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 * VP8/9 encoder support via libvpx
24 */
25
26 #include "config_components.h"
27
28 #define VPX_DISABLE_CTRL_TYPECHECKS 1
29 #define VPX_CODEC_DISABLE_COMPAT 1
30 #include <vpx/vpx_encoder.h>
31 #include <vpx/vp8cx.h>
32
51
52 #define IS_VP9(avctx) (CONFIG_LIBVPX_VP9_ENCODER && avctx->codec_id == AV_CODEC_ID_VP9)
53 #define IS_VP8(avctx) (CONFIG_LIBVPX_VP8_ENCODER && avctx->codec_id == AV_CODEC_ID_VP8)
54
55 /**
56 * Portion of struct vpx_codec_cx_pkt from vpx_encoder.h.
57 * One encoded frame returned from the library.
58 */
60 void *
buf;
/**< compressed data buffer */
61 size_t sz;
/**< length of compressed data */
63 (in timebase units) */
64 uint32_t
flags;
/**< flags for this frame */
66 int have_sse;
/**< true if we have pending sse[] */
68 };
69
73
76
79
91 int have_sse;
/**< true if we have pending sse[] */
94
97 /**
98 * VP8 specific flags, see VP8F_* below.
99 */
101 #define VP8F_ERROR_RESILIENT 0x00000001 ///< Enable measures appropriate for streaming over lossy links
102 #define VP8F_AUTO_ALT_REF 0x00000002
///< Enable automatic alternate reference frame generation
103
105
109
111
119
123
124 // VP8-only
126
127 // VP9-only
142
143 // This FIFO is used to propagate various properties from frames to packets.
145 /**
146 * If the driver does not support ROI then warn the first time we
147 * encounter a frame with ROI side data.
148 */
150 #if CONFIG_LIBVPX_VP9_ENCODER && defined(VPX_CTRL_VP9E_SET_MAX_INTER_BITRATE_PCT)
151 vpx_svc_ref_frame_config_t ref_frame_config;
152 #endif
153 } VPxContext;
154
155 /** String mappings for enum vp8e_enc_control_id */
157 [VP8E_SET_CPUUSED] = "VP8E_SET_CPUUSED",
158 [VP8E_SET_ENABLEAUTOALTREF] = "VP8E_SET_ENABLEAUTOALTREF",
159 [VP8E_SET_NOISE_SENSITIVITY] = "VP8E_SET_NOISE_SENSITIVITY",
160 [VP8E_SET_STATIC_THRESHOLD] = "VP8E_SET_STATIC_THRESHOLD",
161 [VP8E_SET_TOKEN_PARTITIONS] = "VP8E_SET_TOKEN_PARTITIONS",
162 [VP8E_SET_ARNR_MAXFRAMES] = "VP8E_SET_ARNR_MAXFRAMES",
163 [VP8E_SET_ARNR_STRENGTH] = "VP8E_SET_ARNR_STRENGTH",
164 [VP8E_SET_ARNR_TYPE] = "VP8E_SET_ARNR_TYPE",
165 [VP8E_SET_TUNING] = "VP8E_SET_TUNING",
166 [VP8E_SET_CQ_LEVEL] = "VP8E_SET_CQ_LEVEL",
167 [VP8E_SET_MAX_INTRA_BITRATE_PCT] = "VP8E_SET_MAX_INTRA_BITRATE_PCT",
168 [VP8E_SET_SHARPNESS] = "VP8E_SET_SHARPNESS",
169 [VP8E_SET_TEMPORAL_LAYER_ID] = "VP8E_SET_TEMPORAL_LAYER_ID",
170 [VP8E_SET_SCREEN_CONTENT_MODE] = "VP8E_SET_SCREEN_CONTENT_MODE",
171 #if CONFIG_LIBVPX_VP9_ENCODER
172 [VP9E_SET_LOSSLESS] = "VP9E_SET_LOSSLESS",
173 [VP9E_SET_TILE_COLUMNS] = "VP9E_SET_TILE_COLUMNS",
174 [VP9E_SET_TILE_ROWS] = "VP9E_SET_TILE_ROWS",
175 [VP9E_SET_FRAME_PARALLEL_DECODING] = "VP9E_SET_FRAME_PARALLEL_DECODING",
176 [VP9E_SET_AQ_MODE] = "VP9E_SET_AQ_MODE",
177 [VP9E_SET_COLOR_SPACE] = "VP9E_SET_COLOR_SPACE",
178 [VP9E_SET_SVC_LAYER_ID] = "VP9E_SET_SVC_LAYER_ID",
179 #if VPX_ENCODER_ABI_VERSION >= 12
180 [VP9E_SET_SVC_PARAMETERS] = "VP9E_SET_SVC_PARAMETERS",
181 [VP9E_SET_SVC_REF_FRAME_CONFIG] = "VP9E_SET_SVC_REF_FRAME_CONFIG",
182 #endif
183 [VP9E_SET_SVC] = "VP9E_SET_SVC",
184 #if VPX_ENCODER_ABI_VERSION >= 11
185 [VP9E_SET_COLOR_RANGE] = "VP9E_SET_COLOR_RANGE",
186 #endif
187 #if VPX_ENCODER_ABI_VERSION >= 12
188 [VP9E_SET_TARGET_LEVEL] = "VP9E_SET_TARGET_LEVEL",
189 [VP9E_GET_LEVEL] = "VP9E_GET_LEVEL",
190 #endif
191 #ifdef VPX_CTRL_VP9E_SET_ROW_MT
192 [VP9E_SET_ROW_MT] = "VP9E_SET_ROW_MT",
193 #endif
194 #ifdef VPX_CTRL_VP9E_SET_TUNE_CONTENT
195 [VP9E_SET_TUNE_CONTENT] = "VP9E_SET_TUNE_CONTENT",
196 #endif
197 #ifdef VPX_CTRL_VP9E_SET_TPL
198 [VP9E_SET_TPL] = "VP9E_SET_TPL",
199 #endif
200 #ifdef VPX_CTRL_VP9E_SET_MIN_GF_INTERVAL
201 [VP9E_SET_MIN_GF_INTERVAL] = "VP9E_SET_MIN_GF_INTERVAL",
202 #endif
203 #endif
204 };
205
207 {
209 const char *
error = vpx_codec_error(&
ctx->encoder);
210 const char *detail = vpx_codec_error_detail(&
ctx->encoder);
211
213 if (detail)
215 }
216
218 const struct vpx_codec_enc_cfg *cfg,
220 {
223
226 " %*s%u\n %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
227 #if CONFIG_LIBVPX_VP9_ENCODER
228 " %*s%u\n %*s%u\n"
229 #endif
230 " %*s{%u/%u}\n %*s%u\n %*s%d\n %*s%u\n",
231 width,
"g_usage:", cfg->g_usage,
232 width,
"g_threads:", cfg->g_threads,
233 width,
"g_profile:", cfg->g_profile,
234 width,
"g_w:", cfg->g_w,
235 width,
"g_h:", cfg->g_h,
236 #if CONFIG_LIBVPX_VP9_ENCODER
237 width,
"g_bit_depth:", cfg->g_bit_depth,
238 width,
"g_input_bit_depth:", cfg->g_input_bit_depth,
239 #endif
240 width,
"g_timebase:", cfg->g_timebase.num, cfg->g_timebase.den,
241 width,
"g_error_resilient:", cfg->g_error_resilient,
242 width,
"g_pass:", cfg->g_pass,
243 width,
"g_lag_in_frames:", cfg->g_lag_in_frames);
245 " %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
247 width,
"rc_dropframe_thresh:", cfg->rc_dropframe_thresh,
248 width,
"rc_resize_allowed:", cfg->rc_resize_allowed,
249 width,
"rc_resize_up_thresh:", cfg->rc_resize_up_thresh,
250 width,
"rc_resize_down_thresh:", cfg->rc_resize_down_thresh,
251 width,
"rc_end_usage:", cfg->rc_end_usage,
252 width,
"rc_twopass_stats_in:", cfg->rc_twopass_stats_in.buf, cfg->rc_twopass_stats_in.sz,
253 width,
"rc_target_bitrate:", cfg->rc_target_bitrate);
255 " %*s%u\n %*s%u\n",
256 width,
"rc_min_quantizer:", cfg->rc_min_quantizer,
257 width,
