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;
58 struct FrameListData *next;
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;
67 uint8_t is_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;
73 struct FrameListData *coded_frame_list;
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
83 int auto_alt_ref;
84
85 int arnr_max_frames;
86 int arnr_strength;
87 int arnr_type;
88
89 int lag_in_frames;
90 int error_resilient;
91 int crf;
92 int max_intra_rate;
93
94 // VP9-only
95 int lossless;
96 int tile_columns;
97 int tile_rows;
98 int frame_parallel;
99 int aq_mode;
100 } VP8Context;
101
102 /** String mappings for enum vp8e_enc_control_id */
103 static const char *const ctlidstr[] = {
104 [VP8E_UPD_ENTROPY] = "VP8E_UPD_ENTROPY",
105 [VP8E_UPD_REFERENCE] = "VP8E_UPD_REFERENCE",
106 [VP8E_USE_REFERENCE] = "VP8E_USE_REFERENCE",
107 [VP8E_SET_ROI_MAP] = "VP8E_SET_ROI_MAP",
108 [VP8E_SET_ACTIVEMAP] = "VP8E_SET_ACTIVEMAP",
109 [VP8E_SET_SCALEMODE] = "VP8E_SET_SCALEMODE",
110 [VP8E_SET_CPUUSED] = "VP8E_SET_CPUUSED",
111 [VP8E_SET_ENABLEAUTOALTREF] = "VP8E_SET_ENABLEAUTOALTREF",
112 [VP8E_SET_NOISE_SENSITIVITY] = "VP8E_SET_NOISE_SENSITIVITY",
113 [VP8E_SET_SHARPNESS] = "VP8E_SET_SHARPNESS",
114 [VP8E_SET_STATIC_THRESHOLD] = "VP8E_SET_STATIC_THRESHOLD",
115 [VP8E_SET_TOKEN_PARTITIONS] = "VP8E_SET_TOKEN_PARTITIONS",
116 [VP8E_GET_LAST_QUANTIZER] = "VP8E_GET_LAST_QUANTIZER",
117 [VP8E_SET_ARNR_MAXFRAMES] = "VP8E_SET_ARNR_MAXFRAMES",
118 [VP8E_SET_ARNR_STRENGTH] = "VP8E_SET_ARNR_STRENGTH",
119 [VP8E_SET_ARNR_TYPE] = "VP8E_SET_ARNR_TYPE",
120 [VP8E_SET_CQ_LEVEL] = "VP8E_SET_CQ_LEVEL",
121 [VP8E_SET_MAX_INTRA_BITRATE_PCT] = "VP8E_SET_MAX_INTRA_BITRATE_PCT",
122 #if CONFIG_LIBVPX_VP9_ENCODER
123 [VP9E_SET_LOSSLESS] = "VP9E_SET_LOSSLESS",
124 [VP9E_SET_TILE_COLUMNS] = "VP9E_SET_TILE_COLUMNS",
125 [VP9E_SET_TILE_ROWS] = "VP9E_SET_TILE_ROWS",
126 [VP9E_SET_FRAME_PARALLEL_DECODING] = "VP9E_SET_FRAME_PARALLEL_DECODING",
127 [VP9E_SET_AQ_MODE] = "VP9E_SET_AQ_MODE",
128 #endif
129 };
130
log_encoder_error(AVCodecContext * avctx,const char * desc)131 static av_cold void log_encoder_error(AVCodecContext *avctx, const char *desc)
132 {
133 VP8Context *ctx = avctx->priv_data;
134 const char *error = vpx_codec_error(&ctx->encoder);
135 const char *detail = vpx_codec_error_detail(&ctx->encoder);
136
137 av_log(avctx, AV_LOG_ERROR, "%s: %s\n", desc, error);
138 if (detail)
139 av_log(avctx, AV_LOG_ERROR, " Additional information: %s\n", detail);
140 }
141
dump_enc_cfg(AVCodecContext * avctx,const struct vpx_codec_enc_cfg * cfg)142 static av_cold void dump_enc_cfg(AVCodecContext *avctx,
143 const struct vpx_codec_enc_cfg *cfg)
144 {
145 int width = -30;
146 int level = AV_LOG_DEBUG;
147
148 av_log(avctx, level, "vpx_codec_enc_cfg\n");
149 av_log(avctx, level, "generic settings\n"
150 " %*s%u\n %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
151 " %*s{%u/%u}\n %*s%u\n %*s%d\n %*s%u\n",
152 width, "g_usage:", cfg->g_usage,
153 width, "g_threads:", cfg->g_threads,
154 width, "g_profile:", cfg->g_profile,
155 width, "g_w:", cfg->g_w,
156 width, "g_h:", cfg->g_h,
157 width, "g_timebase:", cfg->g_timebase.num, cfg->g_timebase.den,
158 width, "g_error_resilient:", cfg->g_error_resilient,
159 width, "g_pass:", cfg->g_pass,
160 width, "g_lag_in_frames:", cfg->g_lag_in_frames);
161 av_log(avctx, level, "rate control settings\n"
162 " %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
163 " %*s%d\n %*s%p(%"SIZE_SPECIFIER")\n %*s%u\n",
164 width, "rc_dropframe_thresh:", cfg->rc_dropframe_thresh,
165 width, "rc_resize_allowed:", cfg->rc_resize_allowed,
166 width, "rc_resize_up_thresh:", cfg->rc_resize_up_thresh,
167 width, "rc_resize_down_thresh:", cfg->rc_resize_down_thresh,
