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 "libvpx.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"
40 
41 /**
42  * Portion of struct vpx_codec_cx_pkt from vpx_encoder.h.
43  * One encoded frame returned from the library.
44  */
45 struct FrameListData {
46  void *buf; /**< compressed data buffer */
47  size_t sz; /**< length of compressed data */
48  void *buf_alpha;
49  size_t sz_alpha;
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 */
55  uint64_t sse[4];
56  int have_sse; /**< true if we have pending sse[] */
57  uint64_t frame_number;
59 };
60 
61 typedef struct VP8EncoderContext {
62  AVClass *class;
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
70  uint64_t sse[4];
71  int have_sse; /**< true if we have pending sse[] */
72  uint64_t frame_number;
74 
75  int cpu_used;
76  /**
77  * VP8 specific flags, see VP8F_* below.
78  */
79  int flags;
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
82 
84 
87  int arnr_type;
88 
91  int crf;
94 
95  // VP9-only
96  int lossless;
98  int tile_rows;
100  int aq_mode;
101 } VP8Context;
102 
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",
129 #endif
130 };
131 
132 static av_cold void log_encoder_error(AVCodecContext *avctx, const char *desc)
133 {
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);
137 
138  av_log(avctx, AV_LOG_ERROR, "%s: %s\n", desc, error);
139  if (detail)
140  av_log(avctx, AV_LOG_ERROR, " Additional information: %s\n", detail);
141 }
142 
144  const struct vpx_codec_enc_cfg *cfg)
145 {
146  int width = -30;
147  int level = AV_LOG_DEBUG;
148 
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"
173  " %*s%u\n %*s%u\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"
177  " %*s%u\n %*s%u\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");
196 }
197 
198 static void coded_frame_add(void *list, struct FrameListData *cx_frame)
199 {
200  struct FrameListData **p = list;
201 
202  while (*p)
203  p = &(*p)->next;
204  *p = cx_frame;
205  cx_frame->next = NULL;
206 }
207 
208 static av_cold void free_coded_frame(struct FrameListData *cx_frame)
209 {
210  av_freep(&cx_frame->buf);
211  if (cx_frame->buf_alpha)
212  av_freep(&cx_frame->buf_alpha);
213  av_freep(&cx_frame);
214 }
215 
216 static av_cold void free_frame_list(struct FrameListData *list)
217 {
218  struct FrameListData *p = list;
219 
220  while (p) {
221  list = list->next;
222  free_coded_frame(p);
223  p = list;
224  }
225 }
226 
228  enum vp8e_enc_control_id id, int val)
229 {
230  VP8Context *ctx = avctx->priv_data;
231  char buf[80];
232  int width = -30;
233  int res;
234 
235  snprintf(buf, sizeof(buf), "%s:", ctlidstr[id]);
236  av_log(avctx, AV_LOG_DEBUG, " %*s%d\n", width, buf, val);
237 
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",
241  ctlidstr[id]);
242  log_encoder_error(avctx, buf);
243  }
244 
245  return res == VPX_CODEC_OK ? 0 : AVERROR(EINVAL);
246 }
247 
248 static av_cold int vp8_free(AVCodecContext *avctx)
249 {
250  VP8Context *ctx = avctx->priv_data;
251 
252  vpx_codec_destroy(&ctx->encoder);
253  if (ctx->is_alpha)
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);
259  return 0;
260 }
261 
262 static av_cold int vpx_init(AVCodecContext *avctx,
263  const struct vpx_codec_iface *iface)
264 {
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;
