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"
34 #include "libavutil/base64.h"
35 #include "libavutil/common.h"
36 #include "libavutil/intreadwrite.h"
37 #include "libavutil/mathematics.h"
38 #include "libavutil/opt.h"
41 * Portion of struct vpx_codec_cx_pkt from vpx_encoder.h.
42 * One encoded frame returned from the library.
44 struct FrameListData {
45 void *buf; /**< compressed data buffer */
46 size_t sz; /**< length of compressed data */
49 int64_t pts; /**< time stamp to show frame
50 (in timebase units) */
51 unsigned long duration; /**< duration to show frame
52 (in timebase units) */
53 uint32_t flags; /**< flags for this frame */
55 int have_sse; /**< true if we have pending sse[] */
56 uint64_t frame_number;
57 struct FrameListData *next;
60 typedef struct VP8EncoderContext {
62 struct vpx_codec_ctx encoder;
63 struct vpx_image rawimg;
64 struct vpx_codec_ctx encoder_alpha;
65 struct vpx_image rawimg_alpha;
67 struct vpx_fixed_buf twopass_stats;
68 int deadline; //i.e., RT/GOOD/BEST
70 int have_sse; /**< true if we have pending sse[] */
71 uint64_t frame_number;
72 struct FrameListData *coded_frame_list;
76 * VP8 specific flags, see VP8F_* below.
79 #define VP8F_ERROR_RESILIENT 0x00000001 ///< Enable measures appropriate for streaming over lossy links
80 #define VP8F_AUTO_ALT_REF 0x00000002 ///< Enable automatic alternate reference frame generation
100 /** String mappings for enum vp8e_enc_control_id */
101 static const char *const ctlidstr[] = {
102 [VP8E_UPD_ENTROPY] = "VP8E_UPD_ENTROPY",
103 [VP8E_UPD_REFERENCE] = "VP8E_UPD_REFERENCE",
104 [VP8E_USE_REFERENCE] = "VP8E_USE_REFERENCE",
105 [VP8E_SET_ROI_MAP] = "VP8E_SET_ROI_MAP",
106 [VP8E_SET_ACTIVEMAP] = "VP8E_SET_ACTIVEMAP",
107 [VP8E_SET_SCALEMODE] = "VP8E_SET_SCALEMODE",
108 [VP8E_SET_CPUUSED] = "VP8E_SET_CPUUSED",
109 [VP8E_SET_ENABLEAUTOALTREF] = "VP8E_SET_ENABLEAUTOALTREF",
110 [VP8E_SET_NOISE_SENSITIVITY] = "VP8E_SET_NOISE_SENSITIVITY",
111 [VP8E_SET_SHARPNESS] = "VP8E_SET_SHARPNESS",
112 [VP8E_SET_STATIC_THRESHOLD] = "VP8E_SET_STATIC_THRESHOLD",
113 [VP8E_SET_TOKEN_PARTITIONS] = "VP8E_SET_TOKEN_PARTITIONS",
114 [VP8E_GET_LAST_QUANTIZER] = "VP8E_GET_LAST_QUANTIZER",
115 [VP8E_SET_ARNR_MAXFRAMES] = "VP8E_SET_ARNR_MAXFRAMES",
116 [VP8E_SET_ARNR_STRENGTH] = "VP8E_SET_ARNR_STRENGTH",
117 [VP8E_SET_ARNR_TYPE] = "VP8E_SET_ARNR_TYPE",
118 [VP8E_SET_CQ_LEVEL] = "VP8E_SET_CQ_LEVEL",
119 [VP8E_SET_MAX_INTRA_BITRATE_PCT] = "VP8E_SET_MAX_INTRA_BITRATE_PCT",
120 #if CONFIG_LIBVPX_VP9_ENCODER
121 [VP9E_SET_LOSSLESS] = "VP9E_SET_LOSSLESS",
122 [VP9E_SET_TILE_COLUMNS] = "VP9E_SET_TILE_COLUMNS",
123 [VP9E_SET_TILE_ROWS] = "VP9E_SET_TILE_ROWS",
124 [VP9E_SET_FRAME_PARALLEL_DECODING] = "VP9E_SET_FRAME_PARALLEL_DECODING",
128 static av_cold void log_encoder_error(AVCodecContext *avctx, const char *desc)
130 VP8Context *ctx = avctx->priv_data;
131 const char *error = vpx_codec_error(&ctx->encoder);
132 const char *detail = vpx_codec_error_detail(&ctx->encoder);
134 av_log(avctx, AV_LOG_ERROR, "%s: %s\n", desc, error);
136 av_log(avctx, AV_LOG_ERROR, " Additional information: %s\n", detail);
139 static av_cold void dump_enc_cfg(AVCodecContext *avctx,
140 const struct vpx_codec_enc_cfg *cfg)
143 int level = AV_LOG_DEBUG;
145 av_log(avctx, level, "vpx_codec_enc_cfg\n");
