FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
libvpxenc.c
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2010, Google, Inc.
3  *
4  * This file is part of FFmpeg.
5  *
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.
10  *
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.
15  *
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
19  */
20 
21 /**
22  * @file
23  * VP8 encoder support via libvpx
24  */
25 
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>
30 
31 #include "avcodec.h"
32 #include "internal.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"
39 
40 /**
41  * Portion of struct vpx_codec_cx_pkt from vpx_encoder.h.
42  * One encoded frame returned from the library.
43  */
44 struct FrameListData {
45  void *buf; /**< compressed data buffer */
46  size_t sz; /**< length of compressed data */
47  void *buf_alpha;
48  size_t sz_alpha;
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 */
54  uint64_t sse[4];
55  int have_sse; /**< true if we have pending sse[] */
56  uint64_t frame_number;
58 };
59 
60 typedef struct VP8EncoderContext {
61  AVClass *class;
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
69  uint64_t sse[4];
70  int have_sse; /**< true if we have pending sse[] */
71  uint64_t frame_number;
73 
74  int cpu_used;
75  /**
76  * VP8 specific flags, see VP8F_* below.
77  */
78  int flags;
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
81 
83 
86  int arnr_type;
87 
90  int crf;
92 } VP8Context;
93 
94 /** String mappings for enum vp8e_enc_control_id */
95 static const char *const ctlidstr[] = {
96  [VP8E_UPD_ENTROPY] = "VP8E_UPD_ENTROPY",
97  [VP8E_UPD_REFERENCE] = "VP8E_UPD_REFERENCE",
98  [VP8E_USE_REFERENCE] = "VP8E_USE_REFERENCE",
99  [VP8E_SET_ROI_MAP] = "VP8E_SET_ROI_MAP",
100  [VP8E_SET_ACTIVEMAP] = "VP8E_SET_ACTIVEMAP",
101  [VP8E_SET_SCALEMODE] = "VP8E_SET_SCALEMODE",
102  [VP8E_SET_CPUUSED] = "VP8E_SET_CPUUSED",
103  [VP8E_SET_ENABLEAUTOALTREF] = "VP8E_SET_ENABLEAUTOALTREF",
104  [VP8E_SET_NOISE_SENSITIVITY] = "VP8E_SET_NOISE_SENSITIVITY",
105  [VP8E_SET_SHARPNESS] = "VP8E_SET_SHARPNESS",
106  [VP8E_SET_STATIC_THRESHOLD] = "VP8E_SET_STATIC_THRESHOLD",
107  [VP8E_SET_TOKEN_PARTITIONS] = "VP8E_SET_TOKEN_PARTITIONS",
108  [VP8E_GET_LAST_QUANTIZER] = "VP8E_GET_LAST_QUANTIZER",
109  [VP8E_SET_ARNR_MAXFRAMES] = "VP8E_SET_ARNR_MAXFRAMES",
110  [VP8E_SET_ARNR_STRENGTH] = "VP8E_SET_ARNR_STRENGTH",
111  [VP8E_SET_ARNR_TYPE] = "VP8E_SET_ARNR_TYPE",
112  [VP8E_SET_CQ_LEVEL] = "VP8E_SET_CQ_LEVEL",
113  [VP8E_SET_MAX_INTRA_BITRATE_PCT] = "VP8E_SET_MAX_INTRA_BITRATE_PCT",
114 };
115 
116 static av_cold void log_encoder_error(AVCodecContext *avctx, const char *desc)
117 {
118  VP8Context *ctx = avctx->priv_data;
119  const char *error = vpx_codec_error(&ctx->encoder);
120  const char *detail = vpx_codec_error_detail(&ctx->encoder);
121 
122  av_log(avctx, AV_LOG_ERROR, "%s: %s\n", desc, error);
123  if (detail)
124  av_log(avctx, AV_LOG_ERROR, " Additional information: %s\n", detail);
125 }
126 
128  const struct vpx_codec_enc_cfg *cfg)
129 {
130  int width = -30;
131  int level = AV_LOG_DEBUG;
132 
133  av_log(avctx, level, "vpx_codec_enc_cfg\n");
134  av_log(avctx, level, "generic settings\n"
135  " %*s%u\n %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
136  " %*s{%u/%u}\n %*s%u\n %*s%d\n %*s%u\n",
137  width, "g_usage:", cfg->g_usage,
138  width, "g_threads:", cfg->g_threads,
139  width, "g_profile:", cfg->g_profile,
140  width, "g_w:", cfg->g_w,
141  width, "g_h:", cfg->g_h,