269  int res;
270 
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());
273 
274  if (avctx->pix_fmt == AV_PIX_FMT_YUVA420P)
275  ctx->is_alpha = 1;
276 
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);
281  }
282 
283  if(!avctx->bit_rate)
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);
287  }
288 
289  dump_enc_cfg(avctx, &enccfg);
290 
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;
297 
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;
302  else
303  enccfg.g_pass = VPX_RC_ONE_PASS;
304 
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;
313 #endif
314  }
315 
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) {
321 #endif
322  } else {
323  if (enccfg.rc_end_usage == VPX_CQ) {
324  enccfg.rc_target_bitrate = 1000000;
325  } else {
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);
330  }
331  }
332 
333  if (avctx->qmin >= 0)
334  enccfg.rc_min_quantizer = avctx->qmin;
335  if (avctx->qmax >= 0)
336  enccfg.rc_max_quantizer = avctx->qmax;
337 
338  if (enccfg.rc_end_usage == VPX_CQ
339 #if CONFIG_LIBVPX_VP9_ENCODER
340  || enccfg.rc_end_usage == VPX_Q
341 #endif
342  ) {
343  if (ctx->crf < enccfg.rc_min_quantizer || ctx->crf > enccfg.rc_max_quantizer) {
344  av_log(avctx, AV_LOG_ERROR,
345  "CQ level %d must be between minimum and maximum quantizer value (%d-%d)\n",
346  ctx->crf, enccfg.rc_min_quantizer, enccfg.rc_max_quantizer);
347  return AVERROR(EINVAL);
348  }
349  }
350 
351  enccfg.rc_dropframe_thresh = avctx->frame_skip_threshold;
352 
353  //0-100 (0 => CBR, 100 => VBR)
354  enccfg.rc_2pass_vbr_bias_pct = round(avctx->qcompress * 100);
355  if (avctx->bit_rate)
356  enccfg.rc_2pass_vbr_minsection_pct =
357  avctx->rc_min_rate * 100LL / avctx->bit_rate;
358  if (avctx->rc_max_rate)
359  enccfg.rc_2pass_vbr_maxsection_pct =
360  avctx->rc_max_rate * 100LL / avctx->bit_rate;
361 
362  if (avctx->rc_buffer_size)
363  enccfg.rc_buf_sz =
364  avctx->rc_buffer_size * 1000LL / avctx->bit_rate;
365  if (avctx->rc_initial_buffer_occupancy)
366  enccfg.rc_buf_initial_sz =
367  avctx->rc_initial_buffer_occupancy * 1000LL / avctx->bit_rate;
368  enccfg.rc_buf_optimal_sz = enccfg.rc_buf_sz * 5 / 6;
369  enccfg.rc_undershoot_pct = round(avctx->rc_buffer_aggressivity * 100);
370 
371  //_enc_init() will balk if kf_min_dist differs from max w/VPX_KF_AUTO
372  if (avctx->keyint_min >= 0 && avctx->keyint_min == avctx->gop_size)
373  enccfg.kf_min_dist = avctx->keyint_min;
374  if (avctx->gop_size >= 0)
375  enccfg.kf_max_dist = avctx->gop_size;
376 
377  if (enccfg.g_pass == VPX_RC_FIRST_PASS)
378  enccfg.g_lag_in_frames = 0;
379  else if (enccfg.g_pass == VPX_RC_LAST_PASS) {
380  int decode_size;
381 
382  if (!avctx->stats_in) {
383  av_log(avctx, AV_LOG_ERROR, "No stats file for second pass\n");
384  return AVERROR_INVALIDDATA;
385  }
386 
387  ctx->twopass_stats.sz = strlen(avctx->stats_in) * 3 / 4;
388  ctx->twopass_stats.buf = av_malloc(ctx->twopass_stats.sz);
389  if (!ctx->twopass_stats.buf) {
390  av_log(avctx, AV_LOG_ERROR,
391  "Stat buffer alloc (%"SIZE_SPECIFIER" bytes) failed\n",
392  ctx->twopass_stats.sz);
393  return AVERROR(ENOMEM);
394  }
395  decode_size = av_base64_decode(ctx->twopass_stats.buf, avctx->stats_in,
396  ctx->twopass_stats.sz);
397  if (decode_size < 0) {
398  av_log(avctx, AV_LOG_ERROR, "Stat buffer decode failed\n");
