame_free(&avctx->coded_frame);
257 av_freep(&avctx->stats_out);
258 free_frame_list(ctx->coded_frame_list);
262 static av_cold int vpx_init(AVCodecContext *avctx,
263 const struct vpx_codec_iface *iface)
265 VP8Context *ctx = avctx->priv_data;
266 struct vpx_codec_enc_cfg enccfg;
267 struct vpx_codec_enc_cfg enccfg_alpha;
268 vpx_codec_flags_t flags = (avctx->flags & CODEC_FLAG_PSNR) ? VPX_CODEC_USE_PSNR : 0;
271 av_log(avctx, AV_LOG_INFO, "%s\n", vpx_codec_version_str());
272 av_log(avctx, AV_LOG_VERBOSE, "%s\n", vpx_codec_build_config());
274 if (avctx->pix_fmt == AV_PIX_FMT_YUVA420P)
277 if ((res = vpx_codec_enc_config_default(iface, &enccfg, 0)) != VPX_CODEC_OK) {
278 av_log(avctx, AV_LOG_ERROR, "Failed to get config: %s\n",
279 vpx_codec_err_to_string(res));
280 return AVERROR(EINVAL);
284 if(avctx->rc_max_rate || avctx->rc_buffer_size || avctx->rc_initial_buffer_occupancy) {
285 av_log( avctx, AV_LOG_ERROR, "Rate control parameters set without a bitrate\n");
286 return AVERROR(EINVAL);
289 dump_enc_cfg(avctx, &enccfg);
291 enccfg.g_w = avctx->width;
292 enccfg.g_h = avctx->height;
293 enccfg.g_timebase.num = avctx->time_base.num;
294 enccfg.g_timebase.den = avctx->time_base.den;
295 enccfg.g_threads = avctx->thread_count;
296 enccfg.g_lag_in_frames= ctx->lag_in_frames;
298 if (avctx->flags & CODEC_FLAG_PASS1)
299 enccfg.g_pass = VPX_RC_FIRST_PASS;
300 else if (avctx->flags & CODEC_FLAG_PASS2)
301 enccfg.g_pass = VPX_RC_LAST_PASS;
303 enccfg.g_pass = VPX_RC_ONE_PASS;
305 if (avctx->rc_min_rate == avctx->rc_max_rate &&
306 avctx->rc_min_rate == avctx->bit_rate && avctx->bit_rate) {
307 enccfg.rc_end_usage = VPX_CBR;
308 } else if (ctx->crf >= 0) {
309 enccfg.rc_end_usage = VPX_CQ;
310 #if CONFIG_LIBVPX_VP9_ENCODER
311 if (!avctx->bit_rate && avctx->codec_id == AV_CODEC_ID_VP9)
312 enccfg.rc_end_usage = VPX_Q;
316 if (avctx->bit_rate) {
317 enccfg.rc_target_bitrate = av_rescale_rnd(avctx->bit_rate, 1, 1000,
319 #if CONFIG_LIBVPX_VP9_ENCODER
320 } else if (enccfg.rc_end_usage == VPX_Q) {
323 if (enccfg.rc_end_usage == VPX_CQ) {
324 enccfg.rc_target_bitrate = 1000000;
326 avctx->bit_rate = enccfg.rc_target_bitrate * 1000;
327 av_log(avctx, AV_LOG_WARNING,
328 "Neither bitrate nor constrained quality specified, using default bitrate of %dkbit/sec\n",
329 enccfg.rc_target_bitrate);
333 if (avctx->codec_id == AV_CODEC_ID_VP9 && ctx->lossless == 1) {
334 enccfg.rc_min_quantizer =
335 enccfg.rc_max_quantizer = 0;
337 if (avctx->qmin >= 0)
338 enccfg.rc_min_quantizer = avctx->qmin;
339 if (avctx->qmax >= 0)
340 enccfg.rc_max_quantizer = avctx->qmax;
343 if (enccfg.rc_end_usage == VPX_CQ
344 #if CONFIG_LIBVPX_VP9_ENCODER
345 || enccfg.rc_end_usage == VPX_Q
348 if (ctx->crf < enccfg.rc_min_quantizer || ctx->crf > enccfg.rc_max_quantizer) {
349 av_log(avctx, AV_LOG_ERROR,
350 "CQ level %d must be between minimum and maximum quantizer value (%d-%d)\n",
351 ctx->crf, enccfg.rc_min_quantizer, enccfg.rc_max_quantizer);
352 return AVERROR(EINVAL);
356 enccfg.rc_dropframe_thresh = avctx->frame_skip_threshold;