146 av_log(avctx, level, "generic settings\n"
147 " %*s%u\n %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
148 " %*s{%u/%u}\n %*s%u\n %*s%d\n %*s%u\n",
149 width, "g_usage:", cfg->g_usage,
150 width, "g_threads:", cfg->g_threads,
151 width, "g_profile:", cfg->g_profile,
152 width, "g_w:", cfg->g_w,
153 width, "g_h:", cfg->g_h,
154 width, "g_timebase:", cfg->g_timebase.num, cfg->g_timebase.den,
155 width, "g_error_resilient:", cfg->g_error_resilient,
156 width, "g_pass:", cfg->g_pass,
157 width, "g_lag_in_frames:", cfg->g_lag_in_frames);
158 av_log(avctx, level, "rate control settings\n"
159 " %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
160 " %*s%d\n %*s%p(%zu)\n %*s%u\n",
161 width, "rc_dropframe_thresh:", cfg->rc_dropframe_thresh,
162 width, "rc_resize_allowed:", cfg->rc_resize_allowed,
163 width, "rc_resize_up_thresh:", cfg->rc_resize_up_thresh,
164 width, "rc_resize_down_thresh:", cfg->rc_resize_down_thresh,
165 width, "rc_end_usage:", cfg->rc_end_usage,
166 width, "rc_twopass_stats_in:", cfg->rc_twopass_stats_in.buf, cfg->rc_twopass_stats_in.sz,
167 width, "rc_target_bitrate:", cfg->rc_target_bitrate);
168 av_log(avctx, level, "quantizer settings\n"
170 width, "rc_min_quantizer:", cfg->rc_min_quantizer,
171 width, "rc_max_quantizer:", cfg->rc_max_quantizer);
172 av_log(avctx, level, "bitrate tolerance\n"
174 width, "rc_undershoot_pct:", cfg->rc_undershoot_pct,
175 width, "rc_overshoot_pct:", cfg->rc_overshoot_pct);
176 av_log(avctx, level, "decoder buffer model\n"
177 " %*s%u\n %*s%u\n %*s%u\n",
178 width, "rc_buf_sz:", cfg->rc_buf_sz,
179 width, "rc_buf_initial_sz:", cfg->rc_buf_initial_sz,
180 width, "rc_buf_optimal_sz:", cfg->rc_buf_optimal_sz);
181 av_log(avctx, level, "2 pass rate control settings\n"
182 " %*s%u\n %*s%u\n %*s%u\n",
183 width, "rc_2pass_vbr_bias_pct:", cfg->rc_2pass_vbr_bias_pct,
184 width, "rc_2pass_vbr_minsection_pct:", cfg->rc_2pass_vbr_minsection_pct,
185 width, "rc_2pass_vbr_maxsection_pct:", cfg->rc_2pass_vbr_maxsection_pct);
186 av_log(avctx, level, "keyframing settings\n"
187 " %*s%d\n %*s%u\n %*s%u\n",
188 width, "kf_mode:", cfg->kf_mode,
189 width, "kf_min_dist:", cfg->kf_min_dist,
190 width, "kf_max_dist:", cfg->kf_max_dist);
191 av_log(avctx, level, "\n");
194 static void coded_frame_add(void *list, struct FrameListData *cx_frame)
196 struct FrameListData **p = list;
201 cx_frame->next = NULL;
204 static av_cold void free_coded_frame(struct FrameListData *cx_frame)
206 av_freep(&cx_frame->buf);
207 if (cx_frame->buf_alpha)
208 av_freep(&cx_frame->buf_alpha);
212 static av_cold void free_frame_list(struct FrameListData *list)
214 struct FrameListData *p = list;
223 static av_cold int codecctl_int(AVCodecContext *avctx,
224 enum vp8e_enc_control_id id, int val)
226 VP8Context *ctx = avctx->priv_data;
231 snprintf(buf, sizeof(buf), "%s:", ctlidstr[id]);
232 av_log(avctx, AV_LOG_DEBUG, " %*s%d\n", width, buf, val);
234 res = vpx_codec_control(&ctx->encoder, id, val);
235 if (res != VPX_CODEC_OK) {
236 snprintf(buf, sizeof(buf), "Failed to set %s codec control",
238 log_encoder_error(avctx, buf);
241 return res == VPX_CODEC_OK ? 0 : AVERROR(EINVAL);
244 static av_cold int vp8_free(AVCodecContext *avctx)
246 VP8Context *ctx = avctx->priv_data;
248 vpx_codec_destroy(&ctx->encoder);
250 vpx_codec_destroy(&ctx->encoder_alpha);