"rc_max_quantizer:", cfg->rc_max_quantizer);
259 " %*s%u\n %*s%u\n",
260 width,
"rc_undershoot_pct:", cfg->rc_undershoot_pct,
261 width,
"rc_overshoot_pct:", cfg->rc_overshoot_pct);
262 av_log(avctx,
level,
"temporal layering settings\n"
263 " %*s%u\n",
width,
"ts_number_layers:", cfg->ts_number_layers);
266 "\n %*s",
width,
"ts_target_bitrate:");
267 for (
i = 0;
i < VPX_TS_MAX_LAYERS;
i++)
269 "%u ", cfg->ts_target_bitrate[
i]);
270 }
271 #if (VPX_ENCODER_ABI_VERSION >= 12) && CONFIG_LIBVPX_VP9_ENCODER
274 "\n %*s",
width,
"layer_target_bitrate:");
275 for (
i = 0;
i < VPX_TS_MAX_LAYERS;
i++)
277 "%u ", cfg->layer_target_bitrate[
i]);
278 }
279 #endif
282 "\n %*s",
width,
"ts_rate_decimator:");
283 for (
i = 0;
i < VPX_TS_MAX_LAYERS;
i++)
287 "\n %*s%u\n",
width,
"ts_periodicity:", cfg->ts_periodicity);
289 "\n %*s",
width,
"ts_layer_id:");
290 for (
i = 0;
i < VPX_TS_MAX_PERIODICITY;
i++)
294 " %*s%u\n %*s%u\n %*s%u\n",
295 width,
"rc_buf_sz:", cfg->rc_buf_sz,
296 width,
"rc_buf_initial_sz:", cfg->rc_buf_initial_sz,
297 width,
"rc_buf_optimal_sz:", cfg->rc_buf_optimal_sz);
298 av_log(avctx,
level,
"2 pass rate control settings\n"
299 " %*s%u\n %*s%u\n %*s%u\n",
300 width,
"rc_2pass_vbr_bias_pct:", cfg->rc_2pass_vbr_bias_pct,
301 width,
"rc_2pass_vbr_minsection_pct:", cfg->rc_2pass_vbr_minsection_pct,
302 width,
"rc_2pass_vbr_maxsection_pct:", cfg->rc_2pass_vbr_maxsection_pct);
303 #if VPX_ENCODER_ABI_VERSION >= 14
305 width,
"rc_2pass_vbr_corpus_complexity:", cfg->rc_2pass_vbr_corpus_complexity);
306 #endif
308 " %*s%d\n %*s%u\n %*s%u\n",
309 width,
"kf_mode:", cfg->kf_mode,
310 width,
"kf_min_dist:", cfg->kf_min_dist,
311 width,
"kf_max_dist:", cfg->kf_max_dist);
313 }
314
316 {
318
323 }
324
326 {
329 }
330
332 {
334
339 }
340 }
341
343 {
346 }
347
349 {
354 }
355
358 {
360 const struct vpx_codec_enc_cfg *enccfg =
ctx->encoder.config.enc;
361
364
366 // Keep HDR10+ if it has bit depth higher than 8 and
367 // it has PQ trc (SMPTE2084).
370
371 if (sd) {
375 }
376 }
377
384 }
385
389
390 return 0;
394 }
395
397 {
401
403 return 0;
406 "Mismatching timestamps: libvpx %"PRId64" queued %"PRId64"; "
407 "this is a bug, please report it\n",
pkt->
pts, fd.
pts);
409 }
410
416 }
417
423 }
424
426 }
427
431
433 }
434
436 enum vp8e_enc_control_id
id,
int val)
437 {
439 char buf[80];
441 int res;
442
445
446 res = vpx_codec_control(&
ctx->encoder,
id,
val);
447 if (res != VPX_CODEC_OK) {
448 snprintf(buf,
sizeof(buf),
"Failed to set %s codec control",
452 }
453
455 int res_alpha = vpx_codec_control(&
ctx->encoder_alpha,
id,
val);
456 if (res_alpha != VPX_CODEC_OK) {
457 snprintf(buf,
sizeof(buf),
"Failed to set %s alpha codec control",
461 }
462 }
463
464 return 0;
465 }
466
467 #if VPX_ENCODER_ABI_VERSION >= 12
469 enum vp8e_enc_control_id
id,
int *
val)
470 {
472 char buf[80];
474 int res;
475
478
479 res = vpx_codec_control(&
ctx->encoder,
id,
val);
480 if (res != VPX_CODEC_OK) {
481 snprintf(buf,
sizeof(buf),
"Failed to set %s codec control",
485 }
486
488 int res_alpha = vpx_codec_control(&
ctx->encoder_alpha,
id,
val);
489 if (res_alpha != VPX_CODEC_OK) {
490 snprintf(buf,
sizeof(buf),
"Failed to set %s alpha codec control",
494 }
495 }
496
497 return 0;
498 }
499 #endif
500
502 {
504
505 #if VPX_ENCODER_ABI_VERSION >= 12
508 int level_out = 0;
509 if (!codecctl_intp(avctx, VP9E_GET_LEVEL, &level_out))
511 }
512 #endif
513
515
516 vpx_codec_destroy(&
ctx->encoder);
518 vpx_codec_destroy(&
ctx->encoder_alpha);
521 }
528 return 0;
529 }
530
532 {
533 int dest_idx = 0;
534 char *saveptr =
NULL;
536
537 while (token && dest_idx < max_entries) {
538 dest[dest_idx++] = strtoul(token,
NULL, 10);
540 }
541 }
542
543 #if CONFIG_LIBVPX_VP9_ENCODER && defined(VPX_CTRL_VP9E_SET_MAX_INTER_BITRATE_PCT)
544 static void vp8_ts_parse_int64_array(
int64_t *dest,
char *
value,
size_t value_len,
int max_entries)
545 {
546 int dest_idx = 0;
547 char *saveptr =
NULL;
549
550 while (token && dest_idx < max_entries) {
551 dest[dest_idx++] = strtoull(token,
NULL, 10);
553 }
554 }
555 #endif
556
558 int *layer_flags, int *flag_periodicity)
559 {
560 switch (layering_mode) {
561 case 2: {
562 /**
563 * 2-layers, 2-frame period.
564 */
565 static const int ids[2] = { 0, 1 };
566 cfg->ts_periodicity = 2;
567 *flag_periodicity = 2;
568 cfg->ts_number_layers = 2;
569 cfg->ts_rate_decimator[0] = 2;
570 cfg->ts_rate_decimator[1] = 1;
571 memcpy(cfg->ts_layer_id, ids, sizeof(ids));
572
573 layer_flags[0] =
574 VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
575 VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
576 layer_flags[1] =
577 VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_GF |
578 VP8_EFLAG_NO_UPD_LAST |
579 VP8_EFLAG_NO_REF_ARF | VP8_EFLAG_NO_REF_GF;
580 break;
581 }
582 case 3: {
583 /**
584 * 3-layers structure with one reference frame.
585 * This works same as temporal_layering_mode 3.
586 *
587 * 3-layers, 4-frame period.
588 */
589 static const int ids[4] = { 0, 2, 1, 2 };
590 cfg->ts_periodicity = 4;
591 *flag_periodicity = 4;
592 cfg->ts_number_layers = 3;
593 cfg->ts_rate_decimator[0] = 4;
594 cfg->ts_rate_decimator[1] = 2;
595 cfg->ts_rate_decimator[2] = 1;
596 memcpy(cfg->ts_layer_id, ids, sizeof(ids));
597
598 /**
599 * 0=L, 1=GF, 2=ARF,
600 * Intra-layer prediction disabled.