399  return AVERROR_INVALIDDATA;
400  }
401 
402  ctx->twopass_stats.sz = decode_size;
403  enccfg.rc_twopass_stats_in = ctx->twopass_stats;
404  }
405 
406  /* 0-3: For non-zero values the encoder increasingly optimizes for reduced
407  complexity playback on low powered devices at the expense of encode
408  quality. */
409  if (avctx->profile != FF_PROFILE_UNKNOWN)
410  enccfg.g_profile = avctx->profile;
411 
412  enccfg.g_error_resilient = ctx->error_resilient || ctx->flags & VP8F_ERROR_RESILIENT;
413 
414  dump_enc_cfg(avctx, &enccfg);
415  /* Construct Encoder Context */
416  res = vpx_codec_enc_init(&ctx->encoder, iface, &enccfg, flags);
417  if (res != VPX_CODEC_OK) {
418  log_encoder_error(avctx, "Failed to initialize encoder");
419  return AVERROR(EINVAL);
420  }
421 
422  if (ctx->is_alpha) {
423  enccfg_alpha = enccfg;
424  res = vpx_codec_enc_init(&ctx->encoder_alpha, iface, &enccfg_alpha, flags);
425  if (res != VPX_CODEC_OK) {
426  log_encoder_error(avctx, "Failed to initialize alpha encoder");
427  return AVERROR(EINVAL);
428  }
429  }
430 
431  //codec control failures are currently treated only as warnings
432  av_log(avctx, AV_LOG_DEBUG, "vpx_codec_control\n");
433  codecctl_int(avctx, VP8E_SET_CPUUSED, ctx->cpu_used);
434  if (ctx->flags & VP8F_AUTO_ALT_REF)
435  ctx->auto_alt_ref = 1;
436  if (ctx->auto_alt_ref >= 0)
437  codecctl_int(avctx, VP8E_SET_ENABLEAUTOALTREF, ctx->auto_alt_ref);
438  if (ctx->arnr_max_frames >= 0)
439  codecctl_int(avctx, VP8E_SET_ARNR_MAXFRAMES, ctx->arnr_max_frames);
440  if (ctx->arnr_strength >= 0)
441  codecctl_int(avctx, VP8E_SET_ARNR_STRENGTH, ctx->arnr_strength);
442  if (ctx->arnr_type >= 0)
443  codecctl_int(avctx, VP8E_SET_ARNR_TYPE, ctx->arnr_type);
444  codecctl_int(avctx, VP8E_SET_NOISE_SENSITIVITY, avctx->noise_reduction);
445  if (avctx->codec_id == AV_CODEC_ID_VP8)
446  codecctl_int(avctx, VP8E_SET_TOKEN_PARTITIONS, av_log2(avctx->slices));
447 #if FF_API_MPV_OPT
449  if (avctx->mb_threshold) {
450  av_log(avctx, AV_LOG_WARNING, "The mb_threshold option is deprecated, "
451  "use the static-thresh private option instead.\n");
452  ctx->static_thresh = avctx->mb_threshold;
453  }
455 #endif
456  codecctl_int(avctx, VP8E_SET_STATIC_THRESHOLD, ctx->static_thresh);
457  if (ctx->crf >= 0)
458  codecctl_int(avctx, VP8E_SET_CQ_LEVEL, ctx->crf);
459  if (ctx->max_intra_rate >= 0)
460  codecctl_int(avctx, VP8E_SET_MAX_INTRA_BITRATE_PCT, ctx->max_intra_rate);
461 
462 #if CONFIG_LIBVPX_VP9_ENCODER
463  if (avctx->codec_id == AV_CODEC_ID_VP9) {
464  if (ctx->lossless >= 0)
465  codecctl_int(avctx, VP9E_SET_LOSSLESS, ctx->lossless);
466  if (ctx->tile_columns >= 0)
467  codecctl_int(avctx, VP9E_SET_TILE_COLUMNS, ctx->tile_columns);
468  if (ctx->tile_rows >= 0)
469  codecctl_int(avctx, VP9E_SET_TILE_ROWS, ctx->tile_rows);
470  if (ctx->frame_parallel >= 0)
471  codecctl_int(avctx, VP9E_SET_FRAME_PARALLEL_DECODING, ctx->frame_parallel);
472  if (ctx->aq_mode >= 0)
473  codecctl_int(avctx, VP9E_SET_AQ_MODE, ctx->aq_mode);
474  }
475 #endif
476 
477  av_log(avctx, AV_LOG_DEBUG, "Using deadline: %d\n", ctx->deadline);
478 
479  //provide dummy value to initialize wrapper, values will be updated each _encode()
480  vpx_img_wrap(&ctx->rawimg, VPX_IMG_FMT_I420, avctx->width, avctx->height, 1,