168 width, "rc_end_usage:", cfg->rc_end_usage,
169 width, "rc_twopass_stats_in:", cfg->rc_twopass_stats_in.buf, cfg->rc_twopass_stats_in.sz,
170 width, "rc_target_bitrate:", cfg->rc_target_bitrate);
171 av_log(avctx, level, "quantizer settings\n"
172 " %*s%u\n %*s%u\n",
173 width, "rc_min_quantizer:", cfg->rc_min_quantizer,
174 width, "rc_max_quantizer:", cfg->rc_max_quantizer);
175 av_log(avctx, level, "bitrate tolerance\n"
176 " %*s%u\n %*s%u\n",
177 width, "rc_undershoot_pct:", cfg->rc_undershoot_pct,
178 width, "rc_overshoot_pct:", cfg->rc_overshoot_pct);
179 av_log(avctx, level, "decoder buffer model\n"
180 " %*s%u\n %*s%u\n %*s%u\n",
181 width, "rc_buf_sz:", cfg->rc_buf_sz,
182 width, "rc_buf_initial_sz:", cfg->rc_buf_initial_sz,
183 width, "rc_buf_optimal_sz:", cfg->rc_buf_optimal_sz);
184 av_log(avctx, level, "2 pass rate control settings\n"
185 " %*s%u\n %*s%u\n %*s%u\n",
186 width, "rc_2pass_vbr_bias_pct:", cfg->rc_2pass_vbr_bias_pct,
187 width, "rc_2pass_vbr_minsection_pct:", cfg->rc_2pass_vbr_minsection_pct,
188 width, "rc_2pass_vbr_maxsection_pct:", cfg->rc_2pass_vbr_maxsection_pct);
189 av_log(avctx, level, "keyframing settings\n"
190 " %*s%d\n %*s%u\n %*s%u\n",
191 width, "kf_mode:", cfg->kf_mode,
192 width, "kf_min_dist:", cfg->kf_min_dist,
193 width, "kf_max_dist:", cfg->kf_max_dist);
194 av_log(avctx, level, "\n");
195 }
196
coded_frame_add(void * list,struct FrameListData * cx_frame)197 static void coded_frame_add(void *list, struct FrameListData *cx_frame)
198 {
199 struct FrameListData **p = list;
200
201 while (*p)
202 p = &(*p)->next;
203 *p = cx_frame;
204 cx_frame->next = NULL;
205 }
206
free_coded_frame(struct FrameListData * cx_frame)207 static av_cold void free_coded_frame(struct FrameListData *cx_frame)
208 {
209 av_freep(&cx_frame->buf);
210 if (cx_frame->buf_alpha)
211 av_freep(&cx_frame->buf_alpha);
212 av_freep(&cx_frame);
213 }
214
free_frame_list(struct FrameListData * list)215 static av_cold void free_frame_list(struct FrameListData *list)
216 {
217 struct FrameListData *p = list;
218
219 while (p) {
220 list = list->next;
221 free_coded_frame(p);
222 p = list;
223 }
224 }
225
codecctl_int(AVCodecContext * avctx,enum vp8e_enc_control_id id,int val)226 static av_cold int codecctl_int(AVCodecContext *avctx,
227 enum vp8e_enc_control_id id, int val)
228 {
229 VP8Context *ctx = avctx->priv_data;
230 char buf[80];
231 int width = -30;
232 int res;
233
234 snprintf(buf, sizeof(buf), "%s:", ctlidstr[id]);
235 av_log(avctx, AV_LOG_DEBUG, " %*s%d\n", width, buf, val);
236
237 res = vpx_codec_control(&ctx->encoder, id, val);
238 if (res != VPX_CODEC_OK) {
239 snprintf(buf, sizeof(buf), "Failed to set %s codec control",
240 ctlidstr[id]);
241 log_encoder_error(avctx, buf);
242 }
243
244 return res == VPX_CODEC_OK ? 0 : AVERROR(EINVAL);
245 }
246
vp8_free(AVCodecContext * avctx)247 static av_cold int vp8_free(AVCodecContext *avctx)
248 {
249 VP8Context *ctx = avctx->priv_data;
250
251 vpx_codec_destroy(&ctx->encoder);
252 if (ctx->is_alpha)
253 vpx_codec_destroy(&ctx->encoder_alpha);
254 av_freep(&ctx->twopass_stats.buf);
255 av_freep(&avctx->coded_frame);
256 av_freep(&avctx->stats_out);
257 free_frame_list(ctx->coded_frame_list);
258 return 0;
259 }
260
vpx_init(AVCodecContext * avctx,const struct vpx_codec_iface * iface)261 static av_cold int vpx_init(AVCodecContext *avctx,
262 const struct vpx_codec_iface *iface)
263 {
264 VP8Context *ctx = avctx->priv_data;
265 struct vpx_codec_enc_cfg enccfg;
266 struct vpx_codec_enc_cfg enccfg_alpha;
267 vpx_codec_flags_t flags = (avctx->flags & CODEC_FLAG_PSNR) ? VPX_CODEC_USE_PSNR : 0;
268 int res;
269
270 av_log(avctx, AV_LOG_INFO, "%s\n", vpx_codec_version_str());
