lavc: use a separate field for exporting audio encoder padding

Currently, the amount of padding inserted at the beginning by some audio encoders, is exported through AVCodecContext.delay. However - the term 'delay' is heavily overloaded and can have multiple different meanings even in the case of audio encoding. - this field has entirely different meanings, depending on whether the codec context is used for encoding or decoding (and has yet another different meaning for video), preventing generic handling of the codec context. Therefore, add a new field -- AVCodecContext.initial_padding. It could conceivably be used for decoding as well at a later point.

Anton Khirnov committed 12 years ago

Commit 2df0c321 parent c80a81622 files · +63 −40Browse files

Unified · Split

22 files

doc/APIchanges +4 −0
libavcodec/aacenc.c +1 −1
libavcodec/ac3enc.c +1 −1
libavcodec/audio_frame_queue.c +2 −2
libavcodec/avcodec.h +18 −10
libavcodec/g722enc.c +2 −2
libavcodec/libfaac.c +1 −1
libavcodec/libfdk-aacenc.c +1 −1
libavcodec/libmp3lame.c +1 −1
libavcodec/libopencore-amr.c +2 −2
libavcodec/libopusenc.c +2 −2
libavcodec/libspeexenc.c +1 −1
libavcodec/libtwolame.c +2 −2
libavcodec/libvo-aacenc.c +1 −1
libavcodec/libvo-amrwbenc.c +2 −2
libavcodec/libvorbis.c +2 −2
libavcodec/mpegaudioenc.c +2 −2
libavcodec/nellymoserenc.c +1 −1
libavcodec/ra144enc.c +1 −1
libavcodec/utils.c +9 −0
libavcodec/version.h +5 −2
libavcodec/wmaenc.c +2 −3

doc/APIchanges+4−0

@@ -13,6 +13,10 @@ libavutil: 2014-08-09 · 12 unmodified lines
13	13
14	14	API changes, most recent first:
15	15
	16	+2014-10-13 - xxxxxxx - lavc 55.03.0 - avcodec.h
	17	+ Add AVCodecContext.initial_padding. Deprecate the use of AVCodecContext.delay
	18	+ for audio encoding.
	19	+
16	20	2014-09-xx - xxxxxxx - lavu 54.04.0 - pixdesc.h
17	21	Add API to return the name of frame and context color properties.
18	22

libavcodec/aacenc.c+1−1

@@ -777,7 +777,7 @@ static av_cold int aac_encode_init(AVCodecContext *avctx) · 776 unmodified lines
777	777	for (i = 0; i < 428; i++)
778	778	ff_aac_pow34sf_tab[i] = sqrt(ff_aac_pow2sf_tab[i] * sqrt(ff_aac_pow2sf_tab[i]));
779	779
780		− avctx->delay = 1024;
	780	+ avctx->initial_padding = 1024;
781	781	ff_af_queue_init(avctx, &s->afq);
782	782
783	783	return 0;

libavcodec/ac3enc.c+1−1

@@ -2436,7 +2436,7 @@ av_cold int ff_ac3_encode_init(AVCodecContext *avctx) · 2435 unmodified lines
2436	2436	return ret;
2437	2437
2438	2438	avctx->frame_size = AC3_BLOCK_SIZE * s->num_blocks;
2439		− avctx->delay = AC3_BLOCK_SIZE;
	2439	+ avctx->initial_padding = AC3_BLOCK_SIZE;
2440	2440
2441	2441	s->bitstream_mode = avctx->audio_service_type;
2442	2442	if (s->bitstream_mode == AV_AUDIO_SERVICE_TYPE_KARAOKE)

libavcodec/audio_frame_queue.c+2−2

@@ -29,8 +29,8 @@ av_cold void ff_af_queue_init(AVCodecContext avctx, AudioFrameQueue afq) · 28 unmodified lines
29	29	{
30	30	afq->avctx = avctx;
31	31	afq->next_pts = AV_NOPTS_VALUE;
32		− afq->remaining_delay = avctx->delay;
33		− afq->remaining_samples = avctx->delay;
	32	+ afq->remaining_delay = avctx->initial_padding;
	33	+ afq->remaining_samples = avctx->initial_padding;
34	34	afq->frame_queue = NULL;
35	35	}
36	36