481  (unsigned char*)1);
482 
483  if (ctx->is_alpha)
484  vpx_img_wrap(&ctx->rawimg_alpha, VPX_IMG_FMT_I420, avctx->width, avctx->height, 1,
485  (unsigned char*)1);
486 
487  avctx->coded_frame = av_frame_alloc();
488  if (!avctx->coded_frame) {
489  av_log(avctx, AV_LOG_ERROR, "Error allocating coded frame\n");
490  vp8_free(avctx);
491  return AVERROR(ENOMEM);
492  }
493  return 0;
494 }
495 
496 static inline void cx_pktcpy(struct FrameListData *dst,
497  const struct vpx_codec_cx_pkt *src,
498  const struct vpx_codec_cx_pkt *src_alpha,
499  VP8Context *ctx)
500 {
501  dst->pts = src->data.frame.pts;
502  dst->duration = src->data.frame.duration;
503  dst->flags = src->data.frame.flags;
504  dst->sz = src->data.frame.sz;
505  dst->buf = src->data.frame.buf;
506  dst->have_sse = 0;
507  /* For alt-ref frame, don't store PSNR or increment frame_number */
508  if (!(dst->flags & VPX_FRAME_IS_INVISIBLE)) {
509  dst->frame_number = ++ctx->frame_number;
510  dst->have_sse = ctx->have_sse;
511  if (ctx->have_sse) {
512  /* associate last-seen SSE to the frame. */
513  /* Transfers ownership from ctx to dst. */
514  /* WARNING! This makes the assumption that PSNR_PKT comes
515  just before the frame it refers to! */
516  memcpy(dst->sse, ctx->sse, sizeof(dst->sse));
517  ctx->have_sse = 0;
518  }
519  } else {
520  dst->frame_number = -1; /* sanity marker */
521  }
522  if (src_alpha) {
523  dst->buf_alpha = src_alpha->data.frame.buf;
524  dst->sz_alpha = src_alpha->data.frame.sz;
525  }
526  else {
527  dst->buf_alpha = NULL;
528  dst->sz_alpha = 0;
529  }
530 }
531 
532 /**
533  * Store coded frame information in format suitable for return from encode2().
534  *
535  * Write information from @a cx_frame to @a pkt
536  * @return packet data size on success
537  * @return a negative AVERROR on error
538  */
539 static int storeframe(AVCodecContext *avctx, struct FrameListData *cx_frame,
540  AVPacket *pkt, AVFrame *coded_frame)
541 {
542  int ret = ff_alloc_packet2(avctx, pkt, cx_frame->sz);
543  uint8_t *side_data;
544  if (ret >= 0) {
545  memcpy(pkt->data, cx_frame->buf, pkt->size);
546  pkt->pts = pkt->dts = cx_frame->pts;
547  coded_frame->pts = cx_frame->pts;
548  coded_frame->key_frame = !!(cx_frame->flags & VPX_FRAME_IS_KEY);
549 
550  if (coded_frame->key_frame) {
551  coded_frame->pict_type = AV_PICTURE_TYPE_I;
552  pkt->flags |= AV_PKT_FLAG_KEY;
553  } else
554  coded_frame->pict_type = AV_PICTURE_TYPE_P;
555 
556  if (cx_frame->have_sse) {
557  int i;
558  /* Beware of the Y/U/V/all order! */
559  coded_frame->error[0] = cx_frame->sse[1];
560  coded_frame->error[1] = cx_frame->sse[2];
561  coded_frame->error[2] = cx_frame->sse[3];
562  coded_frame->error[3] = 0; // alpha
563  for (i = 0; i < 4; ++i) {
564  avctx->error[i] += coded_frame->error[i];
565  }
566  cx_frame->have_sse = 0;
567  }
568  if (cx_frame->sz_alpha > 0) {
569  side_data = av_packet_new_side_data(pkt,
571  cx_frame->sz_alpha + 8);
572  if(!side_data) {
573  av_free_packet(pkt);
574  av_free(pkt);
575  return AVERROR(ENOMEM);
576  }
577  AV_WB64(side_data, 1);
578  memcpy(side_data + 8, cx_frame->buf_alpha, cx_frame->sz_alpha);
579  }
580  } else {
581  return ret;
582  }
583  return pkt->size;
584 }
585 
586 /**
587  * Queue multiple output frames from the encoder, returning the front-most.