271 av_log(avctx, AV_LOG_VERBOSE, "%s\n", vpx_codec_build_config());
272
273 if (avctx->pix_fmt == AV_PIX_FMT_YUVA420P)
274 ctx->is_alpha = 1;
275
276 if ((res = vpx_codec_enc_config_default(iface, &enccfg, 0)) != VPX_CODEC_OK) {
277 av_log(avctx, AV_LOG_ERROR, "Failed to get config: %s\n",
278 vpx_codec_err_to_string(res));
279 return AVERROR(EINVAL);
280 }
281
282 if(!avctx->bit_rate)
283 if(avctx->rc_max_rate || avctx->rc_buffer_size || avctx->rc_initial_buffer_occupancy) {
284 av_log( avctx, AV_LOG_ERROR, "Rate control parameters set without a bitrate\n");
285 return AVERROR(EINVAL);
286 }
287
288 dump_enc_cfg(avctx, &enccfg);
289
290 enccfg.g_w = avctx->width;
291 enccfg.g_h = avctx->height;
292 enccfg.g_timebase.num = avctx->time_base.num;
293 enccfg.g_timebase.den = avctx->time_base.den;
294 enccfg.g_threads = avctx->thread_count;
295 enccfg.g_lag_in_frames= ctx->lag_in_frames;
296
297 if (avctx->flags & CODEC_FLAG_PASS1)
298 enccfg.g_pass = VPX_RC_FIRST_PASS;
299 else if (avctx->flags & CODEC_FLAG_PASS2)
300 enccfg.g_pass = VPX_RC_LAST_PASS;
301 else
302 enccfg.g_pass = VPX_RC_ONE_PASS;
303
304 if (avctx->rc_min_rate == avctx->rc_max_rate &&
305 avctx->rc_min_rate == avctx->bit_rate && avctx->bit_rate) {
306 enccfg.rc_end_usage = VPX_CBR;
307 } else if (ctx->crf >= 0) {
308 enccfg.rc_end_usage = VPX_CQ;
309 #if CONFIG_LIBVPX_VP9_ENCODER
310 if (!avctx->bit_rate && avctx->codec_id == AV_CODEC_ID_VP9)
311 enccfg.rc_end_usage = VPX_Q;
312 #endif
313 }
314
315 if (avctx->bit_rate) {
316 enccfg.rc_target_bitrate = av_rescale_rnd(avctx->bit_rate, 1, 1000,
317 AV_ROUND_NEAR_INF);
318 #if CONFIG_LIBVPX_VP9_ENCODER
319 } else if (enccfg.rc_end_usage == VPX_Q) {
320 #endif
321 } else {
322 if (enccfg.rc_end_usage == VPX_CQ) {
323 enccfg.rc_target_bitrate = 1000000;
324 } else {
325 avctx->bit_rate = enccfg.rc_target_bitrate * 1000;
326 av_log(avctx, AV_LOG_WARNING,
327 "Neither bitrate nor constrained quality specified, using default bitrate of %dkbit/sec\n",
328 enccfg.rc_target_bitrate);
329 }
330 }
331
332 if (avctx->qmin >= 0)
333 enccfg.rc_min_quantizer = avctx->qmin;
334 if (avctx->qmax >= 0)
335 enccfg.rc_max_quantizer = avctx->qmax;
336
337 if (enccfg.rc_end_usage == VPX_CQ
338 #if CONFIG_LIBVPX_VP9_ENCODER
339 || enccfg.rc_end_usage == VPX_Q
340 #endif
341 ) {
342 if (ctx->crf < enccfg.rc_min_quantizer || ctx->crf > enccfg.rc_max_quantizer) {
343 av_log(avctx, AV_LOG_ERROR,
344 "CQ level %d must be between minimum and maximum quantizer value (%d-%d)\n",
345 ctx->crf, enccfg.rc_min_quantizer, enccfg.rc_max_quantizer);
346 return AVERROR(EINVAL);
347 }
348 }
349
350 enccfg.rc_dropframe_thresh = avctx->frame_skip_threshold;
351
352 //0-100 (0 => CBR, 100 => VBR)
353 enccfg.rc_2pass_vbr_bias_pct = round(avctx->qcompress * 100);
354 if (avctx->bit_rate)
355 enccfg.rc_2pass_vbr_minsection_pct =
356 avctx->rc_min_rate * 100LL / avctx->bit_rate;
357 if (avctx->rc_max_rate)
358 enccfg.rc_2pass_vbr_maxsection_pct =
359 avctx->rc_max_rate * 100LL / avctx->bit_rate;
360
361 if (avctx->rc_buffer_size)
362 enccfg.rc_buf_sz =
363 avctx->rc_buffer_size * 1000LL / avctx->bit_rate;
364 if (avctx->rc_initial_buffer_occupancy)
365 enccfg.rc_buf_initial_sz =
366 avctx->rc_initial_buffer_occupancy * 1000LL / avctx->bit_rate;
367 enccfg.rc_buf_optimal_sz = enccfg.rc_buf_sz * 5 / 6;
368 enccfg.rc_undershoot_pct = round(avctx->rc_buffer_aggressivity * 100);
369
370 //_enc_init() will balk if kf_min_dist differs from max w/VPX_KF_AUTO
371 if (avctx->keyint_min >= 0 && avctx->keyint_min == avctx->gop_size)
372 enccfg.kf_min_dist = avctx->keyint_min;
373 if (avctx->gop_size >= 0)
374 enccfg.kf_max_dist = avctx->gop_size;
375