601 */
602 layer_flags[0] =
603 VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
604 VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
605 layer_flags[1] =
606 VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
607 VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF |
608 VP8_EFLAG_NO_UPD_ARF;
609 layer_flags[2] =
610 VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
611 VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
612 layer_flags[3] =
613 VP8_EFLAG_NO_REF_LAST | VP8_EFLAG_NO_REF_ARF |
614 VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF |
615 VP8_EFLAG_NO_UPD_ARF;
616 break;
617 }
618 case 4: {
619 /**
620 * 3-layers structure.
621 * added dependency between the two TL2 frames (on top of case 3).
622 * 3-layers, 4-frame period.
623 */
624 static const int ids[4] = { 0, 2, 1, 2 };
625 cfg->ts_periodicity = 4;
626 *flag_periodicity = 4;
627 cfg->ts_number_layers = 3;
628 cfg->ts_rate_decimator[0] = 4;
629 cfg->ts_rate_decimator[1] = 2;
630 cfg->ts_rate_decimator[2] = 1;
631 memcpy(cfg->ts_layer_id, ids, sizeof(ids));
632
633 /**
634 * 0=L, 1=GF, 2=ARF, Intra-layer prediction disabled.
635 */
636 layer_flags[0] =
637 VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
638 VP8_EFLAG_NO_UPD_GF | VP8_EFLAG_NO_UPD_ARF;
639 layer_flags[1] =
640 VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
641 VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF;
642 layer_flags[2] =
643 VP8_EFLAG_NO_REF_GF | VP8_EFLAG_NO_REF_ARF |
644 VP8_EFLAG_NO_UPD_ARF | VP8_EFLAG_NO_UPD_LAST;
645 layer_flags[3] =
646 VP8_EFLAG_NO_REF_LAST |
647 VP8_EFLAG_NO_UPD_LAST | VP8_EFLAG_NO_UPD_GF |
648 VP8_EFLAG_NO_UPD_ARF;
649 break;
650 }
651 default:
652 /**
653 * do not change the layer_flags or the flag_periodicity in this case;
654 * it might be that the code is using external flags to be used.
655 */
656 break;
657
658 }
659 }
660
663 {
664 size_t value_len = strlen(
value);
665 int ts_layering_mode = 0;
666
667 if (!value_len)
668 return -1;
669
670 if (!strcmp(
key,
"ts_number_layers"))
671 enccfg->ts_number_layers = strtoul(
value, &
value, 10);
672 else if (!strcmp(
key,
"ts_target_bitrate")) {
675 #if (VPX_ENCODER_ABI_VERSION >= 12) && CONFIG_LIBVPX_VP9_ENCODER
678 #endif
679 }
else if (!strcmp(
key,
"ts_rate_decimator")) {
681 }
else if (!strcmp(
key,
"ts_periodicity")) {
682 enccfg->ts_periodicity = strtoul(
value, &
value, 10);
683 }
else if (!strcmp(
key,
"ts_layer_id")) {
685 }
else if (!strcmp(
key,
"ts_layering_mode")) {
686 /* option for pre-defined temporal structures in function set_temporal_layer_pattern. */
687 ts_layering_mode = strtoul(
value, &
value, 10);
688 }
689
690 #if (VPX_ENCODER_ABI_VERSION >= 12) && CONFIG_LIBVPX_VP9_ENCODER
691 enccfg->temporal_layering_mode = VP9E_TEMPORAL_LAYERING_MODE_BYPASS; // only bypass mode is supported for now.
692 enccfg->ss_number_layers = 1; // TODO: add spatial scalability support.
693 #endif
694 if (ts_layering_mode) {
695 // make sure the ts_layering_mode comes at the end of the ts_parameter string to ensure that
696 // correct configuration is done.
699 }
700
701 return 0;
702 }
703
704 #if CONFIG_LIBVPX_VP9_ENCODER && defined(VPX_CTRL_VP9E_SET_MAX_INTER_BITRATE_PCT)
705 static int vpx_ref_frame_config_set_value(vpx_svc_ref_frame_config_t *ref_frame_config,
706 int ss_number_layers,
char *
key,
char *
value)
707 {
708 size_t value_len = strlen(
value);
709
710 if (!value_len)
712
713 if (!strcmp(
key,
"rfc_update_buffer_slot")) {
715 }
else if (!strcmp(
key,
"rfc_update_last")) {
717 }
else if (!strcmp(
key,
"rfc_update_golden")) {
719 }
else if (!strcmp(
key,
"rfc_update_alt_ref")) {
721 }
else if (!strcmp(
key,
"rfc_lst_fb_idx")) {
723 }
else if (!strcmp(
key,
"rfc_gld_fb_idx")) {
725 }
else if (!strcmp(
key,
"rfc_alt_fb_idx")) {
727 }
else if (!strcmp(
key,
"rfc_reference_last")) {
729 }
else if (!strcmp(
key,
"rfc_reference_golden")) {
731 }
else if (!strcmp(
key,
"rfc_reference_alt_ref")) {
733 }
else if (!strcmp(
key,
"rfc_reference_duration")) {
734 vp8_ts_parse_int64_array(ref_frame_config->duration,
value, value_len, ss_number_layers);
735 }
736
737 return 0;
738 }
739
740 static int vpx_parse_ref_frame_config_element(vpx_svc_ref_frame_config_t *ref_frame_config,
741 int ss_number_layers, const char **buf)
742 {
743 const char key_val_sep[] = "=";
744 const char pairs_sep[] = ":";
748
749 if (
key && *
key && strspn(*buf, key_val_sep)) {
750 (*buf)++;
752 }
753
755 ret = vpx_ref_frame_config_set_value(ref_frame_config, ss_number_layers,
key,
val);
756 else
758
761
763 }
764
765 static int vpx_parse_ref_frame_config(vpx_svc_ref_frame_config_t *ref_frame_config,
766 int ss_number_layers, const char *str)
767 {
769
770 while (*str) {
772 vpx_parse_ref_frame_config_element(ref_frame_config, ss_number_layers, &str);
775
776 if (*str)
777 str++;
778 }
779
781 }
782 #endif
783
784 #if CONFIG_LIBVPX_VP9_ENCODER
786 struct vpx_codec_enc_cfg *enccfg, vpx_codec_flags_t *
flags,
787 vpx_img_fmt_t *img_fmt)
788 {
791 enccfg->g_bit_depth = enccfg->g_input_bit_depth =
desc->comp[0].depth;
795 enccfg->g_profile = 0;
796 *img_fmt = VPX_IMG_FMT_I420;
797 return 0;
799 enccfg->g_profile = 1;
800 *img_fmt = VPX_IMG_FMT_I422;
801 return 0;
803 enccfg->g_profile = 1;
804 *img_fmt = VPX_IMG_FMT_I440;
805 return 0;
807 ctx->vpx_cs = VPX_CS_SRGB;
809 enccfg->g_profile = 1;
810 *img_fmt = VPX_IMG_FMT_I444;
811 return 0;
814 if (codec_caps & VPX_CODEC_CAP_HIGHBITDEPTH) {
815 enccfg->g_profile = 2;
816 *img_fmt = VPX_IMG_FMT_I42016;
817 *
flags |= VPX_CODEC_USE_HIGHBITDEPTH;
818 return 0;
819 }
820 break;
823 if (codec_caps & VPX_CODEC_CAP_HIGHBITDEPTH) {
824 enccfg->g_profile = 3;
825 *img_fmt = VPX_IMG_FMT_I42216;
826 *
flags |= VPX_CODEC_USE_HIGHBITDEPTH;
827 return 0;
828 }
829 break;
832 if (codec_caps & VPX_CODEC_CAP_HIGHBITDEPTH) {
833 enccfg->g_profile = 3;
834 *img_fmt = VPX_IMG_FMT_I44016;
835 *
flags |= VPX_CODEC_USE_HIGHBITDEPTH;
836 return 0;
837 }
838 break;
841 ctx->vpx_cs = VPX_CS_SRGB;
844 if (codec_caps & VPX_CODEC_CAP_HIGHBITDEPTH) {
845 enccfg->g_profile = 3;
846 *img_fmt = VPX_IMG_FMT_I44416;
847 *
flags |= VPX_CODEC_USE_HIGHBITDEPTH;
848 return 0;
849 }
850 break;
851 default:
852 break;
853 }
856 }
857
859 {
860 enum vpx_color_space vpx_cs;
862
864 vpx_cs =
ctx->vpx_cs;
865 } else {
875 default:
878 return;
879 }
880 }
882 }
883
884 #if VPX_ENCODER_ABI_VERSION >= 11
886 {
887 enum vpx_color_range vpx_cr;
892 default:
895 return;
896 }
897
899 }
900 #endif
901 #endif
902
903 /**
904 * Set the target bitrate to VPX library default. Also set CRF to 32 if needed.