142  width, "g_timebase:", cfg->g_timebase.num, cfg->g_timebase.den,
143  width, "g_error_resilient:", cfg->g_error_resilient,
144  width, "g_pass:", cfg->g_pass,
145  width, "g_lag_in_frames:", cfg->g_lag_in_frames);
146  av_log(avctx, level, "rate control settings\n"
147  " %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
148  " %*s%d\n %*s%p(%zu)\n %*s%u\n",
149  width, "rc_dropframe_thresh:", cfg->rc_dropframe_thresh,
150  width, "rc_resize_allowed:", cfg->rc_resize_allowed,
151  width, "rc_resize_up_thresh:", cfg->rc_resize_up_thresh,
152  width, "rc_resize_down_thresh:", cfg->rc_resize_down_thresh,
153  width, "rc_end_usage:", cfg->rc_end_usage,
154  width, "rc_twopass_stats_in:", cfg->rc_twopass_stats_in.buf, cfg->rc_twopass_stats_in.sz,
155  width, "rc_target_bitrate:", cfg->rc_target_bitrate);
156  av_log(avctx, level, "quantizer settings\n"
157  " %*s%u\n %*s%u\n",
158  width, "rc_min_quantizer:", cfg->rc_min_quantizer,
159  width, "rc_max_quantizer:", cfg->rc_max_quantizer);
160  av_log(avctx, level, "bitrate tolerance\n"
161  " %*s%u\n %*s%u\n",
162  width, "rc_undershoot_pct:", cfg->rc_undershoot_pct,
163  width, "rc_overshoot_pct:", cfg->rc_overshoot_pct);
164  av_log(avctx, level, "decoder buffer model\n"
165  " %*s%u\n %*s%u\n %*s%u\n",
166  width, "rc_buf_sz:", cfg->rc_buf_sz,
167  width, "rc_buf_initial_sz:", cfg->rc_buf_initial_sz,
168  width, "rc_buf_optimal_sz:", cfg->rc_buf_optimal_sz);
169  av_log(avctx, level, "2 pass rate control settings\n"
170  " %*s%u\n %*s%u\n %*s%u\n",
171  width, "rc_2pass_vbr_bias_pct:", cfg->rc_2pass_vbr_bias_pct,
172  width, "rc_2pass_vbr_minsection_pct:", cfg->rc_2pass_vbr_minsection_pct,
173  width, "rc_2pass_vbr_maxsection_pct:", cfg->rc_2pass_vbr_maxsection_pct);
174  av_log(avctx, level, "keyframing settings\n"
175  " %*s%d\n %*s%u\n %*s%u\n",
176  width, "kf_mode:", cfg->kf_mode,
177  width, "kf_min_dist:", cfg->kf_min_dist,
178  width, "kf_max_dist:", cfg->kf_max_dist);
179  av_log(avctx, level, "\n");
180 }
181 
182 static void coded_frame_add(void *list, struct FrameListData *cx_frame)
183 {
184  struct FrameListData **p = list;
185 
186  while (*p != NULL)
187  p = &(*p)->next;
188  *p = cx_frame;
189  cx_frame->next = NULL;
190 }
191 
192 static av_cold void free_coded_frame(struct FrameListData *cx_frame)
193 {
194  av_freep(&cx_frame->buf);
195  if (cx_frame->buf_alpha)
196  av_freep(&cx_frame->buf_alpha);
197  av_freep(&cx_frame);
198 }
199 
200 static av_cold void free_frame_list(struct FrameListData *list)
201 {
202  struct FrameListData *p = list;
203 
204  while (p) {
205  list = list->next;
206  free_coded_frame(p);
207  p = list;
208  }
209 }
210 
212  enum vp8e_enc_control_id id, int val)
213 {
214  VP8Context *ctx = avctx->priv_data;
215  char buf[80];
216  int width = -30;
217  int res;
218 
219  snprintf(buf, sizeof(buf), "%s:", ctlidstr[id]);
220  av_log(avctx, AV_LOG_DEBUG, " %*s%d\n", width, buf, val);
221 
222  res = vpx_codec_control(&ctx->encoder, id, val);
223  if (res != VPX_CODEC_OK) {
224  snprintf(buf, sizeof(buf), "Failed to set %s codec control",
225  ctlidstr[id]);
226  log_encoder_error(avctx, buf);
227  }
228 
229  return res == VPX_CODEC_OK ? 0 : AVERROR(EINVAL);
230 }
231 
232 static av_cold int vp8_free(AVCodecContext *avctx)
233 {
234  VP8Context *ctx = avctx->priv_data;
235 
236  vpx_codec_destroy(&ctx->encoder);
237  if (ctx->is_alpha)
238  vpx_codec_destroy(&ctx->encoder_alpha);
239  av_freep(&ctx->twopass_stats.buf);
240  av_freep(&avctx->coded_frame);