376 if (enccfg.g_pass == VPX_RC_FIRST_PASS)
377 enccfg.g_lag_in_frames = 0;
378 else if (enccfg.g_pass == VPX_RC_LAST_PASS) {
379 int decode_size;
380
381 if (!avctx->stats_in) {
382 av_log(avctx, AV_LOG_ERROR, "No stats file for second pass\n");
383 return AVERROR_INVALIDDATA;
384 }
385
386 ctx->twopass_stats.sz = strlen(avctx->stats_in) * 3 / 4;
387 ctx->twopass_stats.buf = av_malloc(ctx->twopass_stats.sz);
388 if (!ctx->twopass_stats.buf) {
389 av_log(avctx, AV_LOG_ERROR,
390 "Stat buffer alloc (%"SIZE_SPECIFIER" bytes) failed\n",
391 ctx->twopass_stats.sz);
392 return AVERROR(ENOMEM);
393 }
394 decode_size = av_base64_decode(ctx->twopass_stats.buf, avctx->stats_in,
395 ctx->twopass_stats.sz);
396 if (decode_size < 0) {
397 av_log(avctx, AV_LOG_ERROR, "Stat buffer decode failed\n");
398 return AVERROR_INVALIDDATA;
399 }
400
401 ctx->twopass_stats.sz = decode_size;
402 enccfg.rc_twopass_stats_in = ctx->twopass_stats;
403 }
404
405 /* 0-3: For non-zero values the encoder increasingly optimizes for reduced
406 complexity playback on low powered devices at the expense of encode
407 quality. */
408 if (avctx->profile != FF_PROFILE_UNKNOWN)
409 enccfg.g_profile = avctx->profile;
410
411 enccfg.g_error_resilient = ctx->error_resilient || ctx->flags & VP8F_ERROR_RESILIENT;
412
413 dump_enc_cfg(avctx, &enccfg);
414 /* Construct Encoder Context */
415 res = vpx_codec_enc_init(&ctx->encoder, iface, &enccfg, flags);
416 if (res != VPX_CODEC_OK) {
417 log_encoder_error(avctx, "Failed to initialize encoder");
418 return AVERROR(EINVAL);
419 }
420
421 if (ctx->is_alpha) {
422 enccfg_alpha = enccfg;
423 res = vpx_codec_enc_init(&ctx->encoder_alpha, iface, &enccfg_alpha, flags);
424 if (res != VPX_CODEC_OK) {
425 log_encoder_error(avctx, "Failed to initialize alpha encoder");
426 return AVERROR(EINVAL);
427 }
428 }
429
430 //codec control failures are currently treated only as warnings
431 av_log(avctx, AV_LOG_DEBUG, "vpx_codec_control\n");
432 codecctl_int(avctx, VP8E_SET_CPUUSED, ctx->cpu_used);
433 if (ctx->flags & VP8F_AUTO_ALT_REF)
434 ctx->auto_alt_ref = 1;
435 if (ctx->auto_alt_ref >= 0)
436 codecctl_int(avctx, VP8E_SET_ENABLEAUTOALTREF, ctx->auto_alt_ref);
437 if (ctx->arnr_max_frames >= 0)
438 codecctl_int(avctx, VP8E_SET_ARNR_MAXFRAMES, ctx->arnr_max_frames);
439 if (ctx->arnr_strength >= 0)
440 codecctl_int(avctx, VP8E_SET_ARNR_STRENGTH, ctx->arnr_strength);
441 if (ctx->arnr_type >= 0)
442 codecctl_int(avctx, VP8E_SET_ARNR_TYPE, ctx->arnr_type);
443 codecctl_int(avctx, VP8E_SET_NOISE_SENSITIVITY, avctx->noise_reduction);
444 if (avctx->codec_id == AV_CODEC_ID_VP8)
445 codecctl_int(avctx, VP8E_SET_TOKEN_PARTITIONS, av_log2(avctx->slices));
446 codecctl_int(avctx, VP8E_SET_STATIC_THRESHOLD, avctx->mb_threshold);
447 if (ctx->crf >= 0)
448 codecctl_int(avctx, VP8E_SET_CQ_LEVEL, ctx->crf);
449 if (ctx->max_intra_rate >= 0)
450 codecctl_int(avctx, VP8E_SET_MAX_INTRA_BITRATE_PCT, ctx->max_intra_rate);
451
452 #if CONFIG_LIBVPX_VP9_ENCODER
453 if (avctx->codec_id == AV_CODEC_ID_VP9) {
454 if (ctx->lossless >= 0)
455 codecctl_int(avctx, VP9E_SET_LOSSLESS, ctx->lossless);
456 if (ctx->tile_columns >= 0)
457 codecctl_int(avctx, VP9E_SET_TILE_COLUMNS, ctx->tile_columns);
458 if (ctx->tile_rows >= 0)
459 codecctl_int(avctx, VP9E_SET_TILE_ROWS, ctx->tile_rows);
460 if (ctx->frame_parallel >= 0)
461 codecctl_int(avctx, VP9E_SET_FRAME_PARALLEL_DECODING, ctx->frame_parallel);
462 if (ctx->aq_mode >= 0)
463 codecctl_int(avctx, VP9E_SET_AQ_MODE, ctx->aq_mode);
464 }
465 #endif
466
467 av_log(avctx, AV_LOG_DEBUG, "Using deadline: %d\n", ctx->deadline);
468
469 //provide dummy value to initialize wrapper, values will be updated each _encode()