libavcodec/avcodec.h+18−10

@@ -1191,16 +1191,7 @@ typedef struct AVCodecContext { · 1190 unmodified lines
1191	1191	* encoded input.
1192	1192	*
1193	1193	* Audio:
1194		− * For encoding, this is the number of "priming" samples added by the
1195		− * encoder to the beginning of the stream. The decoded output will be
1196		− * delayed by this many samples relative to the input to the encoder (or
1197		− * more, if the decoder adds its own padding).
1198		− * The timestamps on the output packets are adjusted by the encoder so
1199		− * that they always refer to the first sample of the data actually
1200		− * contained in the packet, including any added padding.
1201		− * E.g. if the timebase is 1/samplerate and the timestamp of the first
1202		− * input sample is 0, the timestamp of the first output packet will be
1203		− * -delay.
	1194	+ * For encoding, this field is unused (see initial_padding).
1204	1195	*
1205	1196	* For decoding, this is the number of samples the decoder needs to
1206	1197	* output before the decoder's output is valid. When seeking, you should
@@ -2780,6 +2771,23 @@ typedef struct AVCodecContext { · 1573 unmodified lines
2780	2771	* use AVOptions to set this field.
2781	2772	*/
2782	2773	int side_data_only_packets;
	2774	+
	2775	+ /**
	2776	+ * Audio only. The number of "priming" samples (padding) inserted by the
	2777	+ * encoder at the beginning of the audio. I.e. this number of leading
	2778	+ * decoded samples must be discarded by the caller to get the original audio
	2779	+ * without leading padding.
	2780	+ *
	2781	+ * - decoding: unused
	2782	+ * - encoding: Set by libavcodec. The timestamps on the output packets are
	2783	+ * adjusted by the encoder so that they always refer to the
	2784	+ * first sample of the data actually contained in the packet,
	2785	+ * including any added padding. E.g. if the timebase is
	2786	+ * 1/samplerate and the timestamp of the first input sample is
	2787	+ * 0, the timestamp of the first output packet will be
	2788	+ * -initial_padding.
	2789	+ */
	2790	+ int initial_padding;
2783	2791	} AVCodecContext;
2784	2792
2785	2793	/**

libavcodec/g722enc.c+2−2

@@ -106,7 +106,7 @@ static av_cold int g722_encode_init(AVCodecContext * avctx) · 105 unmodified lines
106	106	a common packet size for VoIP applications */
107	107	avctx->frame_size = 320;
108	108	}
109		− avctx->delay = 22;
	109	+ avctx->initial_padding = 22;
110	110
111	111	if (avctx->trellis) {
112	112	/* validate trellis */
@@ -375,7 +375,7 @@ static int g722_encode_frame(AVCodecContext avctx, AVPacket avpkt, · 262 unmodified lines
375	375	}
376	376
377	377	if (frame->pts != AV_NOPTS_VALUE)
378		− avpkt->pts = frame->pts - ff_samples_to_time_base(avctx, avctx->delay);
	378	+ avpkt->pts = frame->pts - ff_samples_to_time_base(avctx, avctx->initial_padding);
379	379	*got_packet_ptr = 1;
380	380	return 0;
381	381	}

libavcodec/libfaac.c+1−1

@@ -157,7 +157,7 @@ static av_cold int Faac_encode_init(AVCodecContext *avctx) · 156 unmodified lines
157	157	goto error;
158	158	}
159	159
160		− avctx->delay = FAAC_DELAY_SAMPLES;
	160	+ avctx->initial_padding = FAAC_DELAY_SAMPLES;
161	161	ff_af_queue_init(avctx, &s->afq);
162	162
163	163	return 0;

libavcodec/libfdk-aacenc.c+1−1

@@ -286,7 +286,7 @@ static av_cold int aac_encode_init(AVCodecContext *avctx) · 285 unmodified lines
286	286	}
287	287
288	288	avctx->frame_size = info.frameLength;
289		− avctx->delay = info.encoderDelay;
	289	+ avctx->initial_padding = info.encoderDelay;
290	290	ff_af_queue_init(avctx, &s->afq);
291	291
292	292	if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER) {