905 */
907 struct vpx_codec_enc_cfg *enccfg)
908 {
911 avctx->
bit_rate = enccfg->rc_target_bitrate * 1000;
912 if (enccfg->rc_end_usage == VPX_CQ) {
914 "Bitrate not specified for constrained quality mode, using default of %dkbit/sec\n",
915 enccfg->rc_target_bitrate);
916 } else {
917 enccfg->rc_end_usage = VPX_CQ;
920 "Neither bitrate nor constrained quality specified, using default CRF of %d and bitrate of %dkbit/sec\n",
921 ctx->crf, enccfg->rc_target_bitrate);
922 }
923 }
924
925
926 #if CONFIG_LIBVPX_VP9_ENCODER
927 /**
928 * Keep the target bitrate at 0 to engage constant quality mode. If CRF is not
929 * set, use 32.
930 */
932 struct vpx_codec_enc_cfg *enccfg)
933 {
936 if (enccfg->rc_end_usage != VPX_Q &&
ctx->lossless < 0) {
937 enccfg->rc_end_usage = VPX_Q;
940 "Neither bitrate nor constrained quality specified, using default CRF of %d\n",
942 }
943 }
944 #endif
945
946 /**
947 * Called when the bitrate is not set. It sets appropriate default values for
948 * bitrate and CRF.
949 */
951 struct vpx_codec_enc_cfg *enccfg)
952 {
954 #if CONFIG_LIBVPX_VP9_ENCODER
956 set_vp9_defaults(avctx, enccfg);
957 return;
958 }
959 #endif
961 }
962
964 const struct vpx_codec_iface *iface)
965 {
967 struct vpx_codec_enc_cfg enccfg = { 0 };
968 struct vpx_codec_enc_cfg enccfg_alpha;
971 int res;
972 vpx_img_fmt_t img_fmt = VPX_IMG_FMT_I420;
973 #if CONFIG_LIBVPX_VP9_ENCODER
974 vpx_codec_caps_t codec_caps = vpx_codec_get_caps(iface);
975 vpx_svc_extra_cfg_t svc_params;
976 #endif
978
981
984
985 if ((res = vpx_codec_enc_config_default(iface, &enccfg, 0)) != VPX_CODEC_OK) {
987 vpx_codec_err_to_string(res));
989 }
990
994
995 #if CONFIG_LIBVPX_VP9_ENCODER
999 }
1000 #endif
1001
1006 }
1007
1009
1010 enccfg.g_w = avctx->
width;
1011 enccfg.g_h = avctx->
height;
1014 enccfg.g_threads =
1016 enccfg.g_lag_in_frames=
ctx->lag_in_frames;
1017
1019 enccfg.g_pass = VPX_RC_FIRST_PASS;
1021 enccfg.g_pass = VPX_RC_LAST_PASS;
1022 else
1023 enccfg.g_pass = VPX_RC_ONE_PASS;
1024
1027 enccfg.rc_end_usage = VPX_CBR;
1028 }
else if (
ctx->crf >= 0) {
1029 enccfg.rc_end_usage = VPX_CQ;
1030 #if CONFIG_LIBVPX_VP9_ENCODER
1032 enccfg.rc_end_usage = VPX_Q;
1033 #endif
1034 }
1035
1039 #if CONFIG_LIBVPX_VP9_ENCODER
1040 enccfg.ss_target_bitrate[0] = enccfg.rc_target_bitrate;
1041 #endif
1042 } else {
1043 // Set bitrate to default value. Also sets CRF to default if needed.
1045 }
1046
1048 enccfg.rc_min_quantizer =
1049 enccfg.rc_max_quantizer = 0;
1050 } else {
1051 if (avctx->
qmin >= 0)
1052 enccfg.rc_min_quantizer = avctx->
qmin;
1053 if (avctx->
qmax >= 0)
1054 enccfg.rc_max_quantizer = avctx->
qmax;
1055 }
1056
1057 if (enccfg.rc_end_usage == VPX_CQ
1058 #if CONFIG_LIBVPX_VP9_ENCODER
1059 || enccfg.rc_end_usage == VPX_Q
1060 #endif
1061 ) {
1062 if (
ctx->crf < enccfg.rc_min_quantizer ||
ctx->crf > enccfg.rc_max_quantizer) {
1064 "CQ level %d must be between minimum and maximum quantizer value (%d-%d)\n",
1065 ctx->crf, enccfg.rc_min_quantizer, enccfg.rc_max_quantizer);
1067 }
1068 }
1069
1070 enccfg.rc_dropframe_thresh =
ctx->drop_threshold;
1071
1072 //0-100 (0 => CBR, 100 => VBR)
1075 enccfg.rc_2pass_vbr_minsection_pct =
1078 enccfg.rc_2pass_vbr_maxsection_pct =
1080 #if CONFIG_LIBVPX_VP9_ENCODER
1082 #if VPX_ENCODER_ABI_VERSION >= 14
1083 if (
ctx->corpus_complexity >= 0)
1084 enccfg.rc_2pass_vbr_corpus_complexity =
ctx->corpus_complexity;
1085 #endif
1086 }
1087 #endif
1088
1090 enccfg.rc_buf_sz =
1093 enccfg.rc_buf_initial_sz =
1095 enccfg.rc_buf_optimal_sz = enccfg.rc_buf_sz * 5 / 6;
1096 if (
ctx->rc_undershoot_pct >= 0)
1097 enccfg.rc_undershoot_pct =
ctx->rc_undershoot_pct;
1098 if (
ctx->rc_overshoot_pct >= 0)
1099 enccfg.rc_overshoot_pct =
ctx->rc_overshoot_pct;
1100
1101 //_enc_init() will balk if kf_min_dist differs from max w/VPX_KF_AUTO
1105 enccfg.kf_max_dist = avctx->
gop_size;
1106
1107 if (enccfg.g_pass == VPX_RC_FIRST_PASS)
1108 enccfg.g_lag_in_frames = 0;
1109 else if (enccfg.g_pass == VPX_RC_LAST_PASS) {
1110 int decode_size,
ret;
1111
1115 }
1116
1117 ctx->twopass_stats.sz = strlen(avctx->
stats_in) * 3 / 4;
1122 ctx->twopass_stats.sz);
1123 ctx->twopass_stats.sz = 0;
1125 }
1127 ctx->twopass_stats.sz);
1128 if (decode_size < 0) {
1131 }
1132
1133 ctx->twopass_stats.sz = decode_size;
1134 enccfg.rc_twopass_stats_in =
ctx->twopass_stats;
1135 }
1136
1137 /* 0-3: For non-zero values the encoder increasingly optimizes for reduced
1138 complexity playback on low powered devices at the expense of encode
1139 quality. */
1141 enccfg.g_profile = avctx->
profile;
1142
1144
1148 "Error parsing option '%s = %s'.\n",
1150 }
1151
1152 /* Construct Encoder Context */
1153 res = vpx_codec_enc_init(&
ctx->encoder, iface, &enccfg,
flags);
1154 if (res != VPX_CODEC_OK) {
1158 }
1160
1161 #if CONFIG_LIBVPX_VP9_ENCODER
1163 memset(&svc_params, 0, sizeof(svc_params));
1164 for (
int i = 0;
i < enccfg.ts_number_layers; ++
i) {
1165 svc_params.max_quantizers[
i] = enccfg.rc_max_quantizer;
1166 svc_params.min_quantizers[
i] = enccfg.rc_min_quantizer;
1167 }
1168 svc_params.scaling_factor_num[0] = enccfg.g_h;
1169 svc_params.scaling_factor_den[0] = enccfg.g_h;
1170 #if VPX_ENCODER_ABI_VERSION >= 12
1172 codecctl_intp(avctx, VP9E_SET_SVC_PARAMETERS, (int *)&svc_params);
1173 #endif
1174 }
1175 #endif
1176 if (
ctx->is_alpha) {
1177 enccfg_alpha = enccfg;
1178 res = vpx_codec_enc_init(&
ctx->encoder_alpha, iface, &enccfg_alpha,
flags);
1179 if (res != VPX_CODEC_OK) {