358 //0-100 (0 => CBR, 100 => VBR)
359 enccfg.rc_2pass_vbr_bias_pct = round(avctx->qcompress * 100);
361 enccfg.rc_2pass_vbr_minsection_pct =
362 avctx->rc_min_rate * 100LL / avctx->bit_rate;
363 if (avctx->rc_max_rate)
364 enccfg.rc_2pass_vbr_maxsection_pct =
365 avctx->rc_max_rate * 100LL / avctx->bit_rate;
367 if (avctx->rc_buffer_size)
369 avctx->rc_buffer_size * 1000LL / avctx->bit_rate;
370 if (avctx->rc_initial_buffer_occupancy)
371 enccfg.rc_buf_initial_sz =
372 avctx->rc_initial_buffer_occupancy * 1000LL / avctx->bit_rate;
373 enccfg.rc_buf_optimal_sz = enccfg.rc_buf_sz * 5 / 6;
374 enccfg.rc_undershoot_pct = round(avctx->rc_buffer_aggressivity * 100);
376 //_enc_init() will balk if kf_min_dist differs from max w/VPX_KF_AUTO
377 if (avctx->keyint_min >= 0 && avctx->keyint_min == avctx->gop_size)
378 enccfg.kf_min_dist = avctx->keyint_min;
379 if (avctx->gop_size >= 0)
380 enccfg.kf_max_dist = avctx->gop_size;
382 if (enccfg.g_pass == VPX_RC_FIRST_PASS)
383 enccfg.g_lag_in_frames = 0;
384 else if (enccfg.g_pass == VPX_RC_LAST_PASS) {
385 int decode_size, ret;
387 if (!avctx->stats_in) {
388 av_log(avctx, AV_LOG_ERROR, "No stats file for second pass\n");
389 return AVERROR_INVALIDDATA;
392 ctx->twopass_stats.sz = strlen(avctx->stats_in) * 3 / 4;
393 ret = av_reallocp(&ctx->twopass_stats.buf, ctx->twopass_stats.sz);
395 av_log(avctx, AV_LOG_ERROR,
396 "Stat buffer alloc (%"SIZE_SPECIFIER" bytes) failed\n",
397 ctx->twopass_stats.sz);
398 ctx->twopass_stats.sz = 0;
401 decode_size = av_base64_decode(ctx->twopass_stats.buf, avctx->stats_in,
402 ctx->twopass_stats.sz);
403 if (decode_size < 0) {
404 av_log(avctx, AV_LOG_ERROR, "Stat buffer decode failed\n");
405 return AVERROR_INVALIDDATA;
408 ctx->twopass_stats.sz = decode_size;
409 enccfg.rc_twopass_stats_in = ctx->twopass_stats;
412 /* 0-3: For non-zero values the encoder increasingly optimizes for reduced
413 complexity playback on low powered devices at the expense of encode
415 if (avctx->profile != FF_PROFILE_UNKNOWN)
416 enccfg.g_profile = avctx->profile;
418 enccfg.g_error_resilient = ctx->error_resilient || ctx->flags & VP8F_ERROR_RESILIENT;
420 dump_enc_cfg(avctx, &enccfg);
421 /* Construct Encoder Context */
422 res = vpx_codec_enc_init(&ctx->encoder, iface, &enccfg, flags);
423 if (res != VPX_CODEC_OK) {
424 log_encoder_error(avctx, "Failed to initialize encoder");
425 return AVERROR(EINVAL);
429 enccfg_alpha = enccfg;
430 res = vpx_codec_enc_init(&ctx->encoder_alpha, iface, &enccfg_alpha, flags);
431 if (res != VPX_CODEC_OK) {
432 log_encoder_error(avctx, "Failed to initialize alpha encoder");
433 return AVERROR(EINVAL);
437 //codec control failures are currently treated only as warnings
438 av_log(avctx, AV_LOG_DEBUG, "vpx_codec_control\n");
439 codecctl_int(avctx, VP8E_SET_CPUUSED, ctx->cpu_used);
440 if (ctx->flags & VP8F_AUTO_ALT_REF)
441 ctx->auto_alt_ref = 1;
442 if (ctx->auto_alt_ref >= 0)
443 codecctl_int(avctx, VP8E_SET_ENABLEAUTOALTREF, ctx->auto_alt_ref);
444 if (ctx->arnr_max_frames >= 0)