libavcodec/libmp3lame.c+1−1

@@ -137,7 +137,7 @@ static av_cold int mp3lame_encode_init(AVCodecContext *avctx) · 136 unmodified lines
137	137	}
138	138
139	139	/* get encoder delay */
140		− avctx->delay = lame_get_encoder_delay(s->gfp) + 528 + 1;
	140	+ avctx->initial_padding = lame_get_encoder_delay(s->gfp) + 528 + 1;
141	141	ff_af_queue_init(avctx, &s->afq);
142	142
143	143	avctx->frame_size = lame_get_framesize(s->gfp);

libavcodec/libopencore-amr.c+2−2

@@ -200,7 +200,7 @@ static av_cold int amr_nb_encode_init(AVCodecContext *avctx) · 199 unmodified lines
200	200	}
201	201
202	202	avctx->frame_size = 160;
203		− avctx->delay = 50;
	203	+ avctx->initial_padding = 50;
204	204	ff_af_queue_init(avctx, &s->afq);
205	205
206	206	s->enc_state = Encoder_Interface_init(s->enc_dtx);
@@ -250,7 +250,7 @@ static int amr_nb_encode_frame(AVCodecContext avctx, AVPacket avpkt, · 43 unmodified lines
250	250	return AVERROR(ENOMEM);
251	251	memcpy(flush_buf, samples, frame->nb_samples * sizeof(*flush_buf));
252	252	samples = flush_buf;
253		− if (frame->nb_samples < avctx->frame_size - avctx->delay)
	253	+ if (frame->nb_samples < avctx->frame_size - avctx->initial_padding)
254	254	s->enc_last_frame = -1;
255	255	}
256	256	if ((ret = ff_af_queue_add(&s->afq, frame)) < 0) {

libavcodec/libopusenc.c+2−2

@@ -87,7 +87,7 @@ static void libopus_write_header(AVCodecContext *avctx, int stream_count, · 86 unmodified lines
87	87	bytestream_put_buffer(&p, "OpusHead", 8);
88	88	bytestream_put_byte(&p, 1); /* Version */
89	89	bytestream_put_byte(&p, channels);
90		− bytestream_put_le16(&p, avctx->delay); /* Lookahead samples at 48kHz */
	90	+ bytestream_put_le16(&p, avctx->initial_padding); /* Lookahead samples at 48kHz */
91	91	bytestream_put_le32(&p, avctx->sample_rate); /* Original sample rate */
92	92	bytestream_put_le16(&p, 0); /* Gain of 0dB is recommended. */
93	93
@@ -277,7 +277,7 @@ static int av_cold libopus_encode_init(AVCodecContext *avctx) · 183 unmodified lines
277	277	goto fail;
278	278	}
279	279
280		− ret = opus_multistream_encoder_ctl(enc, OPUS_GET_LOOKAHEAD(&avctx->delay));
	280	+ ret = opus_multistream_encoder_ctl(enc, OPUS_GET_LOOKAHEAD(&avctx->initial_padding));
281	281	if (ret != OPUS_OK)
282	282	av_log(avctx, AV_LOG_WARNING,
283	283	"Unable to get number of lookahead samples: %s\n",

libavcodec/libspeexenc.c+1−1

@@ -235,7 +235,7 @@ static av_cold int encode_init(AVCodecContext *avctx) · 234 unmodified lines
235	235	s->header.frames_per_packet = s->frames_per_packet;
236	236
237	237	/* set encoding delay */
238		− speex_encoder_ctl(s->enc_state, SPEEX_GET_LOOKAHEAD, &avctx->delay);
	238	+ speex_encoder_ctl(s->enc_state, SPEEX_GET_LOOKAHEAD, &avctx->initial_padding);
239	239	ff_af_queue_init(avctx, &s->afq);
240	240
241	241	/* create header packet bytes from header struct */

libavcodec/libtwolame.c+2−2

@@ -60,7 +60,7 @@ static av_cold int twolame_encode_init(AVCodecContext *avctx) · 59 unmodified lines
60	60	int ret;
61	61
62	62	avctx->frame_size = TWOLAME_SAMPLES_PER_FRAME;
63		− avctx->delay = 512 - 32 + 1;
	63	+ avctx->initial_padding = 512 - 32 + 1;
64	64
65	65	s->glopts = twolame_init();
66	66	if (!s->glopts)
@@ -151,7 +151,7 @@ static int twolame_encode_frame(AVCodecContext avctx, AVPacket avpkt, · 84 unmodified lines
151	151	avpkt->duration = ff_samples_to_time_base(avctx, frame->nb_samples);
152	152	if (frame) {
153	153	if (frame->pts != AV_NOPTS_VALUE)
154		− avpkt->pts = frame->pts - ff_samples_to_time_base(avctx, avctx->delay);
	154	+ avpkt->pts = frame->pts - ff_samples_to_time_base(avctx, avctx->initial_padding);
155	155	} else {
156	156	avpkt->pts = s->next_pts;
157	157	}