251 av_freep(&ctx->twopass_stats.buf);
252 av_freep(&avctx->coded_frame);
253 av_freep(&avctx->stats_out);
254 free_frame_list(ctx->coded_frame_list);
258 static av_cold int vpx_init(AVCodecContext *avctx,
259 const struct vpx_codec_iface *iface)
261 VP8Context *ctx = avctx->priv_data;
262 struct vpx_codec_enc_cfg enccfg;
263 struct vpx_codec_enc_cfg enccfg_alpha;
264 vpx_codec_flags_t flags = (avctx->flags & CODEC_FLAG_PSNR) ? VPX_CODEC_USE_PSNR : 0;
267 av_log(avctx, AV_LOG_INFO, "%s\n", vpx_codec_version_str());
268 av_log(avctx, AV_LOG_VERBOSE, "%s\n", vpx_codec_build_config());
270 if (avctx->pix_fmt == AV_PIX_FMT_YUVA420P)
273 if ((res = vpx_codec_enc_config_default(iface, &enccfg, 0)) != VPX_CODEC_OK) {
274 av_log(avctx, AV_LOG_ERROR, "Failed to get config: %s\n",
275 vpx_codec_err_to_string(res));
276 return AVERROR(EINVAL);
280 if(avctx->rc_max_rate || avctx->rc_buffer_size || avctx->rc_initial_buffer_occupancy) {
281 av_log( avctx, AV_LOG_ERROR, "Rate control parameters set without a bitrate\n");
282 return AVERROR(EINVAL);
285 dump_enc_cfg(avctx, &enccfg);
287 enccfg.g_w = avctx->width;
288 enccfg.g_h = avctx->height;
289 enccfg.g_timebase.num = avctx->time_base.num;
290 enccfg.g_timebase.den = avctx->time_base.den;
291 enccfg.g_threads = avctx->thread_count;
292 enccfg.g_lag_in_frames= ctx->lag_in_frames;
294 if (avctx->flags & CODEC_FLAG_PASS1)
295 enccfg.g_pass = VPX_RC_FIRST_PASS;
296 else if (avctx->flags & CODEC_FLAG_PASS2)
297 enccfg.g_pass = VPX_RC_LAST_PASS;
299 enccfg.g_pass = VPX_RC_ONE_PASS;
301 if (avctx->rc_min_rate == avctx->rc_max_rate &&
302 avctx->rc_min_rate == avctx->bit_rate && avctx->bit_rate)
303 enccfg.rc_end_usage = VPX_CBR;
305 enccfg.rc_end_usage = VPX_CQ;
307 if (avctx->bit_rate) {
308 enccfg.rc_target_bitrate = av_rescale_rnd(avctx->bit_rate, 1, 1000,
311 if (enccfg.rc_end_usage == VPX_CQ) {
312 enccfg.rc_target_bitrate = 1000000;
314 avctx->bit_rate = enccfg.rc_target_bitrate * 1000;
315 av_log(avctx, AV_LOG_WARNING,
316 "Neither bitrate nor constrained quality specified, using default bitrate of %dkbit/sec\n",
317 enccfg.rc_target_bitrate);
321 if (avctx->qmin >= 0)
322 enccfg.rc_min_quantizer = avctx->qmin;
323 if (avctx->qmax >= 0)
324 enccfg.rc_max_quantizer = avctx->qmax;
326 if (enccfg.rc_end_usage == VPX_CQ) {
327 if (ctx->crf < enccfg.rc_min_quantizer || ctx->crf > enccfg.rc_max_quantizer) {
328 av_log(avctx, AV_LOG_ERROR,
329 "CQ level must be between minimum and maximum quantizer value (%d-%d)\n",
330 enccfg.rc_min_quantizer, enccfg.rc_max_quantizer);
331 return AVERROR(EINVAL);
335 enccfg.rc_dropframe_thresh = avctx->frame_skip_threshold;
337 //0-100 (0 => CBR, 100 => VBR)
338 enccfg.rc_2pass_vbr_bias_pct = round(avctx->qcompress * 100);
340 enccfg.rc_2pass_vbr_minsection_pct =
341 avctx->rc_min_rate * 100LL / avctx->bit_rate;
342 if (avctx->rc_max_rate)
343 enccfg.rc_2pass_vbr_maxsection_pct =
344 avctx->rc_max_rate * 100LL / avctx->bit_rate;
346 if (avctx->rc_buffer_size)
348 avctx->rc_buffer_size * 1000LL / avctx->bit_rate;
349 if (avctx->rc_initial_buffer_occupancy)
350 enccfg.rc_buf_initial_sz =
351 avctx->rc_initial_buffer_occupancy * 1000LL / avctx->bit_rate;
352 enccfg.rc_buf_optimal_sz = enccfg.rc_buf_sz * 5 / 6;
353 enccfg.rc_undershoot_pct = round(avctx->rc_buffer_aggressivity * 100);