588  * In cases where vpx_codec_get_cx_data() returns more than 1 frame append
589  * the frame queue. Return the head frame if available.
590  * @return Stored frame size
591  * @return AVERROR(EINVAL) on output size error
592  * @return AVERROR(ENOMEM) on coded frame queue data allocation error
593  */
594 static int queue_frames(AVCodecContext *avctx, AVPacket *pkt_out,
595  AVFrame *coded_frame)
596 {
597  VP8Context *ctx = avctx->priv_data;
598  const struct vpx_codec_cx_pkt *pkt;
599  const struct vpx_codec_cx_pkt *pkt_alpha = NULL;
600  const void *iter = NULL;
601  const void *iter_alpha = NULL;
602  int size = 0;
603 
604  if (ctx->coded_frame_list) {
605  struct FrameListData *cx_frame = ctx->coded_frame_list;
606  /* return the leading frame if we've already begun queueing */
607  size = storeframe(avctx, cx_frame, pkt_out, coded_frame);
608  if (size < 0)
609  return size;
610  ctx->coded_frame_list = cx_frame->next;
611  free_coded_frame(cx_frame);
612  }
613 
614  /* consume all available output from the encoder before returning. buffers
615  are only good through the next vpx_codec call */
616  while ((pkt = vpx_codec_get_cx_data(&ctx->encoder, &iter)) &&
617  (!ctx->is_alpha ||
618  (ctx->is_alpha && (pkt_alpha = vpx_codec_get_cx_data(&ctx->encoder_alpha, &iter_alpha))))) {
619  switch (pkt->kind) {
620  case VPX_CODEC_CX_FRAME_PKT:
621  if (!size) {
622  struct FrameListData cx_frame;
623 
624  /* avoid storing the frame when the list is empty and we haven't yet
625  provided a frame for output */
627  cx_pktcpy(&cx_frame, pkt, pkt_alpha, ctx);
628  size = storeframe(avctx, &cx_frame, pkt_out, coded_frame);
629  if (size < 0)
630  return size;
631  } else {
632  struct FrameListData *cx_frame =
633  av_malloc(sizeof(struct FrameListData));
634 
635  if (!cx_frame) {
636  av_log(avctx, AV_LOG_ERROR,
637  "Frame queue element alloc failed\n");
638  return AVERROR(ENOMEM);
639  }
640  cx_pktcpy(cx_frame, pkt, pkt_alpha, ctx);
641  cx_frame->buf = av_malloc(cx_frame->sz);
642 
643  if (!cx_frame->buf) {
644  av_log(avctx, AV_LOG_ERROR,
645  "Data buffer alloc (%"SIZE_SPECIFIER" bytes) failed\n",
646  cx_frame->sz);
647  av_freep(&cx_frame);
648  return AVERROR(ENOMEM);
649  }
650  memcpy(cx_frame->buf, pkt->data.frame.buf, pkt->data.frame.sz);
651  if (ctx->is_alpha) {
652  cx_frame->buf_alpha = av_malloc(cx_frame->sz_alpha);
653  if (!cx_frame->buf_alpha) {
654  av_log(avctx, AV_LOG_ERROR,
655  "Data buffer alloc (%"SIZE_SPECIFIER" bytes) failed\n",
656  cx_frame->sz_alpha);
657  av_free(cx_frame);
658  return AVERROR(ENOMEM);
659  }
660  memcpy(cx_frame->buf_alpha, pkt_alpha->data.frame.buf, pkt_alpha->data.frame.sz);
661  }
662  coded_frame_add(&ctx->coded_frame_list, cx_frame);
663  }
664  break;
665  case VPX_CODEC_STATS_PKT: {
666  struct vpx_fixed_buf *stats = &ctx->twopass_stats;
667  int err;
668  if ((err = av_reallocp(&stats->buf,
669  stats->sz +
670  pkt->data.twopass_stats.sz)) < 0) {
671  stats->sz = 0;
672  av_log(avctx, AV_LOG_ERROR, "Stat buffer realloc failed\n");
673  return err;
674  }
675  memcpy((uint8_t*)stats->buf + stats->sz,
676  pkt->data.twopass_stats.buf, pkt->data.twopass_stats.sz);
677  stats->sz += pkt->data.twopass_stats.sz;
678  break;
679  }
680  case VPX_CODEC_PSNR_PKT:
681  av_assert0(!ctx->have_sse);
682  ctx->sse[0] = pkt->data.psnr.sse[0];
683  ctx->sse[1] = pkt->data.psnr.sse[1];
684  ctx->sse[2] = pkt->data.psnr.sse[2];
685  ctx->sse[3] = pkt->data.psnr.sse[3];