241  av_freep(&avctx->stats_out);
243  return 0;
244 }
245 
246 static av_cold int vpx_init(AVCodecContext *avctx,
247  const struct vpx_codec_iface *iface)
248 {
249  VP8Context *ctx = avctx->priv_data;
250  struct vpx_codec_enc_cfg enccfg;
251  struct vpx_codec_enc_cfg enccfg_alpha;
252  vpx_codec_flags_t flags = (avctx->flags & CODEC_FLAG_PSNR) ? VPX_CODEC_USE_PSNR : 0;
253  int res;
254 
255  av_log(avctx, AV_LOG_INFO, "%s\n", vpx_codec_version_str());
256  av_log(avctx, AV_LOG_VERBOSE, "%s\n", vpx_codec_build_config());
257 
258  if (avctx->pix_fmt == AV_PIX_FMT_YUVA420P)
259  ctx->is_alpha = 1;
260 
261  if ((res = vpx_codec_enc_config_default(iface, &enccfg, 0)) != VPX_CODEC_OK) {
262  av_log(avctx, AV_LOG_ERROR, "Failed to get config: %s\n",
263  vpx_codec_err_to_string(res));
264  return AVERROR(EINVAL);
265  }
266 
267  if(!avctx->bit_rate)
268  if(avctx->rc_max_rate || avctx->rc_buffer_size || avctx->rc_initial_buffer_occupancy) {
269  av_log( avctx, AV_LOG_ERROR, "Rate control parameters set without a bitrate\n");
270  return AVERROR(EINVAL);
271  }
272 
273  dump_enc_cfg(avctx, &enccfg);
274 
275  enccfg.g_w = avctx->width;
276  enccfg.g_h = avctx->height;
277  enccfg.g_timebase.num = avctx->time_base.num;
278  enccfg.g_timebase.den = avctx->time_base.den;
279  enccfg.g_threads = avctx->thread_count;
280  enccfg.g_lag_in_frames= ctx->lag_in_frames;
281 
282  if (avctx->flags & CODEC_FLAG_PASS1)
283  enccfg.g_pass = VPX_RC_FIRST_PASS;
284  else if (avctx->flags & CODEC_FLAG_PASS2)
285  enccfg.g_pass = VPX_RC_LAST_PASS;
286  else
287  enccfg.g_pass = VPX_RC_ONE_PASS;
288 
289  if (avctx->rc_min_rate == avctx->rc_max_rate &&
290  avctx->rc_min_rate == avctx->bit_rate && avctx->bit_rate)
291  enccfg.rc_end_usage = VPX_CBR;
292  else if (ctx->crf)
293  enccfg.rc_end_usage = VPX_CQ;
294 
295  if (avctx->bit_rate) {
296  enccfg.rc_target_bitrate = av_rescale_rnd(avctx->bit_rate, 1, 1000,
298  } else {
299  if (enccfg.rc_end_usage == VPX_CQ) {
300  enccfg.rc_target_bitrate = 1000000;
301  } else {
302  avctx->bit_rate = enccfg.rc_target_bitrate * 1000;
303  av_log(avctx, AV_LOG_WARNING,
304  "Neither bitrate nor constrained quality specified, using default bitrate of %dkbit/sec\n",
305  enccfg.rc_target_bitrate);
306  }
307  }
308 
309  if (avctx->qmin >= 0)
310  enccfg.rc_min_quantizer = avctx->qmin;
311  if (avctx->qmax > 0)
312  enccfg.rc_max_quantizer = avctx->qmax;
313 
314  if (enccfg.rc_end_usage == VPX_CQ) {
315  if (ctx->crf < enccfg.rc_min_quantizer || ctx->crf > enccfg.rc_max_quantizer) {
316  av_log(avctx, AV_LOG_ERROR,
317  "CQ level must be between minimum and maximum quantizer value (%d-%d)\n",
318  enccfg.rc_min_quantizer, enccfg.rc_max_quantizer);
319  return AVERROR(EINVAL);
320  }
321  }
322 
323  enccfg.rc_dropframe_thresh = avctx->frame_skip_threshold;
324 
325  //0-100 (0 => CBR, 100 => VBR)
326  enccfg.rc_2pass_vbr_bias_pct = round(avctx->qcompress * 100);
327  if (avctx->bit_rate)
328  enccfg.rc_2pass_vbr_minsection_pct =
329  avctx->rc_min_rate * 100LL / avctx->bit_rate;
330  if (avctx->rc_max_rate)
331  enccfg.rc_2pass_vbr_maxsection_pct =
332  avctx->rc_max_rate * 100LL / avctx->bit_rate;
333 
334  if (avctx->rc_buffer_size)
335  enccfg.rc_buf_sz =
336  avctx->rc_buffer_size * 1000LL / avctx->bit_rate;
337  if (avctx->rc_initial_buffer_occupancy)
338  enccfg.rc_buf_initial_sz =
339  avctx->rc_initial_buffer_occupancy * 1000LL / avctx->bit_rate;
340  enccfg.rc_buf_optimal_sz = enccfg.rc_buf_sz * 5 / 6;