355 //_enc_init() will balk if kf_min_dist differs from max w/VPX_KF_AUTO
356 if (avctx->keyint_min >= 0 && avctx->keyint_min == avctx->gop_size)
357 enccfg.kf_min_dist = avctx->keyint_min;
358 if (avctx->gop_size >= 0)
359 enccfg.kf_max_dist = avctx->gop_size;
361 if (enccfg.g_pass == VPX_RC_FIRST_PASS)
362 enccfg.g_lag_in_frames = 0;
363 else if (enccfg.g_pass == VPX_RC_LAST_PASS) {
366 if (!avctx->stats_in) {
367 av_log(avctx, AV_LOG_ERROR, "No stats file for second pass\n");
368 return AVERROR_INVALIDDATA;
371 ctx->twopass_stats.sz = strlen(avctx->stats_in) * 3 / 4;
372 ctx->twopass_stats.buf = av_malloc(ctx->twopass_stats.sz);
373 if (!ctx->twopass_stats.buf) {
374 av_log(avctx, AV_LOG_ERROR,
375 "Stat buffer alloc (%zu bytes) failed\n",
376 ctx->twopass_stats.sz);
377 return AVERROR(ENOMEM);
379 decode_size = av_base64_decode(ctx->twopass_stats.buf, avctx->stats_in,
380 ctx->twopass_stats.sz);
381 if (decode_size < 0) {
382 av_log(avctx, AV_LOG_ERROR, "Stat buffer decode failed\n");
383 return AVERROR_INVALIDDATA;
386 ctx->twopass_stats.sz = decode_size;
387 enccfg.rc_twopass_stats_in = ctx->twopass_stats;
390 /* 0-3: For non-zero values the encoder increasingly optimizes for reduced
391 complexity playback on low powered devices at the expense of encode
393 if (avctx->profile != FF_PROFILE_UNKNOWN)
394 enccfg.g_profile = avctx->profile;
396 enccfg.g_error_resilient = ctx->error_resilient || ctx->flags & VP8F_ERROR_RESILIENT;
398 dump_enc_cfg(avctx, &enccfg);
399 /* Construct Encoder Context */
400 res = vpx_codec_enc_init(&ctx->encoder, iface, &enccfg, flags);
401 if (res != VPX_CODEC_OK) {
402 log_encoder_error(avctx, "Failed to initialize encoder");
403 return AVERROR(EINVAL);
407 enccfg_alpha = enccfg;
408 res = vpx_codec_enc_init(&ctx->encoder_alpha, iface, &enccfg_alpha, flags);
409 if (res != VPX_CODEC_OK) {
410 log_encoder_error(avctx, "Failed to initialize alpha encoder");
411 return AVERROR(EINVAL);
415 //codec control failures are currently treated only as warnings
416 av_log(avctx, AV_LOG_DEBUG, "vpx_codec_control\n");
417 if (ctx->cpu_used != INT_MIN)
418 codecctl_int(avctx, VP8E_SET_CPUUSED, ctx->cpu_used);
419 if (ctx->flags & VP8F_AUTO_ALT_REF)
420 ctx->auto_alt_ref = 1;
421 if (ctx->auto_alt_ref >= 0)
422 codecctl_int(avctx, VP8E_SET_ENABLEAUTOALTREF, ctx->auto_alt_ref);
423 if (ctx->arnr_max_frames >= 0)
424 codecctl_int(avctx, VP8E_SET_ARNR_MAXFRAMES, ctx->arnr_max_frames);
425 if (ctx->arnr_strength >= 0)
426 codecctl_int(avctx, VP8E_SET_ARNR_STRENGTH, ctx->arnr_strength);
427 if (ctx->arnr_type >= 0)
428 codecctl_int(avctx, VP8E_SET_ARNR_TYPE, ctx->arnr_type);
429 codecctl_int(avctx, VP8E_SET_NOISE_SENSITIVITY, avctx->noise_reduction);
430 codecctl_int(avctx, VP8E_SET_TOKEN_PARTITIONS, av_log2(avctx->slices));
431 codecctl_int(avctx, VP8E_SET_STATIC_THRESHOLD, avctx->mb_threshold);
432 codecctl_int(avctx, VP8E_SET_CQ_LEVEL, ctx->crf);
433 if (ctx->max_intra_rate >= 0)
434 codecctl_int(avctx, VP8E_SET_MAX_INTRA_BITRATE_PCT, ctx->max_intra_rate);
436 #if CONFIG_LIBVPX_VP9_ENCODER
437 if (avctx->codec_id == AV_CODEC_ID_VP9) {
438 if (ctx->lossless >= 0)
439 codecctl_int(avctx, VP9E_SET_LOSSLESS, ctx->lossless);
440 if (ctx->tile_columns >= 0)
441 codecctl_int(avctx, VP9E_SET_TILE_COLUMNS, ctx->tile_columns);