445 codecctl_int(avctx, VP8E_SET_ARNR_MAXFRAMES, ctx->arnr_max_frames);
446 if (ctx->arnr_strength >= 0)
447 codecctl_int(avctx, VP8E_SET_ARNR_STRENGTH, ctx->arnr_strength);
448 if (ctx->arnr_type >= 0)
449 codecctl_int(avctx, VP8E_SET_ARNR_TYPE, ctx->arnr_type);
450 codecctl_int(avctx, VP8E_SET_NOISE_SENSITIVITY, avctx->noise_reduction);
451 if (avctx->codec_id == AV_CODEC_ID_VP8)
452 codecctl_int(avctx, VP8E_SET_TOKEN_PARTITIONS, av_log2(avctx->slices));
454 FF_DISABLE_DEPRECATION_WARNINGS
455 if (avctx->mb_threshold) {
456 av_log(avctx, AV_LOG_WARNING, "The mb_threshold option is deprecated, "
457 "use the static-thresh private option instead.\n");
458 ctx->static_thresh = avctx->mb_threshold;
460 FF_ENABLE_DEPRECATION_WARNINGS
462 codecctl_int(avctx, VP8E_SET_STATIC_THRESHOLD, ctx->static_thresh);
464 codecctl_int(avctx, VP8E_SET_CQ_LEVEL, ctx->crf);
465 if (ctx->max_intra_rate >= 0)
466 codecctl_int(avctx, VP8E_SET_MAX_INTRA_BITRATE_PCT, ctx->max_intra_rate);
468 #if CONFIG_LIBVPX_VP9_ENCODER
469 if (avctx->codec_id == AV_CODEC_ID_VP9) {
470 if (ctx->lossless >= 0)
471 codecctl_int(avctx, VP9E_SET_LOSSLESS, ctx->lossless);
472 if (ctx->tile_columns >= 0)
473 codecctl_int(avctx, VP9E_SET_TILE_COLUMNS, ctx->tile_columns);
474 if (ctx->tile_rows >= 0)
475 codecctl_int(avctx, VP9E_SET_TILE_ROWS, ctx->tile_rows);
476 if (ctx->frame_parallel >= 0)
477 codecctl_int(avctx, VP9E_SET_FRAME_PARALLEL_DECODING, ctx->frame_parallel);
478 if (ctx->aq_mode >= 0)
479 codecctl_int(avctx, VP9E_SET_AQ_MODE, ctx->aq_mode);
483 av_log(avctx, AV_LOG_DEBUG, "Using deadline: %d\n", ctx->deadline);
485 //provide dummy value to initialize wrapper, values will be updated each _encode()
486 vpx_img_wrap(&ctx->rawimg, VPX_IMG_FMT_I420, avctx->width, avctx->height, 1,
490 vpx_img_wrap(&ctx->rawimg_alpha, VPX_IMG_FMT_I420, avctx->width, avctx->height, 1,
493 avctx->coded_frame = av_frame_alloc();
494 if (!avctx->coded_frame) {
495 av_log(avctx, AV_LOG_ERROR, "Error allocating coded frame\n");
497 return AVERROR(ENOMEM);
502 static inline void cx_pktcpy(struct FrameListData *dst,
503 const struct vpx_codec_cx_pkt *src,
504 const struct vpx_codec_cx_pkt *src_alpha,
507 dst->pts = src->data.frame.pts;
508 dst->duration = src->data.frame.duration;
509 dst->flags = src->data.frame.flags;
510 dst->sz = src->data.frame.sz;
511 dst->buf = src->data.frame.buf;
513 /* For alt-ref frame, don't store PSNR or increment frame_number */
514 if (!(dst->flags & VPX_FRAME_IS_INVISIBLE)) {
515 dst->frame_number = ++ctx->frame_number;
516 dst->have_sse = ctx->have_sse;
518 /* associate last-seen SSE to the frame. */
519 /* Transfers ownership from ctx to dst. */
520 /* WARNING! This makes the assumption that PSNR_PKT comes
521 just before the frame it refers to! */
522 memcpy(dst->sse, ctx->sse, sizeof(dst->sse));
526 dst->frame_number = -1; /* sanity marker */
529 dst->buf_alpha = src_alpha->data.frame.buf;
530 dst->sz_alpha = src_alpha->data.frame.sz;
533 dst->buf_alpha = NULL;
539 * Store coded frame information in format suitable for return from encode2().