1182 }
1183 }
1184
1185 //codec control failures are currently treated only as warnings
1189 ctx->auto_alt_ref = 1;
1190 if (
ctx->auto_alt_ref >= 0)
1193 if (
ctx->arnr_max_frames >= 0)
1195 if (
ctx->arnr_strength >= 0)
1197 if (
ctx->arnr_type >= 0)
1201
1203 av_log(avctx,
AV_LOG_ERROR,
"Transparency encoding with auto_alt_ref does not work\n");
1205 }
1206
1207 if (
ctx->sharpness >= 0)
1209
1211 codecctl_int(avctx, VP8E_SET_NOISE_SENSITIVITY,
ctx->noise_sensitivity);
1213 }
1217 if (
ctx->max_intra_rate >= 0)
1218 codecctl_int(avctx, VP8E_SET_MAX_INTRA_BITRATE_PCT,
ctx->max_intra_rate);
1219
1220 #if CONFIG_LIBVPX_VP9_ENCODER
1222 if (
ctx->lossless >= 0)
1224 if (
ctx->tile_columns >= 0)
1226 if (
ctx->tile_rows >= 0)
1228 if (
ctx->frame_parallel >= 0)
1229 codecctl_int(avctx, VP9E_SET_FRAME_PARALLEL_DECODING,
ctx->frame_parallel);
1230 if (
ctx->aq_mode >= 0)
1232 set_colorspace(avctx);
1233 #if VPX_ENCODER_ABI_VERSION >= 11
1235 #endif
1236 #if VPX_ENCODER_ABI_VERSION >= 12
1238 #endif
1239 #ifdef VPX_CTRL_VP9E_SET_ROW_MT
1240 if (
ctx->row_mt >= 0)
1242 #endif
1243 #ifdef VPX_CTRL_VP9E_SET_TUNE_CONTENT
1244 if (
ctx->tune_content >= 0)
1246 #endif
1247 #ifdef VPX_CTRL_VP9E_SET_TPL
1248 if (
ctx->tpl_model >= 0)
1250 #endif
1251 #ifdef VPX_CTRL_VP9E_SET_MIN_GF_INTERVAL
1252 if (
ctx->min_gf_interval >= 0)
1254 #endif
1255 }
1256 #endif
1258 if (
ctx->screen_content_mode == 2 &&
ctx->is_alpha) {
1260 "Transparency encoding with screen mode with aggressive rate control not supported\n");
1262 }
1263 codecctl_int(avctx, VP8E_SET_SCREEN_CONTENT_MODE,
ctx->screen_content_mode);
1264 }
1265
1267
1268 //provide dummy value to initialize wrapper, values will be updated each _encode()
1269 vpx_img_wrap(&
ctx->rawimg, img_fmt, avctx->
width, avctx->
height, 1,
1270 (unsigned char*)1);
1271 #if CONFIG_LIBVPX_VP9_ENCODER
1273 ctx->rawimg.bit_depth = enccfg.g_bit_depth;
1274 #endif
1275
1277 if (!cpb_props)
1279
1280 if (enccfg.rc_end_usage == VPX_CBR ||
1281 enccfg.g_pass != VPX_RC_ONE_PASS) {
1285 }
1287
1288 return 0;
1289 }
1290
1292 const struct vpx_codec_cx_pkt *
src,
1294 {
1295 dst->pts =
src->data.frame.pts;
1296 dst->flags =
src->data.frame.flags;
1297 dst->sz =
src->data.frame.sz;
1298 dst->buf =
src->data.frame.buf;
1300 /* For alt-ref frame, don't store PSNR */
1301 if (!(
dst->flags & VPX_FRAME_IS_INVISIBLE)) {
1302 dst->have_sse =
ctx->have_sse;
1303 if (
ctx->have_sse) {
1304 /* associate last-seen SSE to the frame. */
1305 /* Transfers ownership from ctx to dst. */
1306 /* WARNING! This makes the assumption that PSNR_PKT comes
1307 just before the frame it refers to! */
1308 memcpy(
dst->sse,
ctx->sse,
sizeof(
dst->sse));
1310 }
1311 }
1312 }
1313
1314 /**
1315 * Store coded frame information in format suitable for return from encode2().
1316 *
1317 * Write information from @a cx_frame to @a pkt
1318 * @return packet data size on success
1319 * @return a negative AVERROR on error
1320 */
1323 {
1326 uint8_t *side_data;
1327 int pict_type;
1329
1332
1335
1336 if (!!(cx_frame->
flags & VPX_FRAME_IS_KEY)) {
1339 } else {
1341 }
1342
1343 ret = vpx_codec_control(&
ctx->encoder, VP8E_GET_LAST_QUANTIZER_64, &
quality);
1344 if (
ret != VPX_CODEC_OK)
1347 cx_frame->
have_sse ? 3 : 0, pict_type);
1348
1350 /* Beware of the Y/U/V/all order! */
1351 for (
int i = 0;
i < 3; ++
i)
1354 }
1355 if (alpha_cx_frame) {
1358 alpha_cx_frame->
sz + 8);
1359 if (!side_data) {
1362 }
1364 memcpy(side_data + 8, alpha_cx_frame->
buf, alpha_cx_frame->
sz);
1365 }
1369
1371 }
1372
1373 /**
1374 * Queue multiple output frames from the encoder, returning the front-most.
1375 * In cases where vpx_codec_get_cx_data() returns more than 1 frame append
1376 * the frame queue. Return the head frame if available.
1377 * @return Stored frame size
1378 * @return AVERROR(EINVAL) on output size error
1379 * @return AVERROR(ENOMEM) on coded frame queue data allocation error
1380 */
1383 {
1385 const struct vpx_codec_cx_pkt *
pkt;
1386 const void *iter =
NULL;
1388
1389 if (!
ctx->is_alpha && *frame_list) {
1391 /* return the leading frame if we've already begun queueing */
1395 *frame_list = cx_frame->
next;
1397 }
1398
1399 /* consume all available output from the encoder before returning. buffers
1400 are only good through the next vpx_codec call */
1401 while (
pkt = vpx_codec_get_cx_data(encoder, &iter)) {
1402 switch (
pkt->kind) {
1403 case VPX_CODEC_CX_FRAME_PKT:
1404 if (!
ctx->is_alpha && !
size) {
1406
1407 /* avoid storing the frame when the list is empty and we haven't yet
1408 provided a frame for output */
1414 } else {
1416
1417 if (!cx_frame) {
1419 "Frame queue element alloc failed\n");
1421 }
1424
1425 if (!cx_frame->
buf) {
1431 }
1434 }
1435 break;
1436 case VPX_CODEC_STATS_PKT: {
1437 struct vpx_fixed_buf *
stats = &
ctx->twopass_stats;
1439 if (!pkt_out)
1440 break;
1442 &
ctx->twopass_stats_size,
1450 }
1455 break;
1456 }
1457 case VPX_CODEC_PSNR_PKT:
1458 if (!pkt_out)
1459 break;
1466 break;
1467 case VPX_CODEC_CUSTOM_PKT:
1468 //ignore unsupported/unrecognized packet types
1469 break;
1470 }
1471 }
1472
1474 }
1475
1477 vpx_roi_map_t *roi_map, int block_size, int segment_cnt)
1478 {
1479 /**
1480 * range of vpx_roi_map_t.delta_q[i] is [-63, 63]
1481 */
1482 #define MAX_DELTA_Q 63
1483
1485 int nb_rois;
1486 uint32_t self_size;
1487 int segment_id;
1488
1489 /* record the mapping from delta_q to "segment id + 1" in segment_mapping[].