442 if (ctx->tile_rows >= 0)
443 codecctl_int(avctx, VP9E_SET_TILE_ROWS, ctx->tile_rows);
444 if (ctx->frame_parallel >= 0)
445 codecctl_int(avctx, VP9E_SET_FRAME_PARALLEL_DECODING, ctx->frame_parallel);
449 av_log(avctx, AV_LOG_DEBUG, "Using deadline: %d\n", ctx->deadline);
451 //provide dummy value to initialize wrapper, values will be updated each _encode()
452 vpx_img_wrap(&ctx->rawimg, VPX_IMG_FMT_I420, avctx->width, avctx->height, 1,
456 vpx_img_wrap(&ctx->rawimg_alpha, VPX_IMG_FMT_I420, avctx->width, avctx->height, 1,
459 avctx->coded_frame = av_frame_alloc();
460 if (!avctx->coded_frame) {
461 av_log(avctx, AV_LOG_ERROR, "Error allocating coded frame\n");
463 return AVERROR(ENOMEM);
468 static inline void cx_pktcpy(struct FrameListData *dst,
469 const struct vpx_codec_cx_pkt *src,
470 const struct vpx_codec_cx_pkt *src_alpha,
473 dst->pts = src->data.frame.pts;
474 dst->duration = src->data.frame.duration;
475 dst->flags = src->data.frame.flags;
476 dst->sz = src->data.frame.sz;
477 dst->buf = src->data.frame.buf;
479 /* For alt-ref frame, don't store PSNR or increment frame_number */
480 if (!(dst->flags & VPX_FRAME_IS_INVISIBLE)) {
481 dst->frame_number = ++ctx->frame_number;
482 dst->have_sse = ctx->have_sse;
484 /* associate last-seen SSE to the frame. */
485 /* Transfers ownership from ctx to dst. */
486 /* WARNING! This makes the assumption that PSNR_PKT comes
487 just before the frame it refers to! */
488 memcpy(dst->sse, ctx->sse, sizeof(dst->sse));
492 dst->frame_number = -1; /* sanity marker */
495 dst->buf_alpha = src_alpha->data.frame.buf;
496 dst->sz_alpha = src_alpha->data.frame.sz;
499 dst->buf_alpha = NULL;
505 * Store coded frame information in format suitable for return from encode2().
507 * Write information from @a cx_frame to @a pkt
508 * @return packet data size on success
509 * @return a negative AVERROR on error
511 static int storeframe(AVCodecContext *avctx, struct FrameListData *cx_frame,
512 AVPacket *pkt, AVFrame *coded_frame)
514 int ret = ff_alloc_packet2(avctx, pkt, cx_frame->sz);
517 memcpy(pkt->data, cx_frame->buf, pkt->size);
518 pkt->pts = pkt->dts = cx_frame->pts;
519 coded_frame->pts = cx_frame->pts;
520 coded_frame->key_frame = !!(cx_frame->flags & VPX_FRAME_IS_KEY);
522 if (coded_frame->key_frame) {
523 coded_frame->pict_type = AV_PICTURE_TYPE_I;
524 pkt->flags |= AV_PKT_FLAG_KEY;
526 coded_frame->pict_type = AV_PICTURE_TYPE_P;
528 if (cx_frame->have_sse) {
530 /* Beware of the Y/U/V/all order! */
531 coded_frame->error[0] = cx_frame->sse[1];
532 coded_frame->error[1] = cx_frame->sse[2];
533 coded_frame->error[2] = cx_frame->sse[3];
534 coded_frame->error[3] = 0; // alpha
535 for (i = 0; i < 4; ++i) {
536 avctx->error[i] += coded_frame->error[i];
538 cx_frame->have_sse = 0;
540 if (cx_frame->sz_alpha > 0) {
541 side_data = av_packet_new_side_data(pkt,
542 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
543 cx_frame->sz_alpha + 8);
544 if(side_data == NULL) {
547 return AVERROR(ENOMEM);
549 AV_WB64(side_data, 1);
550 memcpy(side_data + 8, cx_frame->buf_alpha, cx_frame->sz_alpha);
559 * Queue multiple output frames from the encoder, returning the front-most.
560 * In cases where vpx_codec_get_cx_data() returns more than 1 frame append
561 * the frame queue. Return the head frame if available.
