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
38
39 /**
40 * Portion of struct vpx_codec_cx_pkt from vpx_encoder.h.
41 * One encoded frame returned from the library.
42 */
44 void *
buf;
/**< compressed data buffer */
45 size_t sz;
/**< length of compressed data */
46 int64_t
pts;
/**< time stamp to show frame
47 (in timebase units) */
48 unsigned long duration;
/**< duration to show frame
49 (in timebase units) */
50 uint32_t
flags;
/**< flags for this frame */
52 };
53
54 typedef struct VP8EncoderContext {
56 struct vpx_codec_ctx encoder;
57 struct vpx_image rawimg;
58 struct vpx_fixed_buf twopass_stats;
61
63 /**
64 * VP8 specific flags, see VP8F_* below.
65 */
67 #define VP8F_ERROR_RESILIENT 0x00000001 ///< Enable measures appropriate for streaming over lossy links
68 #define VP8F_AUTO_ALT_REF 0x00000002
///< Enable automatic alternate reference frame generation
69
71
75
81
82 /** String mappings for enum vp8e_enc_control_id */
84 [VP8E_UPD_ENTROPY] = "VP8E_UPD_ENTROPY",
85 [VP8E_UPD_REFERENCE] = "VP8E_UPD_REFERENCE",
86 [VP8E_USE_REFERENCE] = "VP8E_USE_REFERENCE",
87 [VP8E_SET_ROI_MAP] = "VP8E_SET_ROI_MAP",
88 [VP8E_SET_ACTIVEMAP] = "VP8E_SET_ACTIVEMAP",
89 [VP8E_SET_SCALEMODE] = "VP8E_SET_SCALEMODE",
90 [VP8E_SET_CPUUSED] = "VP8E_SET_CPUUSED",
91 [VP8E_SET_ENABLEAUTOALTREF] = "VP8E_SET_ENABLEAUTOALTREF",
92 [VP8E_SET_NOISE_SENSITIVITY] = "VP8E_SET_NOISE_SENSITIVITY",
93 [VP8E_SET_SHARPNESS] = "VP8E_SET_SHARPNESS",
94 [VP8E_SET_STATIC_THRESHOLD] = "VP8E_SET_STATIC_THRESHOLD",
95 [VP8E_SET_TOKEN_PARTITIONS] = "VP8E_SET_TOKEN_PARTITIONS",
96 [VP8E_GET_LAST_QUANTIZER] = "VP8E_GET_LAST_QUANTIZER",
97 [VP8E_SET_ARNR_MAXFRAMES] = "VP8E_SET_ARNR_MAXFRAMES",
98 [VP8E_SET_ARNR_STRENGTH] = "VP8E_SET_ARNR_STRENGTH",
99 [VP8E_SET_ARNR_TYPE] = "VP8E_SET_ARNR_TYPE",
100 [VP8E_SET_CQ_LEVEL] = "VP8E_SET_CQ_LEVEL",
101 [VP8E_SET_MAX_INTRA_BITRATE_PCT] = "VP8E_SET_MAX_INTRA_BITRATE_PCT",
102 };
103
105 {
107 const char *error = vpx_codec_error(&ctx->
encoder);
108 const char *detail = vpx_codec_error_detail(&ctx->
encoder);
109
111 if (detail)
113 }
114
116 const struct vpx_codec_enc_cfg *cfg)
117 {
120
121 av_log(avctx, level,
"vpx_codec_enc_cfg\n");
122 av_log(avctx, level,
"generic settings\n"
123 " %*s%u\n %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
124 " %*s{%u/%u}\n %*s%u\n %*s%d\n %*s%u\n",
125 width, "g_usage:", cfg->g_usage,
126 width, "g_threads:", cfg->g_threads,
127 width, "g_profile:", cfg->g_profile,
128 width, "g_w:", cfg->g_w,
129 width, "g_h:", cfg->g_h,
130 width, "g_timebase:", cfg->g_timebase.num, cfg->g_timebase.den,
131 width, "g_error_resilient:", cfg->g_error_resilient,
132 width, "g_pass:", cfg->g_pass,
133 width, "g_lag_in_frames:", cfg->g_lag_in_frames);
134 av_log(avctx, level,
"rate control settings\n"
135 " %*s%u\n %*s%u\n %*s%u\n %*s%u\n"
136 " %*s%d\n %*s%p(%zu)\n %*s%u\n",
137 width, "rc_dropframe_thresh:", cfg->rc_dropframe_thresh,
138 width, "rc_resize_allowed:", cfg->rc_resize_allowed,
139 width, "rc_resize_up_thresh:", cfg->rc_resize_up_thresh,
140 width, "rc_resize_down_thresh:", cfg->rc_resize_down_thresh,
141 width, "rc_end_usage:", cfg->rc_end_usage,