470 vpx_img_wrap(&ctx->rawimg, VPX_IMG_FMT_I420, avctx->width, avctx->height, 1,
471 (unsigned char*)1);
472
473 if (ctx->is_alpha)
474 vpx_img_wrap(&ctx->rawimg_alpha, VPX_IMG_FMT_I420, avctx->width, avctx->height, 1,
475 (unsigned char*)1);
476
477 avctx->coded_frame = av_frame_alloc();
478 if (!avctx->coded_frame) {
479 av_log(avctx, AV_LOG_ERROR, "Error allocating coded frame\n");
480 vp8_free(avctx);
481 return AVERROR(ENOMEM);
482 }
483 return 0;
484 }
485
cx_pktcpy(struct FrameListData * dst,const struct vpx_codec_cx_pkt * src,const struct vpx_codec_cx_pkt * src_alpha,VP8Context * ctx)486 static inline void cx_pktcpy(struct FrameListData *dst,
487 const struct vpx_codec_cx_pkt *src,
488 const struct vpx_codec_cx_pkt *src_alpha,
489 VP8Context *ctx)
490 {
491 dst->pts = src->data.frame.pts;
492 dst->duration = src->data.frame.duration;
493 dst->flags = src->data.frame.flags;
494 dst->sz = src->data.frame.sz;
495 dst->buf = src->data.frame.buf;
496 dst->have_sse = 0;
497 /* For alt-ref frame, don't store PSNR or increment frame_number */
498 if (!(dst->flags & VPX_FRAME_IS_INVISIBLE)) {
499 dst->frame_number = ++ctx->frame_number;
500 dst->have_sse = ctx->have_sse;
501 if (ctx->have_sse) {
502 /* associate last-seen SSE to the frame. */
503 /* Transfers ownership from ctx to dst. */
504 /* WARNING! This makes the assumption that PSNR_PKT comes
505 just before the frame it refers to! */
506 memcpy(dst->sse, ctx->sse, sizeof(dst->sse));
507 ctx->have_sse = 0;
508 }
509 } else {
510 dst->frame_number = -1; /* sanity marker */
511 }
512 if (src_alpha) {
513 dst->buf_alpha = src_alpha->data.frame.buf;
514 dst->sz_alpha = src_alpha->data.frame.sz;
515 }
516 else {
517 dst->buf_alpha = NULL;
518 dst->sz_alpha = 0;
519 }
520 }
521
522 /**
523 * Store coded frame information in format suitable for return from encode2().
524 *
525 * Write information from @a cx_frame to @a pkt
526 * @return packet data size on success
527 * @return a negative AVERROR on error
528 */
storeframe(AVCodecContext * avctx,struct FrameListData * cx_frame,AVPacket * pkt,AVFrame * coded_frame)529 static int storeframe(AVCodecContext *avctx, struct FrameListData *cx_frame,
530 AVPacket *pkt, AVFrame *coded_frame)
531 {
532 int ret = ff_alloc_packet2(avctx, pkt, cx_frame->sz);
533 uint8_t *side_data;
534 if (ret >= 0) {
535 memcpy(pkt->data, cx_frame->buf, pkt->size);
536 pkt->pts = pkt->dts = cx_frame->pts;
537 coded_frame->pts = cx_frame->pts;
538 coded_frame->key_frame = !!(cx_frame->flags & VPX_FRAME_IS_KEY);
539
540 if (coded_frame->key_frame) {
541 coded_frame->pict_type = AV_PICTURE_TYPE_I;
542 pkt->flags |= AV_PKT_FLAG_KEY;
543 } else
544 coded_frame->pict_type = AV_PICTURE_TYPE_P;
545
546 if (cx_frame->have_sse) {
547 int i;
548 /* Beware of the Y/U/V/all order! */
549 coded_frame->error[0] = cx_frame->sse[1];
550 coded_frame->error[1] = cx_frame->sse[2];
551 coded_frame->error[2] = cx_frame->sse[3];
552 coded_frame->error[3] = 0; // alpha
553 for (i = 0; i < 4; ++i) {
554 avctx->error[i] += coded_frame->error[i];
555 }
556 cx_frame->have_sse = 0;
557 }
558 if (cx_frame->sz_alpha > 0) {
559 side_data = av_packet_new_side_data(pkt,
560 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
561 cx_frame->sz_alpha + 8);
562 if(!side_data) {
563 av_free_packet(pkt);
564 av_free(pkt);
565 return AVERROR(ENOMEM);
566 }
567 AV_WB64(side_data, 1);
568 memcpy(side_data + 8, cx_frame->buf_alpha, cx_frame->sz_alpha);
569 }
570 } else {
571 return ret;
572 }
573 return pkt->size;
574 }
575
576 /**
577 * Queue multiple output frames from the encoder, returning the front-most.
578 * In cases where vpx_codec_get_cx_data() returns more than 1 frame append
579 * the frame queue. Return the head frame if available.