541 * Write information from @a cx_frame to @a pkt
542 * @return packet data size on success
543 * @return a negative AVERROR on error
545 static int storeframe(AVCodecContext *avctx, struct FrameListData *cx_frame,
546 AVPacket *pkt, AVFrame *coded_frame)
548 int ret = ff_alloc_packet2(avctx, pkt, cx_frame->sz);
551 memcpy(pkt->data, cx_frame->buf, pkt->size);
552 pkt->pts = pkt->dts = cx_frame->pts;
553 coded_frame->pts = cx_frame->pts;
554 coded_frame->key_frame = !!(cx_frame->flags & VPX_FRAME_IS_KEY);
556 if (coded_frame->key_frame) {
557 coded_frame->pict_type = AV_PICTURE_TYPE_I;
558 pkt->flags |= AV_PKT_FLAG_KEY;
560 coded_frame->pict_type = AV_PICTURE_TYPE_P;
562 if (cx_frame->have_sse) {
564 /* Beware of the Y/U/V/all order! */
565 coded_frame->error[0] = cx_frame->sse[1];
566 coded_frame->error[1] = cx_frame->sse[2];
567 coded_frame->error[2] = cx_frame->sse[3];
568 coded_frame->error[3] = 0; // alpha
569 for (i = 0; i < 4; ++i) {
570 avctx->error[i] += coded_frame->error[i];
572 cx_frame->have_sse = 0;
574 if (cx_frame->sz_alpha > 0) {
575 side_data = av_packet_new_side_data(pkt,
576 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
577 cx_frame->sz_alpha + 8);
581 return AVERROR(ENOMEM);
583 AV_WB64(side_data, 1);
584 memcpy(side_data + 8, cx_frame->buf_alpha, cx_frame->sz_alpha);
593 * Queue multiple output frames from the encoder, returning the front-most.
594 * In cases where vpx_codec_get_cx_data() returns more than 1 frame append
595 * the frame queue. Return the head frame if available.
596 * @return Stored frame size
597 * @return AVERROR(EINVAL) on output size error
598 * @return AVERROR(ENOMEM) on coded frame queue data allocation error
600 static int queue_frames(AVCodecContext *avctx, AVPacket *pkt_out,
601 AVFrame *coded_frame)
603 VP8Context *ctx = avctx->priv_data;
604 const struct vpx_codec_cx_pkt *pkt;
605 const struct vpx_codec_cx_pkt *pkt_alpha = NULL;
606 const void *iter = NULL;
607 const void *iter_alpha = NULL;
610 if (ctx->coded_frame_list) {
611 struct FrameListData *cx_frame = ctx->coded_frame_list;
612 /* return the leading frame if we've already begun queueing */
613 size = storeframe(avctx, cx_frame, pkt_out, coded_frame);
616 ctx->coded_frame_list = cx_frame->next;
617 free_coded_frame(cx_frame);
620 /* consume all available output from the encoder before returning. buffers
621 are only good through the next vpx_codec call */
622 while ((pkt = vpx_codec_get_cx_data(&ctx->encoder, &iter)) &&
624 (ctx->is_alpha && (pkt_alpha = vpx_codec_get_cx_data(&ctx->encoder_alpha, &iter_alpha))))) {
626 case VPX_CODEC_CX_FRAME_PKT:
628 struct FrameListData cx_frame;
630 /* avoid storing the frame when the list is empty and we haven't yet
631 provided a frame for output */
632 av_assert0(!ctx->coded_frame_list);
633 cx_pktcpy(&cx_frame, pkt, pkt_alpha, ctx);
634 size = storeframe(avctx, &cx_frame, pkt_out, coded_frame);
638 struct FrameListData *cx_frame =
639 av_malloc(sizeof(struct FrameListData));
642 av_log(avctx, AV_LOG_ERROR,
643 "Frame queue element alloc failed\n");
644 return AVERROR(ENOMEM);
646 cx_pktcpy(cx_frame, pkt, pkt_alpha, ctx);
647 cx_frame->buf = av_malloc(cx_frame->sz);
649 if (!cx_frame->buf) {
650 av_log(avctx, AV_LOG_ERROR,
651 "Data buffer alloc (%"SIZE_SPECIFIER" bytes) failed\n",
654 return AVERROR(ENOMEM);
656 memcpy(cx_frame->buf, pkt->data.frame.buf, pkt->data.frame.sz);
658 cx_frame->buf_alpha = av_malloc(cx_frame->sz_alpha);
659 if (!cx_frame->buf_alpha) {