142 width, "rc_twopass_stats_in:", cfg->rc_twopass_stats_in.buf, cfg->rc_twopass_stats_in.sz,
143 width, "rc_target_bitrate:", cfg->rc_target_bitrate);
144 av_log(avctx, level,
"quantizer settings\n"
145 " %*s%u\n %*s%u\n",
146 width, "rc_min_quantizer:", cfg->rc_min_quantizer,
147 width, "rc_max_quantizer:", cfg->rc_max_quantizer);
148 av_log(avctx, level,
"bitrate tolerance\n"
149 " %*s%u\n %*s%u\n",
150 width, "rc_undershoot_pct:", cfg->rc_undershoot_pct,
151 width, "rc_overshoot_pct:", cfg->rc_overshoot_pct);
152 av_log(avctx, level,
"decoder buffer model\n"
153 " %*s%u\n %*s%u\n %*s%u\n",
154 width, "rc_buf_sz:", cfg->rc_buf_sz,
155 width, "rc_buf_initial_sz:", cfg->rc_buf_initial_sz,
156 width, "rc_buf_optimal_sz:", cfg->rc_buf_optimal_sz);
157 av_log(avctx, level,
"2 pass rate control settings\n"
158 " %*s%u\n %*s%u\n %*s%u\n",
159 width, "rc_2pass_vbr_bias_pct:", cfg->rc_2pass_vbr_bias_pct,
160 width, "rc_2pass_vbr_minsection_pct:", cfg->rc_2pass_vbr_minsection_pct,
161 width, "rc_2pass_vbr_maxsection_pct:", cfg->rc_2pass_vbr_maxsection_pct);
162 av_log(avctx, level,
"keyframing settings\n"
163 " %*s%d\n %*s%u\n %*s%u\n",
164 width, "kf_mode:", cfg->kf_mode,
165 width, "kf_min_dist:", cfg->kf_min_dist,
166 width, "kf_max_dist:", cfg->kf_max_dist);
167 av_log(avctx, level,
"\n");
168 }
169
171 {
173
176 *p = cx_frame;
178 }
179
181 {
184 }
185
187 {
189
190 while (p) {
193 p = list;
194 }
195 }
196
198 enum vp8e_enc_control_id id, int val)
199 {
203 int res;
204
207
208 res = vpx_codec_control(&ctx->
encoder,
id, val);
209 if (res != VPX_CODEC_OK) {
210 snprintf(buf,
sizeof(buf),
"Failed to set %s codec control",
213 }
214
215 return res == VPX_CODEC_OK ? 0 :
AVERROR(EINVAL);
216 }
217
219 {
221
222 vpx_codec_destroy(&ctx->
encoder);
227 return 0;
228 }
229
231 {
233 const struct vpx_codec_iface *iface = &vpx_codec_vp8_cx_algo;
234 struct vpx_codec_enc_cfg enccfg;
235 int res;
236
239
240 if ((res = vpx_codec_enc_config_default(iface, &enccfg, 0)) != VPX_CODEC_OK) {
242 vpx_codec_err_to_string(res));
244 }
245
250 }
251
253
254 enccfg.g_w = avctx->
width;
255 enccfg.g_h = avctx->
height;
260
262 enccfg.g_pass = VPX_RC_FIRST_PASS;
264 enccfg.g_pass = VPX_RC_LAST_PASS;
265 else
266 enccfg.g_pass = VPX_RC_ONE_PASS;
267
270 enccfg.rc_end_usage = VPX_CBR;
272 enccfg.rc_end_usage = VPX_CQ;
273
277 } else {
278 if (enccfg.rc_end_usage == VPX_CQ) {
279 enccfg.rc_target_bitrate = 1000000;
280 } else {
281 avctx->
bit_rate = enccfg.rc_target_bitrate * 1000;
283 "Neither bitrate nor constrained quality specified, using default bitrate of %dkbit/sec\n",
284 enccfg.rc_target_bitrate);
285 }
286 }
287
289 enccfg.rc_min_quantizer = avctx->
qmin;
291 enccfg.rc_max_quantizer = avctx->
qmax;
293
294 //0-100 (0 => CBR, 100 => VBR)
297 enccfg.rc_2pass_vbr_minsection_pct =
300 enccfg.rc_2pass_vbr_maxsection_pct =
302
304 enccfg.rc_buf_sz =
307 enccfg.rc_buf_initial_sz =
309 enccfg.rc_buf_optimal_sz = enccfg.rc_buf_sz * 5 / 6;
311
312 //_enc_init() will balk if kf_min_dist differs from max w/VPX_KF_AUTO
316 enccfg.kf_max_dist = avctx->
gop_size;
317
318 if (enccfg.g_pass == VPX_RC_FIRST_PASS)
319 enccfg.g_lag_in_frames = 0;
320 else if (enccfg.g_pass == VPX_RC_LAST_PASS) {
321 int decode_size;
322
326 }