341  enccfg.rc_undershoot_pct = round(avctx->rc_buffer_aggressivity * 100);
342 
343  //_enc_init() will balk if kf_min_dist differs from max w/VPX_KF_AUTO
344  if (avctx->keyint_min >= 0 && avctx->keyint_min == avctx->gop_size)
345  enccfg.kf_min_dist = avctx->keyint_min;
346  if (avctx->gop_size >= 0)
347  enccfg.kf_max_dist = avctx->gop_size;
348 
349  if (enccfg.g_pass == VPX_RC_FIRST_PASS)
350  enccfg.g_lag_in_frames = 0;
351  else if (enccfg.g_pass == VPX_RC_LAST_PASS) {
352  int decode_size;
353 
354  if (!avctx->stats_in) {
355  av_log(avctx, AV_LOG_ERROR, "No stats file for second pass\n");
356  return AVERROR_INVALIDDATA;
357  }
358 
359  ctx->twopass_stats.sz = strlen(avctx->stats_in) * 3 / 4;
360  ctx->twopass_stats.buf = av_malloc(ctx->twopass_stats.sz);
361  if (!ctx->twopass_stats.buf) {
362  av_log(avctx, AV_LOG_ERROR,
363  "Stat buffer alloc (%zu bytes) failed\n",
364  ctx->twopass_stats.sz);
365  return AVERROR(ENOMEM);
366  }
367  decode_size = av_base64_decode(ctx->twopass_stats.buf, avctx->stats_in,
368  ctx->twopass_stats.sz);
369  if (decode_size < 0) {
370  av_log(avctx, AV_LOG_ERROR, "Stat buffer decode failed\n");
371  return AVERROR_INVALIDDATA;
372  }
373 
374  ctx->twopass_stats.sz = decode_size;
375  enccfg.rc_twopass_stats_in = ctx->twopass_stats;
376  }
377 
378  /* 0-3: For non-zero values the encoder increasingly optimizes for reduced
379  complexity playback on low powered devices at the expense of encode
380  quality. */
381  if (avctx->profile != FF_PROFILE_UNKNOWN)
382  enccfg.g_profile = avctx->profile;
383 
384  enccfg.g_error_resilient = ctx->error_resilient || ctx->flags & VP8F_ERROR_RESILIENT;
385 
386  dump_enc_cfg(avctx, &enccfg);
387  /* Construct Encoder Context */
388  res = vpx_codec_enc_init(&ctx->encoder, iface, &enccfg, flags);
389  if (res != VPX_CODEC_OK) {
390  log_encoder_error(avctx, "Failed to initialize encoder");
391  return AVERROR(EINVAL);
392  }
393 
394  if (ctx->is_alpha) {
395  enccfg_alpha = enccfg;
396  res = vpx_codec_enc_init(&ctx->encoder_alpha, iface, &enccfg_alpha, flags);
397  if (res != VPX_CODEC_OK) {
398  log_encoder_error(avctx, "Failed to initialize alpha encoder");
399  return AVERROR(EINVAL);
400  }
401  }
402 
403  //codec control failures are currently treated only as warnings
404  av_log(avctx, AV_LOG_DEBUG, "vpx_codec_control\n");
405  if (ctx->cpu_used != INT_MIN)
406  codecctl_int(avctx, VP8E_SET_CPUUSED, ctx->cpu_used);
407  if (ctx->flags & VP8F_AUTO_ALT_REF)
408  ctx->auto_alt_ref = 1;
409  if (ctx->auto_alt_ref >= 0)
410  codecctl_int(avctx, VP8E_SET_ENABLEAUTOALTREF, ctx->auto_alt_ref);
411  if (ctx->arnr_max_frames >= 0)
412  codecctl_int(avctx, VP8E_SET_ARNR_MAXFRAMES, ctx->arnr_max_frames);
413  if (ctx->arnr_strength >= 0)
414  codecctl_int(avctx, VP8E_SET_ARNR_STRENGTH, ctx->arnr_strength);
415  if (ctx->arnr_type >= 0)
416  codecctl_int(avctx, VP8E_SET_ARNR_TYPE, ctx->arnr_type);
417  codecctl_int(avctx, VP8E_SET_NOISE_SENSITIVITY, avctx->noise_reduction);
418  codecctl_int(avctx, VP8E_SET_TOKEN_PARTITIONS, av_log2(avctx->slices));
419  codecctl_int(avctx, VP8E_SET_STATIC_THRESHOLD, avctx->mb_threshold);
420  codecctl_int(avctx, VP8E_SET_CQ_LEVEL, ctx->crf);
421  if (ctx->max_intra_rate >= 0)
422  codecctl_int(avctx, VP8E_SET_MAX_INTRA_BITRATE_PCT, ctx->max_intra_rate);
423 
424  av_log(avctx, AV_LOG_DEBUG, "Using deadline: %d\n", ctx->deadline);
425 
426  //provide dummy value to initialize wrapper, values will be updated each _encode()
427  vpx_img_wrap(&ctx->rawimg, VPX_IMG_FMT_I420, avctx->width, avctx->height, 1,