660 av_log(avctx, AV_LOG_ERROR,
661 "Data buffer alloc (%"SIZE_SPECIFIER" bytes) failed\n",
664 return AVERROR(ENOMEM);
666 memcpy(cx_frame->buf_alpha, pkt_alpha->data.frame.buf, pkt_alpha->data.frame.sz);
668 coded_frame_add(&ctx->coded_frame_list, cx_frame);
671 case VPX_CODEC_STATS_PKT: {
672 struct vpx_fixed_buf *stats = &ctx->twopass_stats;
674 if ((err = av_reallocp(&stats->buf,
676 pkt->data.twopass_stats.sz)) < 0) {
678 av_log(avctx, AV_LOG_ERROR, "Stat buffer realloc failed\n");
681 memcpy((uint8_t*)stats->buf + stats->sz,
682 pkt->data.twopass_stats.buf, pkt->data.twopass_stats.sz);
683 stats->sz += pkt->data.twopass_stats.sz;
686 case VPX_CODEC_PSNR_PKT:
687 av_assert0(!ctx->have_sse);
688 ctx->sse[0] = pkt->data.psnr.sse[0];
689 ctx->sse[1] = pkt->data.psnr.sse[1];
690 ctx->sse[2] = pkt->data.psnr.sse[2];
691 ctx->sse[3] = pkt->data.psnr.sse[3];
694 case VPX_CODEC_CUSTOM_PKT:
695 //ignore unsupported/unrecognized packet types
703 static int vp8_encode(AVCodecContext *avctx, AVPacket *pkt,
704 const AVFrame *frame, int *got_packet)
706 VP8Context *ctx = avctx->priv_data;
707 struct vpx_image *rawimg = NULL;
708 struct vpx_image *rawimg_alpha = NULL;
709 int64_t timestamp = 0;
711 vpx_enc_frame_flags_t flags = 0;
714 rawimg = &ctx->rawimg;
715 rawimg->planes[VPX_PLANE_Y] = frame->data[0];
716 rawimg->planes[VPX_PLANE_U] = frame->data[1];
717 rawimg->planes[VPX_PLANE_V] = frame->data[2];
718 rawimg->stride[VPX_PLANE_Y] = frame->linesize[0];
719 rawimg->stride[VPX_PLANE_U] = frame->linesize[1];
720 rawimg->stride[VPX_PLANE_V] = frame->linesize[2];
722 uint8_t *u_plane, *v_plane;
723 rawimg_alpha = &ctx->rawimg_alpha;
724 rawimg_alpha->planes[VPX_PLANE_Y] = frame->data[3];
725 u_plane = av_malloc(frame->linesize[1] * frame->height);
726 v_plane = av_malloc(frame->linesize[2] * frame->height);
727 if (!u_plane || !v_plane) {
730 return AVERROR(ENOMEM);
732 memset(u_plane, 0x80, frame->linesize[1] * frame->height);
733 rawimg_alpha->planes[VPX_PLANE_U] = u_plane;
734 memset(v_plane, 0x80, frame->linesize[2] * frame->height);
735 rawimg_alpha->planes[VPX_PLANE_V] = v_plane;
736 rawimg_alpha->stride[VPX_PLANE_Y] = frame->linesize[0];
737 rawimg_alpha->stride[VPX_PLANE_U] = frame->linesize[1];
738 rawimg_alpha->stride[VPX_PLANE_V] = frame->linesize[2];
740 timestamp = frame->pts;
741 if (frame->pict_type == AV_PICTURE_TYPE_I)
742 flags |= VPX_EFLAG_FORCE_KF;
745 res = vpx_codec_encode(&ctx->encoder, rawimg, timestamp,
746 avctx->ticks_per_frame, flags, ctx->deadline);
747 if (res != VPX_CODEC_OK) {
748 log_encoder_error(avctx, "Error encoding frame");
749 return AVERROR_INVALIDDATA;
753 res = vpx_codec_encode(&ctx->encoder_alpha, rawimg_alpha, timestamp,
754 avctx->ticks_per_frame, flags, ctx->deadline);
755 if (res != VPX_CODEC_OK) {
756 log_encoder_error(avctx, "Error encoding alpha frame");
757 return AVERROR_INVALIDDATA;
761 coded_size = queue_frames(avctx, pkt, avctx->coded_frame);
763 if (!frame && avctx->flags & CODEC_FLAG_PASS1) {
764 unsigned int b64_size = AV_BASE64_SIZE(ctx->twopass_stats.sz);
766 avctx->stats_out = av_malloc(b64_size);
767 if (!avctx->stats_out) {
768 av_log(avctx, AV_LOG_ERROR, "Stat buffer alloc (%d bytes) failed\n",
770 return AVERROR(ENOMEM);
772 av_base64_encode(avctx->stats_out, b64_size, ctx->twopass_stats.buf,
773 ctx->twopass_stats.sz);
777 av_freep(&rawimg_alpha->planes[VPX_PLANE_U]);
778 av_freep(&rawimg_alpha->planes[VPX_PLANE_V]);
781 *got_packet = !!coded_size;