1490 * the range of delta_q is [-MAX_DELTA_Q, MAX_DELTA_Q],
1491 * and its corresponding array index is [0, 2 * MAX_DELTA_Q],
1492 * and so the length of the mapping array is 2 * MAX_DELTA_Q + 1.
1493 * "segment id + 1", so we can say there's no mapping if the value of array element is zero.
1494 */
1496
1497 memset(roi_map, 0, sizeof(*roi_map));
1498
1499 /* segment id 0 in roi_map is reserved for the areas not covered by AVRegionOfInterest.
1500 * segment id 0 in roi_map is also for the areas with AVRegionOfInterest.qoffset near 0.
1501 * (delta_q of segment id 0 is 0).
1502 */
1504 segment_id = 1;
1505
1508 if (!self_size || sd->
size % self_size) {
1511 }
1512 nb_rois = sd->
size / self_size;
1513
1514 /* This list must be iterated from zero because regions are
1515 * defined in order of decreasing importance. So discard less
1516 * important areas if they exceed the segment count.
1517 */
1518 for (
int i = 0;
i < nb_rois;
i++) {
1520 int mapping_index;
1521
1526 }
1527
1530
1532 if (!segment_mapping[mapping_index]) {
1533 if (segment_id == segment_cnt) {
1535 "ROI only supports %d segments (and segment 0 is reserved for non-ROIs), skipping the left ones.\n",
1536 segment_cnt);
1537 break;
1538 }
1539
1540 segment_mapping[mapping_index] = segment_id + 1;
1541 roi_map->delta_q[segment_id] =
delta_q;
1542 segment_id++;
1543 }
1544 }
1545
1546 roi_map->rows = (frame_height + block_size - 1) / block_size;
1547 roi_map->cols = (frame_width + block_size - 1) / block_size;
1548 roi_map->roi_map =
av_calloc(roi_map->rows * roi_map->cols,
sizeof(*roi_map->roi_map));
1549 if (!roi_map->roi_map) {
1552 }
1553
1554 /* This list must be iterated in reverse, so for the case that
1555 * two regions are overlapping, the more important area takes effect.
1556 */
1557 for (
int i = nb_rois - 1;
i >= 0;
i--) {
1559 int mapping_value;
1560 int starty, endy, startx, endx;
1561
1563
1564 starty =
av_clip(roi->
top / block_size, 0, roi_map->rows);
1565 endy =
av_clip((roi->
bottom + block_size - 1) / block_size, 0, roi_map->rows);
1566 startx =
av_clip(roi->
left / block_size, 0, roi_map->cols);
1567 endx =
av_clip((roi->
right + block_size - 1) / block_size, 0, roi_map->cols);
1568
1571
1573 if (mapping_value) {
1574 for (int y = starty; y < endy; y++)
1575 for (int x = startx; x < endx; x++)
1576 roi_map->roi_map[x + y * roi_map->cols] = mapping_value - 1;
1577 }
1578 }
1579
1580 return 0;
1581 }
1582
1584 {
1586
1587 #ifdef VPX_CTRL_VP9E_SET_ROI_MAP
1588 int version = vpx_codec_version();
1589 int major = VPX_VERSION_MAJOR(
version);
1590 int minor = VPX_VERSION_MINOR(
version);
1591 int patch = VPX_VERSION_PATCH(
version);
1592
1593 if (major > 1 || (major == 1 && minor > 8) || (major == 1 && minor == 8 && patch >= 1)) {
1594 vpx_roi_map_t roi_map;
1595 const int segment_cnt = 8;
1596 const int block_size = 8;
1598
1599 if (
ctx->aq_mode > 0 ||
ctx->cpu_used < 5 ||
ctx->deadline != VPX_DL_REALTIME) {
1600 if (!
ctx->roi_warned) {
1601 ctx->roi_warned = 1;
1603 "and deadline is REALTIME, so skipping ROI.\n");
1605 }
1606 }
1607
1608 ret =
set_roi_map(avctx, sd, frame_width, frame_height, &roi_map, block_size, segment_cnt);
1612 }
1613
1614 memset(roi_map.ref_frame, -1, sizeof(roi_map.ref_frame));
1615
1616 if (vpx_codec_control(&
ctx->encoder, VP9E_SET_ROI_MAP, &roi_map)) {
1619 }
1622 }
1623 #endif
1624
1625 if (!
ctx->roi_warned) {
1626 ctx->roi_warned = 1;
1627 av_log(avctx,
AV_LOG_WARNING,
"ROI is not supported, please upgrade libvpx to version >= 1.8.1. "
1628 "You may need to rebuild ffmpeg.\n");
1629 }
1630 return 0;
1631 }
1632
1634 {
1635 vpx_roi_map_t roi_map;
1636 const int segment_cnt = 4;
1637 const int block_size = 16;
1639
1640 int ret =
set_roi_map(avctx, sd, frame_width, frame_height, &roi_map, block_size, segment_cnt);
1644 }
1645
1646 if (vpx_codec_control(&
ctx->encoder, VP8E_SET_ROI_MAP, &roi_map)) {
1649 }
1650
1653 }
1654
1656 {
1658 struct vpx_image *rawimg_alpha = &
ctx->rawimg_alpha;
1659 unsigned char **
planes = rawimg_alpha->planes;
1660 int *
stride = rawimg_alpha->stride;
1661
1662 if (!
planes[VPX_PLANE_U] ||
1664 width != (
int)rawimg_alpha->d_w ||
1665 height != (
int)rawimg_alpha->d_h) {
1668
1669 vpx_img_wrap(rawimg_alpha, VPX_IMG_FMT_I420,
width,
height, 1,
1670 (unsigned char*)1);
1675
1678 }
1679
1680 return 0;
1681 }
1682
1685 {
1687 struct vpx_image *rawimg =
NULL;
1688 struct vpx_image *rawimg_alpha =
NULL;
1690 int res, coded_size;
1691 vpx_enc_frame_flags_t
flags = 0;
1692 const struct vpx_codec_enc_cfg *enccfg =
ctx->encoder.config.enc;
1693 vpx_svc_layer_id_t layer_id;
1694 int layer_id_valid = 0;
1696
1697 if (avctx->
qmax >= 0 && enccfg->rc_max_quantizer != avctx->
qmax) {
1698 struct vpx_codec_enc_cfg cfg = *enccfg;
1699 cfg.rc_max_quantizer = avctx->
qmax;
1700 res = vpx_codec_enc_config_set(&
ctx->encoder, &cfg);
1701 if (res != VPX_CODEC_OK) {
1704 }
1705 }
1706
1709 rawimg = &
ctx->rawimg;
1710 rawimg->planes[VPX_PLANE_Y] =
frame->data[0];
1711 rawimg->planes[VPX_PLANE_U] =
frame->data[1];
1712 rawimg->planes[VPX_PLANE_V] =
frame->data[2];
1713 rawimg->stride[VPX_PLANE_Y] =
frame->linesize[0];
1714 rawimg->stride[VPX_PLANE_U] =
frame->linesize[1];
1715 rawimg->stride[VPX_PLANE_V] =
frame->linesize[2];
1716 if (
ctx->is_alpha) {
1717 rawimg_alpha = &
ctx->rawimg_alpha;
1719 if (res < 0)
1720 return res;
1721 rawimg_alpha->planes[VPX_PLANE_Y] =
frame->data[3];
1722 rawimg_alpha->stride[VPX_PLANE_Y] =
frame->linesize[3];
1723 }
1724 timestamp =
frame->pts;
1725 #if VPX_IMAGE_ABI_VERSION >= 4
1726 switch (
frame->color_range) {
1728 rawimg->range = VPX_CR_STUDIO_RANGE;
1729 break;
1731 rawimg->range = VPX_CR_FULL_RANGE;
1732 break;
1733 }
1734 #endif
1736 flags |= VPX_EFLAG_FORCE_KF;
1737 if (
frame->metadata) {
1739 if (en) {
1741 }
1742
1743 memset(&layer_id, 0, sizeof(layer_id));
1744
1746 if (en) {
1747 layer_id.temporal_layer_id = strtoul(en->
value,
NULL, 10);
1748 #ifdef VPX_CTRL_VP9E_SET_MAX_INTER_BITRATE_PCT
1749 layer_id.temporal_layer_id_per_spatial[0] = layer_id.temporal_layer_id;
1750 #endif
1751 layer_id_valid = 1;
1752 }
1753 #if CONFIG_LIBVPX_VP9_ENCODER && defined(VPX_CTRL_VP9E_SET_MAX_INTER_BITRATE_PCT)
1755
1756 if (en) {
1758 int ret = vpx_parse_ref_frame_config(&
ctx->ref_frame_config,
1759 enccfg->ss_number_layers, en->
value);
1762 "Error parsing ref_frame_config option %s.\n", en->
value);
1764 }
1765
1766 codecctl_intp(avctx, VP9E_SET_SVC_REF_FRAME_CONFIG, (
int *)&
ctx->ref_frame_config);
1767 } else {
1769 "Ignoring ref-frame-config for a non-VP9 codec\n");
1770 }
1771 }
1772 #endif
1773 }
1774
1775 if (sd) {
1778 } else {
1780 }
1781 }
1782
1785 if (res < 0)
1786 return res;
1787 }
1788 }
1789
1790 // this is for encoding with preset temporal layering patterns defined in
1791 // set_temporal_layer_pattern function.