libavcodec/libvo-aacenc.c+1−1

@@ -61,7 +61,7 @@ static av_cold int aac_encode_init(AVCodecContext *avctx) · 60 unmodified lines
61	61	int index, ret;
62	62
63	63	avctx->frame_size = FRAME_SIZE;
64		− avctx->delay = ENC_DELAY;
	64	+ avctx->initial_padding = ENC_DELAY;
65	65	s->last_frame = 2;
66	66	ff_af_queue_init(avctx, &s->afq);
67	67

libavcodec/libvo-amrwbenc.c+2−2

@@ -93,7 +93,7 @@ static av_cold int amr_wb_encode_init(AVCodecContext *avctx) · 92 unmodified lines
93	93	s->last_bitrate = avctx->bit_rate;
94	94
95	95	avctx->frame_size = 320;
96		− avctx->delay = 80;
	96	+ avctx->initial_padding = 80;
97	97
98	98	s->state = E_IF_init();
99	99
@@ -131,7 +131,7 @@ static int amr_wb_encode_frame(AVCodecContext avctx, AVPacket avpkt, · 31 unmodified lines
131	131	}
132	132
133	133	if (frame->pts != AV_NOPTS_VALUE)
134		− avpkt->pts = frame->pts - ff_samples_to_time_base(avctx, avctx->delay);
	134	+ avpkt->pts = frame->pts - ff_samples_to_time_base(avctx, avctx->initial_padding);
135	135
136	136	avpkt->size = size;
137	137	*got_packet_ptr = 1;

libavcodec/libvorbis.c+2−2

@@ -322,8 +322,8 @@ static int libvorbis_encode_frame(AVCodecContext avctx, AVPacket avpkt, · 321 unmodified lines
322	322	if (duration > 0) {
323	323	/* we do not know encoder delay until we get the first packet from
324	324	* libvorbis, so we have to update the AudioFrameQueue counts */
325		− if (!avctx->delay) {
326		− avctx->delay = duration;
	325	+ if (!avctx->initial_padding) {
	326	+ avctx->initial_padding = duration;
327	327	s->afq.remaining_delay += duration;
328	328	s->afq.remaining_samples += duration;
329	329	}

libavcodec/mpegaudioenc.c+2−2

@@ -84,7 +84,7 @@ static av_cold int MPA_encode_init(AVCodecContext *avctx) · 83 unmodified lines
84	84	bitrate = bitrate / 1000;
85	85	s->nb_channels = channels;
86	86	avctx->frame_size = MPA_FRAME_SIZE;
87		− avctx->delay = 512 - 32 + 1;
	87	+ avctx->initial_padding = 512 - 32 + 1;
88	88
89	89	/* encoding freq */
90	90	s->lsf = 0;
@@ -735,7 +735,7 @@ static int MPA_encode_frame(AVCodecContext avctx, AVPacket avpkt, · 644 unmodified lines
735	735	encode_frame(s, bit_alloc, padding);
736	736
737	737	if (frame->pts != AV_NOPTS_VALUE)
738		− avpkt->pts = frame->pts - ff_samples_to_time_base(avctx, avctx->delay);
	738	+ avpkt->pts = frame->pts - ff_samples_to_time_base(avctx, avctx->initial_padding);
739	739
740	740	avpkt->size = put_bits_count(&s->pb) / 8;
741	741	*got_packet_ptr = 1;

libavcodec/nellymoserenc.c+1−1

@@ -165,7 +165,7 @@ static av_cold int encode_init(AVCodecContext *avctx) · 164 unmodified lines
165	165	}
166	166
167	167	avctx->frame_size = NELLY_SAMPLES;
168		− avctx->delay = NELLY_BUF_LEN;
	168	+ avctx->initial_padding = NELLY_BUF_LEN;
169	169	ff_af_queue_init(avctx, &s->afq);
170	170	s->avctx = avctx;
171	171	if ((ret = ff_mdct_init(&s->mdct_ctx, 8, 0, 32768.0)) < 0)