580 * @return Stored frame size
581 * @return AVERROR(EINVAL) on output size error
582 * @return AVERROR(ENOMEM) on coded frame queue data allocation error
583 */
queue_frames(AVCodecContext * avctx,AVPacket * pkt_out,AVFrame * coded_frame)584 static int queue_frames(AVCodecContext *avctx, AVPacket *pkt_out,
585 AVFrame *coded_frame)
586 {
587 VP8Context *ctx = avctx->priv_data;
588 const struct vpx_codec_cx_pkt *pkt;
589 const struct vpx_codec_cx_pkt *pkt_alpha = NULL;
590 const void *iter = NULL;
591 const void *iter_alpha = NULL;
592 int size = 0;
593
594 if (ctx->coded_frame_list) {
595 struct FrameListData *cx_frame = ctx->coded_frame_list;
596 /* return the leading frame if we've already begun queueing */
597 size = storeframe(avctx, cx_frame, pkt_out, coded_frame);
598 if (size < 0)
599 return size;
600 ctx->coded_frame_list = cx_frame->next;
601 free_coded_frame(cx_frame);
602 }
603
604 /* consume all available output from the encoder before returning. buffers
605 are only good through the next vpx_codec call */
606 while ((pkt = vpx_codec_get_cx_data(&ctx->encoder, &iter)) &&
607 (!ctx->is_alpha ||
608 (ctx->is_alpha && (pkt_alpha = vpx_codec_get_cx_data(&ctx->encoder_alpha, &iter_alpha))))) {
609 switch (pkt->kind) {
610 case VPX_CODEC_CX_FRAME_PKT:
611 if (!size) {
612 struct FrameListData cx_frame;
613
614 /* avoid storing the frame when the list is empty and we haven't yet
615 provided a frame for output */
616 av_assert0(!ctx->coded_frame_list);
617 cx_pktcpy(&cx_frame, pkt, pkt_alpha, ctx);
618 size = storeframe(avctx, &cx_frame, pkt_out, coded_frame);
619 if (size < 0)
620 return size;
621 } else {
622 struct FrameListData *cx_frame =
623 av_malloc(sizeof(struct FrameListData));
624
625 if (!cx_frame) {
626 av_log(avctx, AV_LOG_ERROR,
627 "Frame queue element alloc failed\n");
628 return AVERROR(ENOMEM);
629 }
630 cx_pktcpy(cx_frame, pkt, pkt_alpha, ctx);
631 cx_frame->buf = av_malloc(cx_frame->sz);
632
633 if (!cx_frame->buf) {
634 av_log(avctx, AV_LOG_ERROR,
635 "Data buffer alloc (%"SIZE_SPECIFIER" bytes) failed\n",
636 cx_frame->sz);
637 av_free(cx_frame);
638 return AVERROR(ENOMEM);
639 }
640 memcpy(cx_frame->buf, pkt->data.frame.buf, pkt->data.frame.sz);
641 if (ctx->is_alpha) {
642 cx_frame->buf_alpha = av_malloc(cx_frame->sz_alpha);
643 if (!cx_frame->buf_alpha) {
644 av_log(avctx, AV_LOG_ERROR,
645 "Data buffer alloc (%"SIZE_SPECIFIER" bytes) failed\n",
646 cx_frame->sz_alpha);
647 av_free(cx_frame);
648 return AVERROR(ENOMEM);
649 }
650 memcpy(cx_frame->buf_alpha, pkt_alpha->data.frame.buf, pkt_alpha->data.frame.sz);
651 }
652 coded_frame_add(&ctx->coded_frame_list, cx_frame);
653 }
654 break;
655 case VPX_CODEC_STATS_PKT: {
656 struct vpx_fixed_buf *stats = &ctx->twopass_stats;
657 int err;
658 if ((err = av_reallocp(&stats->buf,
659 stats->sz +
660 pkt->data.twopass_stats.sz)) < 0) {
661 stats->sz = 0;
662 av_log(avctx, AV_LOG_ERROR, "Stat buffer realloc failed\n");
663 return err;
664 }
665 memcpy((uint8_t*)stats->buf + stats->sz,
666 pkt->data.twopass_stats.buf, pkt->data.twopass_stats.sz);
667 stats->sz += pkt->data.twopass_stats.sz;
668 break;
669 }
670 case VPX_CODEC_PSNR_PKT:
671 av_assert0(!ctx->have_sse);
672 ctx->sse[0] = pkt->data.psnr.sse[0];
673 ctx->sse[1] = pkt->data.psnr.sse[1];
674 ctx->sse[2] = pkt->data.psnr.sse[2];
675 ctx->sse[3] = pkt->data.psnr.sse[3];
676 ctx->have_sse = 1;
677 break;
678 case VPX_CODEC_CUSTOM_PKT:
679 //ignore unsupported/unrecognized packet types
680 break;
681 }
682 }
683
684 return size;
685 }
686
vp8_encode(AVCodecContext * avctx,AVPacket * pkt,const AVFrame * frame,int * got_packet)687 static int vp8_encode(AVCodecContext *avctx, AVPacket *pkt,
688 const AVFrame *frame, int *got_packet)
689 {
690 VP8Context *ctx = avctx->priv_data;
691 struct vpx_image *rawimg = NULL;
692 struct vpx_image *rawimg_alpha = NULL;
693 int64_t timestamp = 0;
694 int res, coded_size;
695 vpx_enc_frame_flags_t flags = 0;
696
697 if (frame) {
698 rawimg = &ctx->rawimg;
699 rawimg->planes[VPX_PLANE_Y] = frame->data[0];
700 rawimg->planes[VPX_PLANE_U] = frame->data[1];
701 rawimg->planes[VPX_PLANE_V] = frame->data[2];
702 rawimg->stride[VPX_PLANE_Y] = frame->linesize[0];
703 rawimg->stride[VPX_PLANE_U] = frame->linesize[1];
704 rawimg->stride[VPX_PLANE_V] = frame->linesize[2];
705 if (ctx->is_alpha) {
706 uint8_t *u_plane, *v_plane;
707 rawimg_alpha = &ctx->rawimg_alpha;
708 rawimg_alpha->planes[VPX_PLANE_Y] = frame->data[3];
709 u_plane = av_malloc(frame->linesize[1] * frame->height);
710 memset(u_plane, 0x80, frame->linesize[1] * frame->height);