1792 if (enccfg->ts_number_layers > 1 &&
ctx->ts_layer_flags) {
1793 if (
flags & VPX_EFLAG_FORCE_KF) {
1794 // keyframe, reset temporal layering.
1795 ctx->current_temporal_idx = 0;
1796 flags = VPX_EFLAG_FORCE_KF;
1797 } else {
1799 }
1800
1801 /* get the flags from the temporal layer configuration. */
1802 flags |=
ctx->ts_layer_flags[
ctx->current_temporal_idx];
1803
1804 memset(&layer_id, 0, sizeof(layer_id));
1805 #if VPX_ENCODER_ABI_VERSION >= 12
1806 layer_id.spatial_layer_id = 0;
1807 #endif
1808 layer_id.temporal_layer_id = enccfg->ts_layer_id[
ctx->current_temporal_idx];
1809 #ifdef VPX_CTRL_VP9E_SET_MAX_INTER_BITRATE_PCT
1810 layer_id.temporal_layer_id_per_spatial[0] = layer_id.temporal_layer_id;
1811 #endif
1812 layer_id_valid = 1;
1813 }
1814
1815 if (layer_id_valid) {
1817 codecctl_int(avctx, VP8E_SET_TEMPORAL_LAYER_ID, layer_id.temporal_layer_id);
1818 }
1819 #if CONFIG_LIBVPX_VP9_ENCODER && VPX_ENCODER_ABI_VERSION >= 12
1821 codecctl_intp(avctx, VP9E_SET_SVC_LAYER_ID, (int *)&layer_id);
1822 }
1823 #endif
1824 }
1825
1828 "Frame duration too large: %"PRId64
"\n",
frame->duration);
1833 else {
1835 }
1836
1837 res = vpx_codec_encode(&
ctx->encoder, rawimg, timestamp,
1839 if (res != VPX_CODEC_OK) {
1842 }
1843
1844 if (
ctx->is_alpha) {
1845 res = vpx_codec_encode(&
ctx->encoder_alpha, rawimg_alpha, timestamp,
1847 if (res != VPX_CODEC_OK) {
1850 }
1851 }
1852
1854 if (
ctx->is_alpha) {
1856
1857 if (
ctx->coded_frame_list &&
ctx->alpha_coded_frame_list) {
1861 /* return the leading frame if we've already begun queueing */
1862 coded_size =
storeframe(avctx, cx_frame, alpha_cx_frame,
pkt);
1863 if (coded_size < 0)
1864 return coded_size;
1865 ctx->coded_frame_list = cx_frame->
next;
1866 ctx->alpha_coded_frame_list = alpha_cx_frame->
next;
1869 }
1870 }
1871
1874
1878 b64_size);
1880 }
1882 ctx->twopass_stats.sz);
1883 }
else if (enccfg->ts_number_layers > 1 &&
ctx->ts_layer_flags) {
1884 ctx->current_temporal_idx = (
ctx->current_temporal_idx + 1) % enccfg->ts_periodicity;
1885 }
1886
1887 *got_packet = !!coded_size;
1888 return 0;
1889 }
1890
1891 #define OFFSET(x) offsetof(VPxContext, x)
1892 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
1893
1894 #define COMMON_OPTIONS \
1895 { "lag-in-frames", "Number of frames to look ahead for " \
1896 "alternate reference frame selection", OFFSET(lag_in_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
1897 { "arnr-maxframes", "altref noise reduction max frame count", OFFSET(arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
1898 { "arnr-strength", "altref noise reduction filter strength", OFFSET(arnr_strength), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
1899 { "arnr-type", "altref noise reduction filter type", OFFSET(arnr_type), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE, .unit = "arnr_type"}, \
1900 { "backward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0, VE, .unit = "arnr_type" }, \
1901 { "forward", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0, VE, .unit = "arnr_type" }, \
1902 { "centered", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = 3}, 0, 0, VE, .unit = "arnr_type" }, \
1903 { "tune", "Tune the encoding to a specific scenario", OFFSET(tune), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE, .unit = "tune"}, \
1904 { "psnr", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VP8_TUNE_PSNR}, 0, 0, VE, .unit = "tune"}, \
1905 { "ssim", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VP8_TUNE_SSIM}, 0, 0, VE, .unit = "tune"}, \
1906 { "deadline", "Time to spend encoding, in microseconds.", OFFSET(deadline), AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, .unit = "quality"}, \
1907 { "best", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_BEST_QUALITY}, 0, 0, VE, .unit = "quality"}, \
1908 { "good", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_GOOD_QUALITY}, 0, 0, VE, .unit = "quality"}, \
1909 { "realtime", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = VPX_DL_REALTIME}, 0, 0, VE, .unit = "quality"}, \
1910 { "error-resilient", "Error resilience configuration", OFFSET(error_resilient), AV_OPT_TYPE_FLAGS, {.i64 = 0}, INT_MIN, INT_MAX, VE, .unit = "er"}, \
1911 { "max-intra-rate", "Maximum I-frame bitrate (pct) 0=unlimited", OFFSET(max_intra_rate), AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX, VE}, \
1912 { "default", "Improve resiliency against losses of whole frames", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_DEFAULT}, 0, 0, VE, .unit = "er"}, \
1913 { "partitions", "The frame partitions are independently decodable " \
1914 "by the bool decoder, meaning that partitions can be decoded even " \
1915 "though earlier partitions have been lost. Note that intra prediction" \
1916 " is still done over the partition boundary.", 0, AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_PARTITIONS}, 0, 0, VE, .unit = "er"}, \
1917 { "crf", "Select the quality for constant quality mode", offsetof(VPxContext, crf), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 63, VE }, \
1918 { "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 }, \
1919 { "drop-threshold", "Frame drop threshold", offsetof(VPxContext, drop_threshold), AV_OPT_TYPE_INT, {.i64 = 0 }, INT_MIN, INT_MAX, VE }, \
1920 { "noise-sensitivity", "Noise sensitivity", OFFSET(noise_sensitivity), AV_OPT_TYPE_INT, {.i64 = 0 }, 0, 4, VE}, \
1921 { "undershoot-pct", "Datarate undershoot (min) target (%)", OFFSET(rc_undershoot_pct), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 100, VE }, \
1922 { "overshoot-pct", "Datarate overshoot (max) target (%)", OFFSET(rc_overshoot_pct), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 1000, VE }, \