785 #define OFFSET(x) offsetof(VP8Context, x)
786 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
788 #ifndef VPX_ERROR_RESILIENT_DEFAULT
789 #define VPX_ERROR_RESILIENT_DEFAULT 1
790 #define VPX_ERROR_RESILIENT_PARTITIONS 2
793 #define COMMON_OPTIONS \
794 { "cpu-used", "Quality/Speed ratio modifier", OFFSET(cpu_used), AV_OPT_TYPE_INT, {.i64 = 1}, -16, 16, VE}, \
795 { "auto-alt-ref", "Enable use of alternate reference " \
796 "frames (2-pass only)", OFFSET(auto_alt_ref), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE}, \
797 { "lag-in-frames", "Number of frames to look ahead for " \
798 "alternate reference frame selection", OFFSET(lag_in_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
799 { "arnr-maxframes", "altref noise reduction max frame count", OFFSET(arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
800 { "arnr-strength", "altref noise reduction filter strength", OFFSET(arnr_strength), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
801 { "arnr-type", "altref noise reduction filter type", OFFSET(arnr_type), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE, "arnr_type"}, \
802 { "backward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0, VE, "arnr_type" }, \
803 { "forward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0, VE, "arnr_type" }, \
804 { "centered", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 3}, 0, 0, VE, "arnr_type" }, \
805 { "deadline", "Time to spend encoding, in microseconds.", OFFSET(deadline), AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, "quality"}, \
806 { "best", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_BEST_QUALITY}, 0, 0, VE, "quality"}, \
807 { "good", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_GOOD_QUALITY}, 0, 0, VE, "quality"}, \
808 { "realtime", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_REALTIME}, 0, 0, VE, "quality"}, \
809 { "error-resilient", "Error resilience configuration", OFFSET(error_resilient), AV_OPT_TYPE_FLAGS, {.i64 = 0}, INT_MIN, INT_MAX, VE, "er"}, \
810 { "max-intra-rate", "Maximum I-frame bitrate (pct) 0=unlimited", OFFSET(max_intra_rate), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
811 { "default", "Improve resiliency against losses of whole frames", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_DEFAULT}, 0, 0, VE, "er"}, \
812 { "partitions", "The frame partitions are independently decodable " \
813 "by the bool decoder, meaning that partitions can be decoded even " \
814 "though earlier partitions have been lost. Note that intra predicition" \
815 " is still done over the partition boundary.", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_PARTITIONS}, 0, 0, VE, "er"}, \
816 { "crf", "Select the quality for constant quality mode", offsetof(VP8Context, crf), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 63, VE }, \
817 { "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 }, \
819 #define LEGACY_OPTIONS \
820 {"speed", "", offsetof(VP8Context, cpu_used), AV_OPT_TYPE_INT, {.i64 = 1}, -16, 16, VE}, \
821 {"quality", "", offsetof(VP8Context, deadline), AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, "quality"}, \
822 {"vp8flags", "", offsetof(VP8Context, flags), FF_OPT_TYPE_FLAGS, {.i64 = 0}, 0, UINT_MAX, VE, "flags"}, \
823 {"error_resilient", "enable error resilience", 0, FF_OPT_TYPE_CONST, {.dbl = VP8F_ERROR_RESILIENT}, INT_MIN, INT_MAX, VE, "flags"}, \
824 {"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"}, \