428  (unsigned char*)1);
429 
430  if (ctx->is_alpha)
431  vpx_img_wrap(&ctx->rawimg_alpha, VPX_IMG_FMT_I420, avctx->width, avctx->height, 1,
432  (unsigned char*)1);
433 
434  avctx->coded_frame = avcodec_alloc_frame();
435  if (!avctx->coded_frame) {
436  av_log(avctx, AV_LOG_ERROR, "Error allocating coded frame\n");
437  vp8_free(avctx);
438  return AVERROR(ENOMEM);
439  }
440  return 0;
441 }
442 
443 static inline void cx_pktcpy(struct FrameListData *dst,
444  const struct vpx_codec_cx_pkt *src,
445  const struct vpx_codec_cx_pkt *src_alpha,
446  VP8Context *ctx)
447 {
448  dst->pts = src->data.frame.pts;
449  dst->duration = src->data.frame.duration;
450  dst->flags = src->data.frame.flags;
451  dst->sz = src->data.frame.sz;
452  dst->buf = src->data.frame.buf;
453  dst->have_sse = 0;
454  /* For alt-ref frame, don't store PSNR or increment frame_number */
455  if (!(dst->flags & VPX_FRAME_IS_INVISIBLE)) {
456  dst->frame_number = ++ctx->frame_number;
457  dst->have_sse = ctx->have_sse;
458  if (ctx->have_sse) {
459  /* associate last-seen SSE to the frame. */
460  /* Transfers ownership from ctx to dst. */
461  /* WARNING! This makes the assumption that PSNR_PKT comes
462  just before the frame it refers to! */
463  memcpy(dst->sse, ctx->sse, sizeof(dst->sse));
464  ctx->have_sse = 0;
465  }
466  } else {
467  dst->frame_number = -1; /* sanity marker */
468  }
469  if (src_alpha) {
470  dst->buf_alpha = src_alpha->data.frame.buf;
471  dst->sz_alpha = src_alpha->data.frame.sz;
472  }
473  else {
474  dst->buf_alpha = NULL;
475  dst->sz_alpha = 0;
476  }
477 }
478 
479 /**
480  * Store coded frame information in format suitable for return from encode2().
481  *
482  * Write information from @a cx_frame to @a pkt
483  * @return packet data size on success
484  * @return a negative AVERROR on error
485  */
486 static int storeframe(AVCodecContext *avctx, struct FrameListData *cx_frame,
487  AVPacket *pkt, AVFrame *coded_frame)
488 {
489  int ret = ff_alloc_packet2(avctx, pkt, cx_frame->sz);
490  uint8_t *side_data;
491  if (ret >= 0) {
492  memcpy(pkt->data, cx_frame->buf, pkt->size);
493  pkt->pts = pkt->dts = cx_frame->pts;
494  coded_frame->pts = cx_frame->pts;
495  coded_frame->key_frame = !!(cx_frame->flags & VPX_FRAME_IS_KEY);
496 
497  if (coded_frame->key_frame) {
498  coded_frame->pict_type = AV_PICTURE_TYPE_I;
499  pkt->flags |= AV_PKT_FLAG_KEY;
500  } else
501  coded_frame->pict_type = AV_PICTURE_TYPE_P;
502 
503  if (cx_frame->have_sse) {
504  int i;
505  /* Beware of the Y/U/V/all order! */
506  coded_frame->error[0] = cx_frame->sse[1];
507  coded_frame->error[1] = cx_frame->sse[2];
508  coded_frame->error[2] = cx_frame->sse[3];
509  coded_frame->error[3] = 0; // alpha
510  for (i = 0; i < 4; ++i) {
511  avctx->error[i] += coded_frame->error[i];
512  }
513  cx_frame->have_sse = 0;
514  }
515  if (cx_frame->sz_alpha > 0) {
516  side_data = av_packet_new_side_data(pkt,
518  cx_frame->sz_alpha + 8);
519  if(side_data == NULL) {
520  av_free_packet(pkt);
521  av_free(pkt);
522  return AVERROR(ENOMEM);
523  }
524  AV_WB64(side_data, 1);
525  memcpy(side_data + 8, cx_frame->buf_alpha, cx_frame->sz_alpha);
526  }
527  } else {
528  return ret;
529  }
530  return pkt->size;
531 }
532 
533 /**
534  * Queue multiple output frames from the encoder, returning the front-most.
535  * In cases where vpx_codec_get_cx_data() returns more than 1 frame append
536  * the frame queue. Return the head frame if available.