libavcodec/ra144enc.c+1−1

@@ -56,7 +56,7 @@ static av_cold int ra144_encode_init(AVCodecContext * avctx) · 55 unmodified lines
56	56	return -1;
57	57	}
58	58	avctx->frame_size = NBLOCKS * BLOCKSIZE;
59		− avctx->delay = avctx->frame_size;
	59	+ avctx->initial_padding = avctx->frame_size;
60	60	avctx->bit_rate = 8000;
61	61	ractx = avctx->priv_data;
62	62	ractx->lpc_coef[0] = ractx->lpc_tables[0];

libavcodec/utils.c+9−0

@@ -1240,6 +1240,11 @@ int attribute_align_arg avcodec_open2(AVCodecContext avctx, const AVCodec code · 1239 unmodified lines
1240	1240	}
1241	1241	}
1242	1242
	1243	+#if FF_API_AUDIOENC_DELAY
	1244	+ if (av_codec_is_encoder(avctx->codec))
	1245	+ avctx->delay = avctx->initial_padding;
	1246	+#endif
	1247	+
1243	1248	if (av_codec_is_decoder(avctx->codec)) {
1244	1249	/* validate channel layout from the decoder */
1245	1250	if (avctx->channel_layout) {
@@ -1447,6 +1452,10 @@ int attribute_align_arg avcodec_encode_audio2(AVCodecContext *avctx, · 201 unmodified lines
1447	1452	end:
1448	1453	av_frame_free(&padded_frame);
1449	1454
	1455	+#if FF_API_AUDIOENC_DELAY
	1456	+ avctx->delay = avctx->initial_padding;
	1457	+#endif
	1458	+
1450	1459	return ret;
1451	1460	}
1452	1461

libavcodec/version.h+5−2

@@ -29,8 +29,8 @@ · 28 unmodified lines
29	29	#include "libavutil/version.h"
30	30
31	31	#define LIBAVCODEC_VERSION_MAJOR 56
32		−#define LIBAVCODEC_VERSION_MINOR 2
33		−#define LIBAVCODEC_VERSION_MICRO 2
	32	+#define LIBAVCODEC_VERSION_MINOR 3
	33	+#define LIBAVCODEC_VERSION_MICRO 0
34	34
35	35	#define LIBAVCODEC_VERSION_INT AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \
36	36	LIBAVCODEC_VERSION_MINOR, \
@@ -153,5 +153,8 @@ · 116 unmodified lines
153	153	#ifndef FF_API_AFD
154	154	#define FF_API_AFD (LIBAVCODEC_VERSION_MAJOR < 57)
155	155	#endif
	156	+#ifndef FF_API_AUDIOENC_DELAY
	157	+#define FF_API_AUDIOENC_DELAY (LIBAVCODEC_VERSION_MAJOR < 58)
	158	+#endif
156	159
157	160	#endif /* AVCODEC_VERSION_H */

libavcodec/wmaenc.c+2−3

@@ -92,8 +92,7 @@ static av_cold int encode_init(AVCodecContext *avctx) · 91 unmodified lines
92	92	avctx->block_align = block_align;
93	93	avctx->bit_rate = avctx->block_align * 8LL * avctx->sample_rate /
94	94	s->frame_len;
95		− avctx->frame_size =
96		− avctx->delay = s->frame_len;
	95	+ avctx->frame_size = avctx->initial_padding = s->frame_len;
97	96
98	97	return 0;
99	98	}
@@ -420,7 +419,7 @@ static int encode_superframe(AVCodecContext avctx, AVPacket avpkt, · 320 unmodified lines
420	419	flush_put_bits(&s->pb);
421	420
422	421	if (frame->pts != AV_NOPTS_VALUE)
423		− avpkt->pts = frame->pts - ff_samples_to_time_base(avctx, avctx->delay);
	422	+ avpkt->pts = frame->pts - ff_samples_to_time_base(avctx, avctx->initial_padding);
424	423
425	424	avpkt->size = avctx->block_align;
426	425	*got_packet_ptr = 1;

lavc: use a separate field for exporting audio encoder padding

Keyboard shortcuts

General

Issues and pull requests

Issues, pull requests and Inbox

Repository code