686  ctx->have_sse = 1;
687  break;
688  case VPX_CODEC_CUSTOM_PKT:
689  //ignore unsupported/unrecognized packet types
690  break;
691  }
692  }
693 
694  return size;
695 }
696 
697 static int vp8_encode(AVCodecContext *avctx, AVPacket *pkt,
698  const AVFrame *frame, int *got_packet)
699 {
700  VP8Context *ctx = avctx->priv_data;
701  struct vpx_image *rawimg = NULL;
702  struct vpx_image *rawimg_alpha = NULL;
703  int64_t timestamp = 0;
704  int res, coded_size;
705  vpx_enc_frame_flags_t flags = 0;
706 
707  if (frame) {
708  rawimg = &ctx->rawimg;
709  rawimg->planes[VPX_PLANE_Y] = frame->data[0];
710  rawimg->planes[VPX_PLANE_U] = frame->data[1];
711  rawimg->planes[VPX_PLANE_V] = frame->data[2];
712  rawimg->stride[VPX_PLANE_Y] = frame->linesize[0];
713  rawimg->stride[VPX_PLANE_U] = frame->linesize[1];
714  rawimg->stride[VPX_PLANE_V] = frame->linesize[2];
715  if (ctx->is_alpha) {
716  uint8_t *u_plane, *v_plane;
717  rawimg_alpha = &ctx->rawimg_alpha;
718  rawimg_alpha->planes[VPX_PLANE_Y] = frame->data[3];
719  u_plane = av_malloc(frame->linesize[1] * frame->height);
720  v_plane = av_malloc(frame->linesize[2] * frame->height);
721  if (!u_plane || !v_plane) {
722  av_free(u_plane);
723  av_free(v_plane);
724  return AVERROR(ENOMEM);
725  }
726  memset(u_plane, 0x80, frame->linesize[1] * frame->height);
727  rawimg_alpha->planes[VPX_PLANE_U] = u_plane;
728  memset(v_plane, 0x80, frame->linesize[2] * frame->height);
729  rawimg_alpha->planes[VPX_PLANE_V] = v_plane;
730  rawimg_alpha->stride[VPX_PLANE_Y] = frame->linesize[0];
731  rawimg_alpha->stride[VPX_PLANE_U] = frame->linesize[1];
732  rawimg_alpha->stride[VPX_PLANE_V] = frame->linesize[2];
733  }
734  timestamp = frame->pts;
735  if (frame->pict_type == AV_PICTURE_TYPE_I)
736  flags |= VPX_EFLAG_FORCE_KF;
737  }
738 
739  res = vpx_codec_encode(&ctx->encoder, rawimg, timestamp,
740  avctx->ticks_per_frame, flags, ctx->deadline);
741  if (res != VPX_CODEC_OK) {
742  log_encoder_error(avctx, "Error encoding frame");
743  return AVERROR_INVALIDDATA;
744  }
745 
746  if (ctx->is_alpha) {
747  res = vpx_codec_encode(&ctx->encoder_alpha, rawimg_alpha, timestamp,
748  avctx->ticks_per_frame, flags, ctx->deadline);
749  if (res != VPX_CODEC_OK) {
750  log_encoder_error(avctx, "Error encoding alpha frame");
751  return AVERROR_INVALIDDATA;
752  }
753  }
754 
755  coded_size = queue_frames(avctx, pkt, avctx->coded_frame);
756 
757  if (!frame && avctx->flags & CODEC_FLAG_PASS1) {
758  unsigned int b64_size = AV_BASE64_SIZE(ctx->twopass_stats.sz);
759 
760  avctx->stats_out = av_malloc(b64_size);
761  if (!avctx->stats_out) {
762  av_log(avctx, AV_LOG_ERROR, "Stat buffer alloc (%d bytes) failed\n",
763  b64_size);
764  return AVERROR(ENOMEM);
765  }
766  av_base64_encode(avctx->stats_out, b64_size, ctx->twopass_stats.buf,
767  ctx->twopass_stats.sz);
768  }
769 
770  if (rawimg_alpha) {
771  av_freep(&rawimg_alpha->planes[VPX_PLANE_U]);
772  av_freep(&rawimg_alpha->planes[VPX_PLANE_V]);
773  }
774 
775  *got_packet = !!coded_size;
776  return 0;
777 }
778 
779 #define OFFSET(x) offsetof(VP8Context, x)
780 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
781 
782 #ifndef VPX_ERROR_RESILIENT_DEFAULT
783 #define VPX_ERROR_RESILIENT_DEFAULT 1
784 #define VPX_ERROR_RESILIENT_PARTITIONS 2
785 #endif
786 
787 #define COMMON_OPTIONS \
788  { "cpu-used", "Quality/Speed ratio modifier", OFFSET(cpu_used), AV_OPT_TYPE_INT, {.i64 = 1}, -16, 16, VE}, \