537  * @return Stored frame size
538  * @return AVERROR(EINVAL) on output size error
539  * @return AVERROR(ENOMEM) on coded frame queue data allocation error
540  */
541 static int queue_frames(AVCodecContext *avctx, AVPacket *pkt_out,
542  AVFrame *coded_frame)
543 {
544  VP8Context *ctx = avctx->priv_data;
545  const struct vpx_codec_cx_pkt *pkt;
546  const struct vpx_codec_cx_pkt *pkt_alpha = NULL;
547  const void *iter = NULL;
548  const void *iter_alpha = NULL;
549  int size = 0;
550 
551  if (ctx->coded_frame_list) {
552  struct FrameListData *cx_frame = ctx->coded_frame_list;
553  /* return the leading frame if we've already begun queueing */
554  size = storeframe(avctx, cx_frame, pkt_out, coded_frame);
555  if (size < 0)
556  return size;
557  ctx->coded_frame_list = cx_frame->next;
558  free_coded_frame(cx_frame);
559  }
560 
561  /* consume all available output from the encoder before returning. buffers
562  are only good through the next vpx_codec call */
563  while ((pkt = vpx_codec_get_cx_data(&ctx->encoder, &iter)) &&
564  (!ctx->is_alpha ||
565  (ctx->is_alpha && (pkt_alpha = vpx_codec_get_cx_data(&ctx->encoder_alpha, &iter_alpha))))) {
566  switch (pkt->kind) {
567  case VPX_CODEC_CX_FRAME_PKT:
568  if (!size) {
569  struct FrameListData cx_frame;
570 
571  /* avoid storing the frame when the list is empty and we haven't yet
572  provided a frame for output */
574  cx_pktcpy(&cx_frame, pkt, pkt_alpha, ctx);
575  size = storeframe(avctx, &cx_frame, pkt_out, coded_frame);
576  if (size < 0)
577  return size;
578  } else {
579  struct FrameListData *cx_frame =
580  av_malloc(sizeof(struct FrameListData));
581 
582  if (!cx_frame) {
583  av_log(avctx, AV_LOG_ERROR,
584  "Frame queue element alloc failed\n");
585  return AVERROR(ENOMEM);
586  }
587  cx_pktcpy(cx_frame, pkt, pkt_alpha, ctx);
588  cx_frame->buf = av_malloc(cx_frame->sz);
589 
590  if (!cx_frame->buf) {
591  av_log(avctx, AV_LOG_ERROR,
592  "Data buffer alloc (%zu bytes) failed\n",
593  cx_frame->sz);
594  av_free(cx_frame);
595  return AVERROR(ENOMEM);
596  }
597  memcpy(cx_frame->buf, pkt->data.frame.buf, pkt->data.frame.sz);
598  if (ctx->is_alpha) {
599  cx_frame->buf_alpha = av_malloc(cx_frame->sz_alpha);
600  if (!cx_frame->buf_alpha) {
601  av_log(avctx, AV_LOG_ERROR,
602  "Data buffer alloc (%zu bytes) failed\n",
603  cx_frame->sz_alpha);
604  av_free(cx_frame);
605  return AVERROR(ENOMEM);
606  }
607  memcpy(cx_frame->buf_alpha, pkt_alpha->data.frame.buf, pkt_alpha->data.frame.sz);
608  }
609  coded_frame_add(&ctx->coded_frame_list, cx_frame);
610  }
611  break;
612  case VPX_CODEC_STATS_PKT: {
613  struct vpx_fixed_buf *stats = &ctx->twopass_stats;
614  stats->buf = av_realloc_f(stats->buf, 1,
615  stats->sz + pkt->data.twopass_stats.sz);
616  if (!stats->buf) {
617  av_log(avctx, AV_LOG_ERROR, "Stat buffer realloc failed\n");
618  return AVERROR(ENOMEM);
619  }
620  memcpy((uint8_t*)stats->buf + stats->sz,
621  pkt->data.twopass_stats.buf, pkt->data.twopass_stats.sz);
622  stats->sz += pkt->data.twopass_stats.sz;
623  break;
624  }
625  case VPX_CODEC_PSNR_PKT:
626  av_assert0(!ctx->have_sse);
627  ctx->sse[0] = pkt->data.psnr.sse[0];
628  ctx->sse[1] = pkt->data.psnr.sse[1];
629  ctx->sse[2] = pkt->data.psnr.sse[2];
630  ctx->sse[3] = pkt->data.psnr.sse[3];
631  ctx->have_sse = 1;
632  break;
633  case VPX_CODEC_CUSTOM_PKT:
634  //ignore unsupported/unrecognized packet types
635  break;
636  }
637  }
638 
639  return size;
640 }
641 
642 static int vp8_encode(AVCodecContext *avctx, AVPacket *pkt,
643  const AVFrame *frame, int *got_packet)
644 {
645  VP8Context *ctx = avctx->priv_data;
646  struct vpx_image *rawimg = NULL;
647  struct vpx_image *rawimg_alpha = NULL;
648  int64_t timestamp = 0;
649  int res, coded_size;
650  vpx_enc_frame_flags_t flags = 0;
651 
652  if (frame) {
653  rawimg = &ctx->rawimg;
654  rawimg->planes[VPX_PLANE_Y] = frame->data[0];
655  rawimg->planes[VPX_PLANE_U] = frame->data[1];