711 rawimg_alpha->planes[VPX_PLANE_U] = u_plane;
712 v_plane = av_malloc(frame->linesize[2] * frame->height);
713 memset(v_plane, 0x80, frame->linesize[2] * frame->height);
714 rawimg_alpha->planes[VPX_PLANE_V] = v_plane;
715 rawimg_alpha->stride[VPX_PLANE_Y] = frame->linesize[0];
716 rawimg_alpha->stride[VPX_PLANE_U] = frame->linesize[1];
717 rawimg_alpha->stride[VPX_PLANE_V] = frame->linesize[2];
718 }
719 timestamp = frame->pts;
720 if (frame->pict_type == AV_PICTURE_TYPE_I)
721 flags |= VPX_EFLAG_FORCE_KF;
722 }
723
724 res = vpx_codec_encode(&ctx->encoder, rawimg, timestamp,
725 avctx->ticks_per_frame, flags, ctx->deadline);
726 if (res != VPX_CODEC_OK) {
727 log_encoder_error(avctx, "Error encoding frame");
728 return AVERROR_INVALIDDATA;
729 }
730
731 if (ctx->is_alpha) {
732 res = vpx_codec_encode(&ctx->encoder_alpha, rawimg_alpha, timestamp,
733 avctx->ticks_per_frame, flags, ctx->deadline);
734 if (res != VPX_CODEC_OK) {
735 log_encoder_error(avctx, "Error encoding alpha frame");
736 return AVERROR_INVALIDDATA;
737 }
738 }
739
740 coded_size = queue_frames(avctx, pkt, avctx->coded_frame);
741
742 if (!frame && avctx->flags & CODEC_FLAG_PASS1) {
743 unsigned int b64_size = AV_BASE64_SIZE(ctx->twopass_stats.sz);
744
745 avctx->stats_out = av_malloc(b64_size);
746 if (!avctx->stats_out) {
747 av_log(avctx, AV_LOG_ERROR, "Stat buffer alloc (%d bytes) failed\n",
748 b64_size);
749 return AVERROR(ENOMEM);
750 }
751 av_base64_encode(avctx->stats_out, b64_size, ctx->twopass_stats.buf,
752 ctx->twopass_stats.sz);
753 }
754
755 if (rawimg_alpha) {
756 av_free(rawimg_alpha->planes[VPX_PLANE_U]);
757 av_free(rawimg_alpha->planes[VPX_PLANE_V]);
758 }
759
760 *got_packet = !!coded_size;
761 return 0;
762 }
763
764 #define OFFSET(x) offsetof(VP8Context, x)
765 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
766
767 #ifndef VPX_ERROR_RESILIENT_DEFAULT
768 #define VPX_ERROR_RESILIENT_DEFAULT 1
769 #define VPX_ERROR_RESILIENT_PARTITIONS 2
770 #endif
771
772 #define COMMON_OPTIONS \
773 { "cpu-used", "Quality/Speed ratio modifier", OFFSET(cpu_used), AV_OPT_TYPE_INT, {.i64 = 1}, -16, 16, VE}, \
774 { "auto-alt-ref", "Enable use of alternate reference " \
775 "frames (2-pass only)", OFFSET(auto_alt_ref), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE}, \
776 { "lag-in-frames", "Number of frames to look ahead for " \
777 "alternate reference frame selection", OFFSET(lag_in_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
778 { "arnr-maxframes", "altref noise reduction max frame count", OFFSET(arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
779 { "arnr-strength", "altref noise reduction filter strength", OFFSET(arnr_strength), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
780 { "arnr-type", "altref noise reduction filter type", OFFSET(arnr_type), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE, "arnr_type"}, \
781 { "backward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0, VE, "arnr_type" }, \
782 { "forward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0, VE, "arnr_type" }, \
783 { "centered", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 3}, 0, 0, VE, "arnr_type" }, \
784 { "deadline", "Time to spend encoding, in microseconds.", OFFSET(deadline), AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, "quality"}, \
785 { "best", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_BEST_QUALITY}, 0, 0, VE, "quality"}, \
786 { "good", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_GOOD_QUALITY}, 0, 0, VE, "quality"}, \
787 { "realtime", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_REALTIME}, 0, 0, VE, "quality"}, \
788 { "error-resilient", "Error resilience configuration", OFFSET(error_resilient), AV_OPT_TYPE_FLAGS, {.i64 = 0}, INT_MIN, INT_MAX, VE, "er"}, \
789 { "max-intra-rate", "Maximum I-frame bitrate (pct) 0=unlimited", OFFSET(max_intra_rate), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
790 { "default", "Improve resiliency against losses of whole frames", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_DEFAULT}, 0, 0, VE, "er"}, \
791 { "partitions", "The frame partitions are independently decodable " \
792 "by the bool decoder, meaning that partitions can be decoded even " \
793 "though earlier partitions have been lost. Note that intra predicition" \
794 " is still done over the partition boundary.", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_PARTITIONS}, 0, 0, VE, "er"}, \
795 { "crf", "Select the quality for constant quality mode", offsetof(VP8Context, crf), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 63, VE }, \