327
332 "Stat buffer alloc (%zu bytes) failed\n",
335 }
338 if (decode_size < 0) {
341 }
342
345 }
346
347 /* 0-3: For non-zero values the encoder increasingly optimizes for reduced
348 complexity playback on low powered devices at the expense of encode
349 quality. */
351 enccfg.g_profile = avctx->
profile;
352
354
356 /* Construct Encoder Context */
357 res = vpx_codec_enc_init(&ctx->
encoder, iface, &enccfg, 0);
358 if (res != VPX_CODEC_OK) {
361 }
362
363 //codec control failures are currently treated only as warnings
383
385
386 //provide dummy value to initialize wrapper, values will be updated each _encode()
388 (unsigned char*)1);
389
395 }
396 return 0;
397 }
398
400 const struct vpx_codec_cx_pkt *src)
401 {
402 dst->
pts = src->data.frame.pts;
403 dst->
duration = src->data.frame.duration;
404 dst->
flags = src->data.frame.flags;
405 dst->
sz = src->data.frame.sz;
406 dst->
buf = src->data.frame.buf;
407 }
408
409 /**
410 * Store coded frame information in format suitable for return from encode2().
411 *
412 * Write information from @a cx_frame to @a pkt
413 * @return packet data size on success
414 * @return a negative AVERROR on error
415 */
418 {
420 if (ret >= 0) {
423 coded_frame->
pts = cx_frame->
pts;
425
429 } else
431 } else {
432 return ret;
433 }
435 }
436
437 /**
438 * Queue multiple output frames from the encoder, returning the front-most.
439 * In cases where vpx_codec_get_cx_data() returns more than 1 frame append
440 * the frame queue. Return the head frame if available.
441 * @return Stored frame size
442 * @return AVERROR(EINVAL) on output size error
443 * @return AVERROR(ENOMEM) on coded frame queue data allocation error
444 */
447 {
449 const struct vpx_codec_cx_pkt *
pkt;
450 const void *iter =
NULL;
452
455 /* return the leading frame if we've already begun queueing */
456 size =
storeframe(avctx, cx_frame, pkt_out, coded_frame);
457 if (size < 0)
461 }
462
463 /* consume all available output from the encoder before returning. buffers
464 are only good through the next vpx_codec call */
465 while ((pkt = vpx_codec_get_cx_data(&ctx->
encoder, &iter))) {
466 switch (pkt->kind) {
467 case VPX_CODEC_CX_FRAME_PKT:
468 if (!size) {
470
471 /* avoid storing the frame when the list is empty and we haven't yet
472 provided a frame for output */
475 size =
storeframe(avctx, &cx_frame, pkt_out, coded_frame);
476 if (size < 0)
478 } else {
481
482 if (!cx_frame) {
484 "Frame queue element alloc failed\n");
486 }
489
490 if (!cx_frame->
buf) {
492 "Data buffer alloc (%zu bytes) failed\n",
496 }
497 memcpy(cx_frame->
buf, pkt->data.frame.buf, pkt->data.frame.sz);
499 }
500 break;
501 case VPX_CODEC_STATS_PKT: {
504 stats->sz + pkt->data.twopass_stats.sz);
505 if (!stats->buf) {
508 }
509 memcpy((
uint8_t*)stats->buf + stats->sz,
510 pkt->data.twopass_stats.buf, pkt->data.twopass_stats.sz);
511 stats->sz += pkt->data.twopass_stats.sz;
512 break;
513 }
514 case VPX_CODEC_PSNR_PKT: //FIXME add support for CODEC_FLAG_PSNR
515 case VPX_CODEC_CUSTOM_PKT:
516 //ignore unsupported/unrecognized packet types
517 break;
518 }
519 }
520
522 }
523
526 {
528 struct vpx_image *rawimg =
NULL;
529 int64_t timestamp = 0;
530 int res, coded_size;