562 * @return Stored frame size
563 * @return AVERROR(EINVAL) on output size error
564 * @return AVERROR(ENOMEM) on coded frame queue data allocation error
566 static int queue_frames(AVCodecContext *avctx, AVPacket *pkt_out,
567 AVFrame *coded_frame)
569 VP8Context *ctx = avctx->priv_data;
570 const struct vpx_codec_cx_pkt *pkt;
571 const struct vpx_codec_cx_pkt *pkt_alpha = NULL;
572 const void *iter = NULL;
573 const void *iter_alpha = NULL;
576 if (ctx->coded_frame_list) {
577 struct FrameListData *cx_frame = ctx->coded_frame_list;
578 /* return the leading frame if we've already begun queueing */
579 size = storeframe(avctx, cx_frame, pkt_out, coded_frame);
582 ctx->coded_frame_list = cx_frame->next;
583 free_coded_frame(cx_frame);
586 /* consume all available output from the encoder before returning. buffers
587 are only good through the next vpx_codec call */
588 while ((pkt = vpx_codec_get_cx_data(&ctx->encoder, &iter)) &&
590 (ctx->is_alpha && (pkt_alpha = vpx_codec_get_cx_data(&ctx->encoder_alpha, &iter_alpha))))) {
592 case VPX_CODEC_CX_FRAME_PKT:
594 struct FrameListData cx_frame;
596 /* avoid storing the frame when the list is empty and we haven't yet
597 provided a frame for output */
598 av_assert0(!ctx->coded_frame_list);
599 cx_pktcpy(&cx_frame, pkt, pkt_alpha, ctx);
600 size = storeframe(avctx, &cx_frame, pkt_out, coded_frame);
604 struct FrameListData *cx_frame =
605 av_malloc(sizeof(struct FrameListData));
608 av_log(avctx, AV_LOG_ERROR,
609 "Frame queue element alloc failed\n");
610 return AVERROR(ENOMEM);
612 cx_pktcpy(cx_frame, pkt, pkt_alpha, ctx);
613 cx_frame->buf = av_malloc(cx_frame->sz);
615 if (!cx_frame->buf) {
616 av_log(avctx, AV_LOG_ERROR,
617 "Data buffer alloc (%zu bytes) failed\n",
620 return AVERROR(ENOMEM);
622 memcpy(cx_frame->buf, pkt->data.frame.buf, pkt->data.frame.sz);
624 cx_frame->buf_alpha = av_malloc(cx_frame->sz_alpha);
625 if (!cx_frame->buf_alpha) {
626 av_log(avctx, AV_LOG_ERROR,
627 "Data buffer alloc (%zu bytes) failed\n",
630 return AVERROR(ENOMEM);
632 memcpy(cx_frame->buf_alpha, pkt_alpha->data.frame.buf, pkt_alpha->data.frame.sz);
634 coded_frame_add(&ctx->coded_frame_list, cx_frame);
637 case VPX_CODEC_STATS_PKT: {
638 struct vpx_fixed_buf *stats = &ctx->twopass_stats;
639 stats->buf = av_realloc_f(stats->buf, 1,
640 stats->sz + pkt->data.twopass_stats.sz);
642 av_log(avctx, AV_LOG_ERROR, "Stat buffer realloc failed\n");
643 return AVERROR(ENOMEM);
645 memcpy((uint8_t*)stats->buf + stats->sz,
646 pkt->data.twopass_stats.buf, pkt->data.twopass_stats.sz);
647 stats->sz += pkt->data.twopass_stats.sz;
650 case VPX_CODEC_PSNR_PKT:
651 av_assert0(!ctx->have_sse);
652 ctx->sse[0] = pkt->data.psnr.sse[0];
653 ctx->sse[1] = pkt->data.psnr.sse[1];
654 ctx->sse[2] = pkt->data.psnr.sse[2];
655 ctx->sse[3] = pkt->data.psnr.sse[3];
658 case VPX_CODEC_CUSTOM_PKT:
659 //ignore unsupported/unrecognized packet types
667 static int vp8_encode(AVCodecContext *avctx, AVPacket *pkt,
668 const AVFrame *frame, int *got_packet)
670 VP8Context *ctx = avctx->priv_data;
671 struct vpx_image *rawimg = NULL;
672 struct vpx_image *rawimg_alpha = NULL;
673 int64_t timestamp = 0;
675 vpx_enc_frame_flags_t flags = 0;
678 rawimg = &ctx->rawimg;
679 rawimg->planes[VPX_PLANE_Y] = frame->data[0];
680 rawimg->planes[VPX_PLANE_U] = frame->data[1];
681 rawimg->planes[VPX_PLANE_V] = frame->data[2];
682 rawimg->stride[VPX_PLANE_Y] = frame->linesize[0];
683 rawimg->stride[VPX_PLANE_U] = frame->linesize[1];
684 rawimg->stride[VPX_PLANE_V] = frame->linesize[2];
686 uint8_t *u_plane, *v_plane;
687 rawimg_alpha = &ctx->rawimg_alpha;
688 rawimg_alpha->planes[VPX_PLANE_Y] = frame->data[3];
689 u_plane = av_malloc(frame->linesize[1] * frame->height);
690 memset(u_plane, 0x80, frame->linesize[1] * frame->height);
691 rawimg_alpha->planes[VPX_PLANE_U] = u_plane;
692 v_plane = av_malloc(frame->linesize[2] * frame->height);