789  { "auto-alt-ref", "Enable use of alternate reference " \
790  "frames (2-pass only)", OFFSET(auto_alt_ref), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE}, \
791  { "lag-in-frames", "Number of frames to look ahead for " \
792  "alternate reference frame selection", OFFSET(lag_in_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
793  { "arnr-maxframes", "altref noise reduction max frame count", OFFSET(arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
794  { "arnr-strength", "altref noise reduction filter strength", OFFSET(arnr_strength), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
795  { "arnr-type", "altref noise reduction filter type", OFFSET(arnr_type), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE, "arnr_type"}, \
796  { "backward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0, VE, "arnr_type" }, \
797  { "forward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0, VE, "arnr_type" }, \
798  { "centered", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 3}, 0, 0, VE, "arnr_type" }, \
799  { "deadline", "Time to spend encoding, in microseconds.", OFFSET(deadline), AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, "quality"}, \
800  { "best", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_BEST_QUALITY}, 0, 0, VE, "quality"}, \
801  { "good", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_GOOD_QUALITY}, 0, 0, VE, "quality"}, \
802  { "realtime", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_REALTIME}, 0, 0, VE, "quality"}, \
803  { "error-resilient", "Error resilience configuration", OFFSET(error_resilient), AV_OPT_TYPE_FLAGS, {.i64 = 0}, INT_MIN, INT_MAX, VE, "er"}, \
804  { "max-intra-rate", "Maximum I-frame bitrate (pct) 0=unlimited", OFFSET(max_intra_rate), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
805  { "default", "Improve resiliency against losses of whole frames", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_DEFAULT}, 0, 0, VE, "er"}, \
806  { "partitions", "The frame partitions are independently decodable " \
807  "by the bool decoder, meaning that partitions can be decoded even " \
808  "though earlier partitions have been lost. Note that intra predicition" \
809  " is still done over the partition boundary.", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_PARTITIONS}, 0, 0, VE, "er"}, \
810  { "crf", "Select the quality for constant quality mode", offsetof(VP8Context, crf), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 63, VE }, \
811  { "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 }, \
812 
813 #define LEGACY_OPTIONS \
814  {"speed", "", offsetof(VP8Context, cpu_used), AV_OPT_TYPE_INT, {.i64 = 1}, -16, 16, VE}, \
815  {"quality", "", offsetof(VP8Context, deadline), AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, "quality"}, \
816  {"vp8flags", "", offsetof(VP8Context, flags), FF_OPT_TYPE_FLAGS, {.i64 = 0}, 0, UINT_MAX, VE, "flags"}, \
817  {"error_resilient", "enable error resilience", 0, FF_OPT_TYPE_CONST, {.dbl = VP8F_ERROR_RESILIENT}, INT_MIN, INT_MAX, VE, "flags"}, \
818  {"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"}, \
819  {"arnr_max_frames", "altref noise reduction max frame count", offsetof(VP8Context, arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 15, VE}, \
820  {"arnr_strength", "altref noise reduction filter strength", offsetof(VP8Context, arnr_strength), AV_OPT_TYPE_INT, {.i64 = 3}, 0, 6, VE}, \