531 vpx_enc_frame_flags_t
flags = 0;
532
533 if (frame) {
535 rawimg->planes[VPX_PLANE_Y] = frame->
data[0];
536 rawimg->planes[VPX_PLANE_U] = frame->
data[1];
537 rawimg->planes[VPX_PLANE_V] = frame->
data[2];
538 rawimg->stride[VPX_PLANE_Y] = frame->
linesize[0];
539 rawimg->stride[VPX_PLANE_U] = frame->
linesize[1];
540 rawimg->stride[VPX_PLANE_V] = frame->
linesize[2];
541 timestamp = frame->
pts;
543 flags |= VPX_EFLAG_FORCE_KF;
544 }
545
546 res = vpx_codec_encode(&ctx->
encoder, rawimg, timestamp,
548 if (res != VPX_CODEC_OK) {
551 }
553
556
560 b64_size);
562 }
565 }
566
567 *got_packet = !!coded_size;
568 return 0;
569 }
570
571 #define OFFSET(x) offsetof(VP8Context, x)
572 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
574 {
"cpu-used",
"Quality/Speed ratio modifier",
OFFSET(cpu_used),
AV_OPT_TYPE_INT, {.i64 = INT_MIN}, INT_MIN, INT_MAX,
VE},
575 { "auto-alt-ref", "Enable use of alternate reference "
577 { "lag-in-frames", "Number of frames to look ahead for "
579 {
"arnr-maxframes",
"altref noise reduction max frame count",
OFFSET(arnr_max_frames),
AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX,
VE},
580 {
"arnr-strength",
"altref noise reduction filter strength",
OFFSET(arnr_strength),
AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX,
VE},
581 {
"arnr-type",
"altref noise reduction filter type",
OFFSET(arnr_type),
AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX,
VE,
"arnr_type"},
585 {
"deadline",
"Time to spend encoding, in microseconds.",
OFFSET(deadline),
AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX,
VE,
"quality"},
589 {
"error-resilient",
"Error resilience configuration",
OFFSET(error_resilient),
AV_OPT_TYPE_FLAGS, {.i64 = 0}, INT_MIN, INT_MAX,
VE,
"er"},
590 {
"max-intra-rate",
"Maximum I-frame bitrate (pct) 0=unlimited",
OFFSET(max_intra_rate),
AV_OPT_TYPE_INT, {.i64 = -1}, -1, INT_MAX,
VE},
591 #ifdef VPX_ERROR_RESILIENT_DEFAULT
592 {
"default",
"Improve resiliency against losses of whole frames", 0,
AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_DEFAULT}, 0, 0,
VE,
"er"},
593 { "partitions", "The frame partitions are independently decodable "
594 "by the bool decoder, meaning that partitions can be decoded even "
595 "though earlier partitions have been lost. Note that intra predicition"
596 " is still done over the partition boundary.", 0,
AV_OPT_TYPE_CONST, {.i64 = VPX_ERROR_RESILIENT_PARTITIONS}, 0, 0,
VE,
"er"},
597 #endif
599 {
"quality",
"", offsetof(
VP8Context, deadline),
AV_OPT_TYPE_INT, {.i64 = VPX_DL_GOOD_QUALITY}, INT_MIN, INT_MAX,
VE,
"quality"},
600 {
"vp8flags",
"", offsetof(
VP8Context,
flags),
FF_OPT_TYPE_FLAGS, {.i64 = 0}, 0, UINT_MAX,
VE,
"flags"},
603 {
"arnr_max_frames",
"altref noise reduction max frame count", offsetof(
VP8Context, arnr_max_frames),
AV_OPT_TYPE_INT, {.i64 = 0}, 0, 15,
VE},
604 {
"arnr_strength",
"altref noise reduction filter strength", offsetof(
VP8Context, arnr_strength),
AV_OPT_TYPE_INT, {.i64 = 3}, 0, 6,
VE},
606 {
"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},
609 };
610
616 };
617
619 { "qmin", "-1" },
620 { "qmax", "-1" },
621 { "g", "-1" },
622 { "keyint_min", "-1" },
624 };
625
637 .priv_class = &class,
639 };