1923 { "ts-parameters", "Temporal scaling configuration using a :-separated list of key=value parameters", OFFSET(vpx_ts_parameters), AV_OPT_TYPE_DICT, {.str=NULL}, 0, 0, VE}, \
1924
1925 #define LEGACY_OPTIONS \
1926 {"speed", "", offsetof(VPxContext, cpu_used), AV_OPT_TYPE_INT, {.i64 = 1}, -16, 16, VE}, \
1927 {"quality", "", offsetof(VPxContext, deadline), AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX, VE, .unit = "quality"}, \
1928 {"vp8flags", "", offsetof(VPxContext, flags), AV_OPT_TYPE_FLAGS, {.i64 = 0}, 0, UINT_MAX, VE, .unit = "flags"}, \
1929 {"error_resilient", "enable error resilience", 0, AV_OPT_TYPE_CONST, {.i64 = VP8F_ERROR_RESILIENT}, INT_MIN, INT_MAX, VE, .unit = "flags"}, \
1930 {"altref", "enable use of alternate reference frames (VP8/2-pass only)", 0, AV_OPT_TYPE_CONST, {.i64 = VP8F_AUTO_ALT_REF}, INT_MIN, INT_MAX, VE, .unit = "flags"}, \
1931 {"arnr_max_frames", "altref noise reduction max frame count", offsetof(VPxContext, arnr_max_frames), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 15, VE}, \
1932 {"arnr_strength", "altref noise reduction filter strength", offsetof(VPxContext, arnr_strength), AV_OPT_TYPE_INT, {.i64 = 3}, 0, 6, VE}, \
1933 {"arnr_type", "altref noise reduction filter type", offsetof(VPxContext, arnr_type), AV_OPT_TYPE_INT, {.i64 = 3}, 1, 3, VE}, \
1934 {"rc_lookahead", "Number of frames to look ahead for alternate reference frame selection", offsetof(VPxContext, lag_in_frames), AV_OPT_TYPE_INT, {.i64 = 25}, 0, 25, VE}, \
1935 {"sharpness", "Increase sharpness at the expense of lower PSNR", offsetof(VPxContext, sharpness), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 7, VE},
1936
1937 #if CONFIG_LIBVPX_VP8_ENCODER
1938 static const AVOption vp8_options[] = {
1940 { "auto-alt-ref", "Enable use of alternate reference "
1943 {
"screen-content-mode",
"Encoder screen content mode",
OFFSET(screen_content_mode),
AV_OPT_TYPE_INT, {.i64 = -1}, -1, 2,
VE},
1946 };
1947 #endif
1948
1949 #if CONFIG_LIBVPX_VP9_ENCODER
1950 static const AVOption vp9_options[] = {
1952 { "auto-alt-ref", "Enable use of alternate reference "
1956 {
"tile-columns",
"Number of tile columns to use, log2",
OFFSET(tile_columns),
AV_OPT_TYPE_INT, {.i64 = -1}, -1, 6,
VE},
1958 {
"frame-parallel",
"Enable frame parallel decodability features",
OFFSET(frame_parallel),
AV_OPT_TYPE_BOOL,{.i64 = -1}, -1, 1,
VE},
1959 #if VPX_ENCODER_ABI_VERSION >= 12
1960 {
"aq-mode",
"adaptive quantization mode",
OFFSET(aq_mode),
AV_OPT_TYPE_INT, {.i64 = -1}, -1, 4,
VE, .unit =
"aq_mode"},
1961 #else
1962 {
"aq-mode",
"adaptive quantization mode",
OFFSET(aq_mode),
AV_OPT_TYPE_INT, {.i64 = -1}, -1, 3,
VE, .unit =
"aq_mode"},
1963 #endif
1965 {
"variance",
"Variance based Aq", 0,
AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0,
VE, .unit =
"aq_mode" },
1966 {
"complexity",
"Complexity based Aq", 0,
AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0,
VE, .unit =
"aq_mode" },
1967 {
"cyclic",
"Cyclic Refresh Aq", 0,
AV_OPT_TYPE_CONST, {.i64 = 3}, 0, 0,
VE, .unit =
"aq_mode" },
1968 #if VPX_ENCODER_ABI_VERSION >= 12
1969 {
"equator360",
"360 video Aq", 0,
AV_OPT_TYPE_CONST, {.i64 = 4}, 0, 0,
VE, .unit =
"aq_mode" },
1971 #endif
1972 #ifdef VPX_CTRL_VP9E_SET_ROW_MT
1974 #endif
1975 #ifdef VPX_CTRL_VP9E_SET_TUNE_CONTENT
1976 #if VPX_ENCODER_ABI_VERSION >= 14
1977 {
"tune-content",
"Tune content type",
OFFSET(tune_content),
AV_OPT_TYPE_INT, {.i64 = -1}, -1, 2,
VE, .unit =
"tune_content" },
1978 #else
1979 {
"tune-content",
"Tune content type",
OFFSET(tune_content),
AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1,
VE, .unit =
"tune_content" },
1980 #endif
1981 {
"default",
"Regular video content", 0,
AV_OPT_TYPE_CONST, {.i64 = 0}, 0, 0,
VE, .unit =
"tune_content" },
1982 {
"screen",
"Screen capture content", 0,
AV_OPT_TYPE_CONST, {.i64 = 1}, 0, 0,
VE, .unit =
"tune_content" },
1983 #if VPX_ENCODER_ABI_VERSION >= 14
1984 {
"film",
"Film content; improves grain retention", 0,
AV_OPT_TYPE_CONST, {.i64 = 2}, 0, 0,
VE, .unit =
"tune_content" },
1985 #endif
1986 #endif
1987 #if VPX_ENCODER_ABI_VERSION >= 14
1988 {
"corpus-complexity",
"corpus vbr complexity midpoint",
OFFSET(corpus_complexity),
AV_OPT_TYPE_INT, {.i64 = -1}, -1, 10000,
VE },
1989 #endif
1990 #ifdef VPX_CTRL_VP9E_SET_TPL
1992 #endif
1993 #ifdef VPX_CTRL_VP9E_SET_MIN_GF_INTERVAL
1994 {
"min-gf-interval",
"Minimum golden/alternate reference frame interval",
OFFSET(min_gf_interval),
AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX,
VE },
1995 #endif
1998 };
1999 #endif
2000
2001 #undef COMMON_OPTIONS
2002 #undef LEGACY_OPTIONS
2003
2005 { "b", "0" },
2006 { "qmin", "-1" },
2007 { "qmax", "-1" },
2008 { "g", "-1" },
2009 { "keyint_min", "-1" },
2011 };
2012
2013 #if CONFIG_LIBVPX_VP8_ENCODER
2015 {
2016 return vpx_init(avctx, vpx_codec_vp8_cx());
2017 }
2018
2019 static const AVClass class_vp8 = {
2022 .option = vp8_options,
2024 };
2025
2034 .priv_data_size = sizeof(VPxContext),
2043 .p.priv_class = &class_vp8,
2045 .p.wrapper_name = "libvpx",
2046 };
2047 #endif /* CONFIG_LIBVPX_VP8_ENCODER */
2048
2049 #if CONFIG_LIBVPX_VP9_ENCODER
2051 {
2052 return vpx_init(avctx, vpx_codec_vp9_cx());
2053 }
2054
2063 };
2064
2083 };
2084
2089 int *out_num)
2090 {
2092 vpx_codec_caps_t codec_caps = vpx_codec_get_caps(vpx_codec_vp9_cx());
2093 if (codec_caps & VPX_CODEC_CAP_HIGHBITDEPTH) {
2094 *
out = vp9_pix_fmts_highbd;
2096 } else {
2097 *
out = vp9_pix_fmts_highcol;
2099 }
2100 return 0;
2101 }
2102
2104 }
2105
2106 static const AVClass class_vp9 = {
2109 .option = vp9_options,
2111 };
2112
2114 .
p.
name =
"libvpx-vp9",
2122 .p.priv_class = &class_vp9,
2123 .p.wrapper_name = "libvpx",
2124 .priv_data_size = sizeof(VPxContext),
2133 .get_supported_config = vp9_get_supported_config,
2134 };
2135 #endif /* CONFIG_LIBVPX_VP9_ENCODER */