656  rawimg->planes[VPX_PLANE_V] = frame->data[2];
657  rawimg->stride[VPX_PLANE_Y] = frame->linesize[0];
658  rawimg->stride[VPX_PLANE_U] = frame->linesize[1];
659  rawimg->stride[VPX_PLANE_V] = frame->linesize[2];
660  if (ctx->is_alpha) {
661  uint8_t *u_plane, *v_plane;
662  rawimg_alpha = &ctx->rawimg_alpha;
663  rawimg_alpha->planes[VPX_PLANE_Y] = frame->data[3];
664  u_plane = av_malloc(frame->linesize[1] * frame->height);
665  memset(u_plane, 0x80, frame->linesize[1] * frame->height);
666  rawimg_alpha->planes[VPX_PLANE_U] = u_plane;
667  v_plane = av_malloc(frame->linesize[2] * frame->height);
668  memset(v_plane, 0x80, frame->linesize[2] * frame->height);
669  rawimg_alpha->planes[VPX_PLANE_V] = v_plane;
670  rawimg_alpha->stride[VPX_PLANE_Y] = frame->linesize[0];
671  rawimg_alpha->stride[VPX_PLANE_U] = frame->linesize[1];
672  rawimg_alpha->stride[VPX_PLANE_V] = frame->linesize[2];
673  }
674  timestamp = frame->pts;
675  if (frame->pict_type == AV_PICTURE_TYPE_I)
676  flags |= VPX_EFLAG_FORCE_KF;
677  }
678 
679  res = vpx_codec_encode(&ctx->encoder, rawimg, timestamp,
680  avctx->ticks_per_frame, flags, ctx->deadline);
681  if (res != VPX_CODEC_OK) {
682  log_encoder_error(avctx, "Error encoding frame");
683  return AVERROR_INVALIDDATA;
684  }
685 
686  if (ctx->is_alpha) {
687  res = vpx_codec_encode(&ctx->encoder_alpha, rawimg_alpha, timestamp,
688  avctx->ticks_per_frame, flags, ctx->deadline);
689  if (res != VPX_CODEC_OK) {
690  log_encoder_error(avctx, "Error encoding alpha frame");
691  return AVERROR_INVALIDDATA;
692  }
693  }
694 
695  coded_size = queue_frames(avctx, pkt, avctx->coded_frame);
696 
697  if (!frame && avctx->flags & CODEC_FLAG_PASS1) {
698  unsigned int b64_size = AV_BASE64_SIZE(ctx->twopass_stats.sz);
699 
700  avctx->stats_out = av_malloc(b64_size);
701  if (!avctx->stats_out) {
702  av_log(avctx, AV_LOG_ERROR, "Stat buffer alloc (%d bytes) failed\n",
703  b64_size);
704  return AVERROR(ENOMEM);
705  }
706  av_base64_encode(avctx->stats_out, b64_size, ctx->twopass_stats.buf,
707  ctx->twopass_stats.sz);
708  }
709 
710  if (rawimg_alpha) {
711  av_free(rawimg_alpha->planes[VPX_PLANE_U]);
712  av_free(rawimg_alpha->planes[VPX_PLANE_V]);
713  }
714 
715  *got_packet = !!coded_size;
716  return 0;
717 }
718 
719 #define OFFSET(x) offsetof(VP8Context, x)
720 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
721 static const AVOption options[] = {
722  { "cpu-used", "Quality/Speed ratio modifier", OFFSET(cpu_used), AV_OPT_TYPE_INT, {.i64 = INT_MIN}, INT_MIN, INT_MAX, VE},
723  { "auto-alt-ref", "Enable use of alternate reference "
724  "frames (2-pass only)", OFFSET(auto_alt_ref), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE},
725  { "lag-in-frames", "Number of frames to look ahead for "
726  "alternate reference frame selection", OFFSET(lag_in_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE},
727  { "arnr-maxframes", "altref noise reduction max frame count", OFFSET(arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE},
728  { "arnr-strength", "altref noise reduction filter strength", OFFSET(arnr_strength), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE},
729  { "arnr-type", "altref noise reduction filter type", OFFSET(arnr_type), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE, "arnr_type"},
730  { "backward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0, VE, "arnr_type" },
731  { "forward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0, VE, "arnr_type" },
732  { "centered", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 3}, 0, 0, VE, "arnr_type" },
733  { "deadline", "Time to spend encoding, in microseconds.", OFFSET(deadline), AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, "quality"},
734  { "best", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_BEST_QUALITY}, 0, 0, VE, "quality"},
735  { "good", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_GOOD_QUALITY}, 0, 0, VE, "quality"},