821  {"arnr_type", "altref noise reduction filter type", offsetof(VP8Context, arnr_type), AV_OPT_TYPE_INT, {.i64 = 3}, 1, 3, VE}, \
822  {"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}, \
823 
824 #if CONFIG_LIBVPX_VP8_ENCODER
825 static const AVOption vp8_options[] = {
828  { NULL }
829 };
830 #endif
831 
832 #if CONFIG_LIBVPX_VP9_ENCODER
833 static const AVOption vp9_options[] = {
835  { "lossless", "Lossless mode", OFFSET(lossless), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE},
836  { "tile-columns", "Number of tile columns to use, log2", OFFSET(tile_columns), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 6, VE},
837  { "tile-rows", "Number of tile rows to use, log2", OFFSET(tile_rows), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 2, VE},
838  { "frame-parallel", "Enable frame parallel decodability features", OFFSET(frame_parallel), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE},
839  { "aq-mode", "adaptive quantization mode", OFFSET(aq_mode), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 3, VE, "aq_mode"},
840  { "none", "Aq not used", 0, AV_OPT_TYPE_CONST, {.i64 = 0}, 0, 0, VE, "aq_mode" }, \
841  { "variance", "Variance based Aq", 0, AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0, VE, "aq_mode" }, \
842  { "complexity", "Complexity based Aq", 0, AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0, VE, "aq_mode" }, \
843  { "cyclic", "Cyclic Refresh Aq", 0, AV_OPT_TYPE_CONST, {.i64 = 3}, 0, 0, VE, "aq_mode" }, \
844  LEGACY_OPTIONS
845  { NULL }
846 };
847 #endif
848 
849 #undef COMMON_OPTIONS
850 #undef LEGACY_OPTIONS
851 
852 static const AVCodecDefault defaults[] = {
853  { "qmin", "-1" },
854  { "qmax", "-1" },
855  { "g", "-1" },
856  { "keyint_min", "-1" },
857  { NULL },
858 };
859 
860 #if CONFIG_LIBVPX_VP8_ENCODER
861 static av_cold int vp8_init(AVCodecContext *avctx)
862 {
863  return vpx_init(avctx, vpx_codec_vp8_cx());
864 }
865 
866 static const AVClass class_vp8 = {
867  .class_name = "libvpx-vp8 encoder",
868  .item_name = av_default_item_name,
869  .option = vp8_options,
870  .version = LIBAVUTIL_VERSION_INT,
871 };
872 
873 AVCodec ff_libvpx_vp8_encoder = {
874  .name = "libvpx",
875  .long_name = NULL_IF_CONFIG_SMALL("libvpx VP8"),
876  .type = AVMEDIA_TYPE_VIDEO,
877  .id = AV_CODEC_ID_VP8,
878  .priv_data_size = sizeof(VP8Context),
879  .init = vp8_init,
880  .encode2 = vp8_encode,
881  .close = vp8_free,
882  .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
884  .priv_class = &class_vp8,
885  .defaults = defaults,
886 };
887 #endif /* CONFIG_LIBVPX_VP8_ENCODER */
888 
889 #if CONFIG_LIBVPX_VP9_ENCODER
890 static av_cold int vp9_init(AVCodecContext *avctx)
891 {
892  return vpx_init(avctx, vpx_codec_vp9_cx());
893 }
894 
895 static const AVClass class_vp9 = {
896  .class_name = "libvpx-vp9 encoder",
897  .item_name = av_default_item_name,
898  .option = vp9_options,
899  .version = LIBAVUTIL_VERSION_INT,
900 };
901 
902 AVCodec ff_libvpx_vp9_encoder = {
903  .name = "libvpx-vp9",
904  .long_name = NULL_IF_CONFIG_SMALL("libvpx VP9"),
905  .type = AVMEDIA_TYPE_VIDEO,
906  .id = AV_CODEC_ID_VP9,
907  .priv_data_size = sizeof(VP8Context),
908  .init = vp9_init,
909  .encode2 = vp8_encode,
910  .close = vp8_free,
911  .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
912  .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE },
913  .priv_class = &class_vp9,
914  .defaults = defaults,
915  .init_static_data = ff_vp9_init_static,
916 };
917 #endif /* CONFIG_LIBVPX_VP9_ENCODER */