825 {"arnr_max_frames", "altref noise reduction max frame count", offsetof(VP8Context, arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 15, VE}, \
826 {"arnr_strength", "altref noise reduction filter strength", offsetof(VP8Context, arnr_strength), AV_OPT_TYPE_INT, {.i64 = 3}, 0, 6, VE}, \
827 {"arnr_type", "altref noise reduction filter type", offsetof(VP8Context, arnr_type), AV_OPT_TYPE_INT, {.i64 = 3}, 1, 3, VE}, \
828 {"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}, \
830 #if CONFIG_LIBVPX_VP8_ENCODER
831 static const AVOption vp8_options[] = {
838 #if CONFIG_LIBVPX_VP9_ENCODER
839 static const AVOption vp9_options[] = {
841 { "lossless", "Lossless mode", OFFSET(lossless), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE},
842 { "tile-columns", "Number of tile columns to use, log2", OFFSET(tile_columns), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 6, VE},
843 { "tile-rows", "Number of tile rows to use, log2", OFFSET(tile_rows), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 2, VE},
844 { "frame-parallel", "Enable frame parallel decodability features", OFFSET(frame_parallel), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE},
845 { "aq-mode", "adaptive quantization mode", OFFSET(aq_mode), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 3, VE, "aq_mode"},
846 { "none", "Aq not used", 0, AV_OPT_TYPE_CONST, {.i64 = 0}, 0, 0, VE, "aq_mode" }, \
847 { "variance", "Variance based Aq", 0, AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0, VE, "aq_mode" }, \
848 { "complexity", "Complexity based Aq", 0, AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0, VE, "aq_mode" }, \
849 { "cyclic", "Cyclic Refresh Aq", 0, AV_OPT_TYPE_CONST, {.i64 = 3}, 0, 0, VE, "aq_mode" }, \
855 #undef COMMON_OPTIONS
856 #undef LEGACY_OPTIONS
858 static const AVCodecDefault defaults[] = {
862 { "keyint_min", "-1" },
866 #if CONFIG_LIBVPX_VP8_ENCODER
867 static av_cold int vp8_init(AVCodecContext *avctx)
869 return vpx_init(avctx, vpx_codec_vp8_cx());
872 static const AVClass class_vp8 = {
873 .class_name = "libvpx-vp8 encoder",
874 .item_name = av_default_item_name,
875 .option = vp8_options,
876 .version = LIBAVUTIL_VERSION_INT,
879 AVCodec ff_libvpx_vp8_encoder = {
881 .long_name = NULL_IF_CONFIG_SMALL("libvpx VP8"),
882 .type = AVMEDIA_TYPE_VIDEO,
883 .id = AV_CODEC_ID_VP8,
884 .priv_data_size = sizeof(VP8Context),
886 .encode2 = vp8_encode,
888 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
889 .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUVA420P, AV_PIX_FMT_NONE },
890 .priv_class = &class_vp8,
891 .defaults = defaults,
893 #endif /* CONFIG_LIBVPX_VP8_ENCODER */
895 #if CONFIG_LIBVPX_VP9_ENCODER
896 static av_cold int vp9_init(AVCodecContext *avctx)
898 return vpx_init(avctx, vpx_codec_vp9_cx());
901 static const AVClass class_vp9 = {
902 .class_name = "libvpx-vp9 encoder",
903 .item_name = av_default_item_name,
904 .option = vp9_options,
905 .version = LIBAVUTIL_VERSION_INT,
908 AVCodec ff_libvpx_vp9_encoder = {
909 .name = "libvpx-vp9",
910 .long_name = NULL_IF_CONFIG_SMALL("libvpx VP9"),
911 .type = AVMEDIA_TYPE_VIDEO,
912 .id = AV_CODEC_ID_VP9,
913 .priv_data_size = sizeof(VP8Context),
915 .encode2 = vp8_encode,
917 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
918 .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE },
919 .priv_class = &class_vp9,
920 .defaults = defaults,
921 .init_static_data = ff_vp9_init_static,
923 #endif /* CONFIG_LIBVPX_VP9_ENCODER */