736  { "realtime", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_REALTIME}, 0, 0, VE, "quality"},
737  { "error-resilient", "Error resilience configuration", OFFSET(error_resilient), AV_OPT_TYPE_FLAGS, {.i64 = 0}, INT_MIN, INT_MAX, VE, "er"},
738  { "max-intra-rate", "Maximum I-frame bitrate (pct) 0=unlimited", OFFSET(max_intra_rate), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE},
739 #ifdef VPX_ERROR_RESILIENT_DEFAULT
740  { "default", "Improve resiliency against losses of whole frames", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_DEFAULT}, 0, 0, VE, "er"},
741  { "partitions", "The frame partitions are independently decodable "
742  "by the bool decoder, meaning that partitions can be decoded even "
743  "though earlier partitions have been lost. Note that intra predicition"
744  " is still done over the partition boundary.", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_PARTITIONS}, 0, 0, VE, "er"},
745 #endif
746 {"speed", "", offsetof(VP8Context, cpu_used), AV_OPT_TYPE_INT, {.i64 = 3}, -16, 16, VE},
747 {"quality", "", offsetof(VP8Context, deadline), AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, "quality"},
748 {"vp8flags", "", offsetof(VP8Context, flags), FF_OPT_TYPE_FLAGS, {.i64 = 0}, 0, UINT_MAX, VE, "flags"},
749 {"error_resilient", "enable error resilience", 0, FF_OPT_TYPE_CONST, {.dbl = VP8F_ERROR_RESILIENT}, INT_MIN, INT_MAX, VE, "flags"},
750 {"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"},
751 {"arnr_max_frames", "altref noise reduction max frame count", offsetof(VP8Context, arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 15, VE},
752 {"arnr_strength", "altref noise reduction filter strength", offsetof(VP8Context, arnr_strength), AV_OPT_TYPE_INT, {.i64 = 3}, 0, 6, VE},
753 {"arnr_type", "altref noise reduction filter type", offsetof(VP8Context, arnr_type), AV_OPT_TYPE_INT, {.i64 = 3}, 1, 3, VE},
754 {"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},
755  { "crf", "Select the quality for constant quality mode", offsetof(VP8Context, crf), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 63, VE },
756  { NULL }
757 };
758 
759 static const AVCodecDefault defaults[] = {
760  { "qmin", "-1" },
761  { "qmax", "-1" },
762  { "g", "-1" },
763  { "keyint_min", "-1" },
764  { NULL },
765 };
766 
767 #if CONFIG_LIBVPX_VP8_ENCODER
768 static av_cold int vp8_init(AVCodecContext *avctx)
769 {
770  return vpx_init(avctx, &vpx_codec_vp8_cx_algo);
771 }
772 
773 static const AVClass class_vp8 = {
774  .class_name = "libvpx-vp8 encoder",
775  .item_name = av_default_item_name,
776  .option = options,
777  .version = LIBAVUTIL_VERSION_INT,
778 };
779 
780 AVCodec ff_libvpx_vp8_encoder = {
781  .name = "libvpx",
782  .long_name = NULL_IF_CONFIG_SMALL("libvpx VP8"),
783  .type = AVMEDIA_TYPE_VIDEO,
784  .id = AV_CODEC_ID_VP8,
785  .priv_data_size = sizeof(VP8Context),
786  .init = vp8_init,
787  .encode2 = vp8_encode,
788  .close = vp8_free,
789  .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
791  .priv_class = &class_vp8,
792  .defaults = defaults,
793 };
794 #endif /* CONFIG_LIBVPX_VP8_ENCODER */
795 
796 #if CONFIG_LIBVPX_VP9_ENCODER
797 static av_cold int vp9_init(AVCodecContext *avctx)
798 {
799  return vpx_init(avctx, &vpx_codec_vp9_cx_algo);
800 }
801 
802 static const AVClass class_vp9 = {
803  .class_name = "libvpx-vp9 encoder",
804  .item_name = av_default_item_name,
805  .option = options,
806  .version = LIBAVUTIL_VERSION_INT,
807 };
808 
809 AVCodec ff_libvpx_vp9_encoder = {
810  .name = "libvpx-vp9",
811  .long_name = NULL_IF_CONFIG_SMALL("libvpx VP9"),
812  .type = AVMEDIA_TYPE_VIDEO,
813  .id = AV_CODEC_ID_VP9,
814  .priv_data_size = sizeof(VP8Context),
815  .init = vp9_init,
816  .encode2 = vp8_encode,
817  .close = vp8_free,
819  .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE },
820  .priv_class = &class_vp9,
821  .defaults = defaults,
822 };
823 #endif /* CONFIG_LIBVPX_VP9_ENCODER */