693 memset(v_plane, 0x80, frame->linesize[2] * frame->height);
694 rawimg_alpha->planes[VPX_PLANE_V] = v_plane;
695 rawimg_alpha->stride[VPX_PLANE_Y] = frame->linesize[0];
696 rawimg_alpha->stride[VPX_PLANE_U] = frame->linesize[1];
697 rawimg_alpha->stride[VPX_PLANE_V] = frame->linesize[2];
699 timestamp = frame->pts;
700 if (frame->pict_type == AV_PICTURE_TYPE_I)
701 flags |= VPX_EFLAG_FORCE_KF;
704 res = vpx_codec_encode(&ctx->encoder, rawimg, timestamp,
705 avctx->ticks_per_frame, flags, ctx->deadline);
706 if (res != VPX_CODEC_OK) {
707 log_encoder_error(avctx, "Error encoding frame");
708 return AVERROR_INVALIDDATA;
712 res = vpx_codec_encode(&ctx->encoder_alpha, rawimg_alpha, timestamp,
713 avctx->ticks_per_frame, flags, ctx->deadline);
714 if (res != VPX_CODEC_OK) {
715 log_encoder_error(avctx, "Error encoding alpha frame");
716 return AVERROR_INVALIDDATA;
720 coded_size = queue_frames(avctx, pkt, avctx->coded_frame);
722 if (!frame && avctx->flags & CODEC_FLAG_PASS1) {
723 unsigned int b64_size = AV_BASE64_SIZE(ctx->twopass_stats.sz);
725 avctx->stats_out = av_malloc(b64_size);
726 if (!avctx->stats_out) {
727 av_log(avctx, AV_LOG_ERROR, "Stat buffer alloc (%d bytes) failed\n",
729 return AVERROR(ENOMEM);
731 av_base64_encode(avctx->stats_out, b64_size, ctx->twopass_stats.buf,
732 ctx->twopass_stats.sz);
736 av_free(rawimg_alpha->planes[VPX_PLANE_U]);
737 av_free(rawimg_alpha->planes[VPX_PLANE_V]);
740 *got_packet = !!coded_size;
744 #define OFFSET(x) offsetof(VP8Context, x)
745 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
747 #ifndef VPX_ERROR_RESILIENT_DEFAULT
748 #define VPX_ERROR_RESILIENT_DEFAULT 1
749 #define VPX_ERROR_RESILIENT_PARTITIONS 2
752 #define COMMON_OPTIONS \
753 { "cpu-used", "Quality/Speed ratio modifier", OFFSET(cpu_used), AV_OPT_TYPE_INT, {.i64 = INT_MIN}, INT_MIN, INT_MAX, VE}, \
754 { "auto-alt-ref", "Enable use of alternate reference " \
755 "frames (2-pass only)", OFFSET(auto_alt_ref), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE}, \
756 { "lag-in-frames", "Number of frames to look ahead for " \
757 "alternate reference frame selection", OFFSET(lag_in_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
758 { "arnr-maxframes", "altref noise reduction max frame count", OFFSET(arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
759 { "arnr-strength", "altref noise reduction filter strength", OFFSET(arnr_strength), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
760 { "arnr-type", "altref noise reduction filter type", OFFSET(arnr_type), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE, "arnr_type"}, \
761 { "backward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0, VE, "arnr_type" }, \
762 { "forward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0, VE, "arnr_type" }, \
763 { "centered", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 3}, 0, 0, VE, "arnr_type" }, \
764 { "deadline", "Time to spend encoding, in microseconds.", OFFSET(deadline), AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, "quality"}, \
765 { "best", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_BEST_QUALITY}, 0, 0, VE, "quality"}, \
766 { "good", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_GOOD_QUALITY}, 0, 0, VE, "quality"}, \
767 { "realtime", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_REALTIME}, 0, 0, VE, "quality"}, \
768 { "error-resilient", "Error resilience configuration", OFFSET(error_resilient), AV_OPT_TYPE_FLAGS, {.i64 = 0}, INT_MIN, INT_MAX, VE, "er"}, \
769 { "max-intra-rate", "Maximum I-frame bitrate (pct) 0=unlimited", OFFSET(max_intra_rate), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
770 { "default", "Improve resiliency against losses of whole frames", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_DEFAULT}, 0, 0, VE, "er"}, \
771 { "partitions", "The frame partitions are independently decodable " \