796
797 #define LEGACY_OPTIONS \
798 {"speed", "", offsetof(VP8Context, cpu_used), AV_OPT_TYPE_INT, {.i64 = 1}, -16, 16, VE}, \
799 {"quality", "", offsetof(VP8Context, deadline), AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, "quality"}, \
800 {"vp8flags", "", offsetof(VP8Context, flags), FF_OPT_TYPE_FLAGS, {.i64 = 0}, 0, UINT_MAX, VE, "flags"}, \
801 {"error_resilient", "enable error resilience", 0, FF_OPT_TYPE_CONST, {.dbl = VP8F_ERROR_RESILIENT}, INT_MIN, INT_MAX, VE, "flags"}, \
802 {"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"}, \
803 {"arnr_max_frames", "altref noise reduction max frame count", offsetof(VP8Context, arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 15, VE}, \
804 {"arnr_strength", "altref noise reduction filter strength", offsetof(VP8Context, arnr_strength), AV_OPT_TYPE_INT, {.i64 = 3}, 0, 6, VE}, \
805 {"arnr_type", "altref noise reduction filter type", offsetof(VP8Context, arnr_type), AV_OPT_TYPE_INT, {.i64 = 3}, 1, 3, VE}, \
806 {"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}, \
807
808 #if CONFIG_LIBVPX_VP8_ENCODER
809 static const AVOption vp8_options[] = {
810 COMMON_OPTIONS
811 LEGACY_OPTIONS
812 { NULL }
813 };
814 #endif
815
816 #if CONFIG_LIBVPX_VP9_ENCODER
817 static const AVOption vp9_options[] = {
818 COMMON_OPTIONS
819 { "lossless", "Lossless mode", OFFSET(lossless), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE},
820 { "tile-columns", "Number of tile columns to use, log2", OFFSET(tile_columns), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 6, VE},
821 { "tile-rows", "Number of tile rows to use, log2", OFFSET(tile_rows), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 2, VE},
822 { "frame-parallel", "Enable frame parallel decodability features", OFFSET(frame_parallel), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, VE},
823 { "aq-mode", "adaptive quantization mode", OFFSET(aq_mode), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 3, VE, "aq_mode"},
824 { "none", "Aq not used", 0, AV_OPT_TYPE_CONST, {.i64 = 0}, 0, 0, VE, "aq_mode" }, \
825 { "variance", "Variance based Aq", 0, AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0, VE, "aq_mode" }, \
826 { "complexity", "Complexity based Aq", 0, AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0, VE, "aq_mode" }, \
827 { "cyclic", "Cyclic Refresh Aq", 0, AV_OPT_TYPE_CONST, {.i64 = 3}, 0, 0, VE, "aq_mode" }, \
828 LEGACY_OPTIONS
829 { NULL }
830 };
831 #endif
832
833 #undef COMMON_OPTIONS
834 #undef LEGACY_OPTIONS
835
836 static const AVCodecDefault defaults[] = {
837 { "qmin", "-1" },
838 { "qmax", "-1" },
839 { "g", "-1" },
840 { "keyint_min", "-1" },
841 { NULL },
842 };
843
844 #if CONFIG_LIBVPX_VP8_ENCODER
vp8_init(AVCodecContext * avctx)845 static av_cold int vp8_init(AVCodecContext *avctx)
846 {
847 return vpx_init(avctx, &vpx_codec_vp8_cx_algo);
848 }
849
850 static const AVClass class_vp8 = {
851 .class_name = "libvpx-vp8 encoder",
852 .item_name = av_default_item_name,
853 .option = vp8_options,
854 .version = LIBAVUTIL_VERSION_INT,
855 };
856
857 AVCodec ff_libvpx_vp8_encoder = {
858 .name = "libvpx",
859 .long_name = NULL_IF_CONFIG_SMALL("libvpx VP8"),
860 .type = AVMEDIA_TYPE_VIDEO,
861 .id = AV_CODEC_ID_VP8,
862 .priv_data_size = sizeof(VP8Context),
863 .init = vp8_init,
864 .encode2 = vp8_encode,
865 .close = vp8_free,
866 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
867 .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUVA420P, AV_PIX_FMT_NONE },
868 .priv_class = &class_vp8,
869 .defaults = defaults,
870 };
871 #endif /* CONFIG_LIBVPX_VP8_ENCODER */
872
873 #if CONFIG_LIBVPX_VP9_ENCODER
vp9_init(AVCodecContext * avctx)874 static av_cold int vp9_init(AVCodecContext *avctx)
875 {
876 return vpx_init(avctx, &vpx_codec_vp9_cx_algo);
877 }
878
879 static const AVClass class_vp9 = {
880 .class_name = "libvpx-vp9 encoder",
881 .item_name = av_default_item_name,
882 .option = vp9_options,
883 .version = LIBAVUTIL_VERSION_INT,
884 };
885
886 AVCodec ff_libvpx_vp9_encoder = {
887 .name = "libvpx-vp9",
888 .long_name = NULL_IF_CONFIG_SMALL("libvpx VP9"),
889 .type = AVMEDIA_TYPE_VIDEO,
890 .id = AV_CODEC_ID_VP9,
891 .priv_data_size = sizeof(VP8Context),
892 .init = vp9_init,
893 .encode2 = vp8_encode,
894 .close = vp8_free,
895 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
896 .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE },
897 .priv_class = &class_vp9,
898 .defaults = defaults,
899 .init_static_data = ff_vp9_init_static,
900 };
901 #endif /* CONFIG_LIBVPX_VP9_ENCODER */
902