772 "by the bool decoder, meaning that partitions can be decoded even " \
773 "though earlier partitions have been lost. Note that intra predicition" \
774 " is still done over the partition boundary.", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_PARTITIONS}, 0, 0, VE, "er"}, \
775 { "crf", "Select the quality for constant quality mode", offsetof(VP8Context, crf), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 63, VE }, \
777 #define LEGACY_OPTIONS \
778 {"speed", "", offsetof(VP8Context, cpu_used), AV_OPT_TYPE_INT, {.i64 = 3}, -16, 16, VE}, \
779 {"quality", "", offsetof(VP8Context, deadline), AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, "quality"}, \
780 {"vp8flags", "", offsetof(VP8Context, flags), FF_OPT_TYPE_FLAGS, {.i64 = 0}, 0, UINT_MAX, VE, "flags"}, \
781 {"error_resilient", "enable error resilience", 0, FF_OPT_TYPE_CONST, {.dbl = VP8F_ERROR_RESILIENT}, INT_MIN, INT_MAX, VE, "flags"}, \
782 {"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"}, \
783 {"arnr_max_frames", "altref noise reduction max frame count", offsetof(VP8Context, arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 15, VE}, \
784 {"arnr_strength", "altref noise reduction filter strength", offsetof(VP8Context, arnr_strength), AV_OPT_TYPE_INT, {.i64 = 3}, 0, 6, VE}, \
785 {"arnr_type", "altref noise reduction filter type", offsetof(VP8Context, arnr_type), AV_OPT_TYPE_INT, {.i64 = 3}, 1, 3, VE}, \
786 {"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}, \
788 #if CONFIG_LIBVPX_VP8_ENCODER
789 static const AVOption vp8_options[] = {
796 #if CONFIG_LIBVPX_VP9_ENCODER
797 static const AVOption vp9_options[] = {
799 { "lossless", "Lossless mode", OFFSET(lossless), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE},
800 { "tile-columns", "Number of tile columns to use, log2", OFFSET(tile_columns), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 6, VE},
801 { "tile-rows", "Number of tile rows to use, log2", OFFSET(tile_rows), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 2, VE},
802 { "frame-parallel", "Enable frame parallel decodability features", OFFSET(frame_parallel), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE},
808 #undef COMMON_OPTIONS
809 #undef LEGACY_OPTIONS
811 static const AVCodecDefault defaults[] = {
815 { "keyint_min", "-1" },
819 #if CONFIG_LIBVPX_VP8_ENCODER
820 static av_cold int vp8_init(AVCodecContext *avctx)
822 return vpx_init(avctx, &vpx_codec_vp8_cx_algo);
825 static const AVClass class_vp8 = {
826 .class_name = "libvpx-vp8 encoder",
827 .item_name = av_default_item_name,
828 .option = vp8_options,
829 .version = LIBAVUTIL_VERSION_INT,
832 AVCodec ff_libvpx_vp8_encoder = {
834 .long_name = NULL_IF_CONFIG_SMALL("libvpx VP8"),
835 .type = AVMEDIA_TYPE_VIDEO,
836 .id = AV_CODEC_ID_VP8,
837 .priv_data_size = sizeof(VP8Context),
839 .encode2 = vp8_encode,
841 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
842 .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUVA420P, AV_PIX_FMT_NONE },
843 .priv_class = &class_vp8,
844 .defaults = defaults,
846 #endif /* CONFIG_LIBVPX_VP8_ENCODER */
848 #if CONFIG_LIBVPX_VP9_ENCODER
849 static av_cold int vp9_init(AVCodecContext *avctx)
851 return vpx_init(avctx, &vpx_codec_vp9_cx_algo);
854 static const AVClass class_vp9 = {
855 .class_name = "libvpx-vp9 encoder",
856 .item_name = av_default_item_name,
857 .option = vp9_options,
858 .version = LIBAVUTIL_VERSION_INT,
861 AVCodec ff_libvpx_vp9_encoder = {
862 .name = "libvpx-vp9",
863 .long_name = NULL_IF_CONFIG_SMALL("libvpx VP9"),
864 .type = AVMEDIA_TYPE_VIDEO,
865 .id = AV_CODEC_ID_VP9,
866 .priv_data_size = sizeof(VP8Context),
868 .encode2 = vp8_encode,
870 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS | CODEC_CAP_EXPERIMENTAL,
871 .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE },
872 .priv_class = &class_vp9,
873 .defaults = defaults,
875 #endif /* CONFIG_LIBVPX_VP9_ENCODER */