libmp3lame: allow joint stereo to be disabled
[libav.git] / libavcodec / libmp3lame.c
CommitLineData
0716b577
J
1/*
2 * Interface to libmp3lame for mp3 encoding
3 * Copyright (c) 2002 Lennert Buytenhek <buytenh@gnu.org>
4 *
2912e87a 5 * This file is part of Libav.
b78e7197 6 *
2912e87a 7 * Libav is free software; you can redistribute it and/or
ff4ec49e
FB
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
b78e7197 10 * version 2.1 of the License, or (at your option) any later version.
0716b577 11 *
2912e87a 12 * Libav is distributed in the hope that it will be useful,
0716b577 13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
ff4ec49e
FB
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
0716b577 16 *
ff4ec49e 17 * You should have received a copy of the GNU Lesser General Public
2912e87a 18 * License along with Libav; if not, write to the Free Software
5509bffa 19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
0716b577 20 */
115329f1 21
983e3246 22/**
ba87f080 23 * @file
983e3246
MN
24 * Interface to libmp3lame for mp3 encoding.
25 */
0716b577 26
e2322252
JR
27#include <lame/lame.h>
28
a903f8f0 29#include "libavutil/channel_layout.h"
1d9c2dc8 30#include "libavutil/common.h"
284ea790 31#include "libavutil/float_dsp.h"
03c804e1 32#include "libavutil/intreadwrite.h"
a7cec3a0
AK
33#include "libavutil/log.h"
34#include "libavutil/opt.h"
0716b577 35#include "avcodec.h"
3d853d7a 36#include "audio_frame_queue.h"
232e16dd 37#include "internal.h"
0716b577 38#include "mpegaudio.h"
1f516c04 39#include "mpegaudiodecheader.h"
0716b577 40
c4db3446 41#define BUFFER_SIZE (7200 + 2 * MPA_FRAME_SIZE + MPA_FRAME_SIZE / 4)
e2322252
JR
42
43typedef struct LAMEContext {
a7cec3a0 44 AVClass *class;
e00959a9 45 AVCodecContext *avctx;
e344c1ea 46 lame_global_flags *gfp;
abd8b9e7 47 uint8_t *buffer;
e344c1ea 48 int buffer_index;
abd8b9e7 49 int buffer_size;
a7cec3a0 50 int reservoir;
729d821f 51 int joint_stereo;
473b297f 52 float *samples_flt[2];
3d853d7a 53 AudioFrameQueue afq;
284ea790 54 AVFloatDSPContext fdsp;
e2322252 55} LAMEContext;
0716b577 56
8dad25eb 57
abd8b9e7
JR
58static int realloc_buffer(LAMEContext *s)
59{
60 if (!s->buffer || s->buffer_size - s->buffer_index < BUFFER_SIZE) {
9b8d11a7 61 int new_size = s->buffer_index + 2 * BUFFER_SIZE, err;
abd8b9e7
JR
62
63 av_dlog(s->avctx, "resizing output buffer: %d -> %d\n", s->buffer_size,
64 new_size);
9b8d11a7 65 if ((err = av_reallocp(&s->buffer, new_size)) < 0) {
abd8b9e7 66 s->buffer_size = s->buffer_index = 0;
9b8d11a7 67 return err;
abd8b9e7 68 }
abd8b9e7
JR
69 s->buffer_size = new_size;
70 }
71 return 0;
72}
73
e2322252 74static av_cold int mp3lame_encode_close(AVCodecContext *avctx)
8dad25eb 75{
e2322252 76 LAMEContext *s = avctx->priv_data;
8dad25eb 77
473b297f
JR
78 av_freep(&s->samples_flt[0]);
79 av_freep(&s->samples_flt[1]);
abd8b9e7 80 av_freep(&s->buffer);
8dad25eb 81
3d853d7a
JR
82 ff_af_queue_close(&s->afq);
83
8dad25eb
JR
84 lame_close(s->gfp);
85 return 0;
86}
87
e2322252 88static av_cold int mp3lame_encode_init(AVCodecContext *avctx)
0716b577 89{
e2322252 90 LAMEContext *s = avctx->priv_data;
8dad25eb 91 int ret;
e344c1ea 92
e00959a9
JR
93 s->avctx = avctx;
94
e2322252 95 /* initialize LAME and get defaults */
e344c1ea 96 if ((s->gfp = lame_init()) == NULL)
8dad25eb 97 return AVERROR(ENOMEM);
e2322252 98
e344c1ea 99 lame_set_num_channels(s->gfp, avctx->channels);
729d821f 100 lame_set_mode(s->gfp, avctx->channels > 1 ? s->joint_stereo ? JOINT_STEREO : STEREO : MONO);
e2322252
JR
101
102 /* sample rate */
103 lame_set_in_samplerate (s->gfp, avctx->sample_rate);
104 lame_set_out_samplerate(s->gfp, avctx->sample_rate);
105
106 /* algorithmic quality */
107 if (avctx->compression_level == FF_COMPRESSION_DEFAULT)
afffd357 108 lame_set_quality(s->gfp, 5);
e2322252 109 else
afffd357 110 lame_set_quality(s->gfp, avctx->compression_level);
e2322252
JR
111
112 /* rate control */
c4db3446 113 if (avctx->flags & CODEC_FLAG_QSCALE) {
c57c770d 114 lame_set_VBR(s->gfp, vbr_default);
c4db3446 115 lame_set_VBR_quality(s->gfp, avctx->global_quality / (float)FF_QP2LAMBDA);
232e16dd
JR
116 } else {
117 if (avctx->bit_rate)
118 lame_set_brate(s->gfp, avctx->bit_rate / 1000);
c57c770d 119 }
e2322252
JR
120
121 /* do not get a Xing VBR header frame from LAME */
e344c1ea 122 lame_set_bWriteVbrTag(s->gfp,0);
e2322252
JR
123
124 /* bit reservoir usage */
a7cec3a0 125 lame_set_disable_reservoir(s->gfp, !s->reservoir);
e2322252
JR
126
127 /* set specified parameters */
8dad25eb
JR
128 if (lame_init_params(s->gfp) < 0) {
129 ret = -1;
130 goto error;
131 }
0716b577 132
3d853d7a
JR
133 /* get encoder delay */
134 avctx->delay = lame_get_encoder_delay(s->gfp) + 528 + 1;
135 ff_af_queue_init(avctx, &s->afq);
136
e2322252 137 avctx->frame_size = lame_get_framesize(s->gfp);
3d853d7a 138
473b297f
JR
139 /* allocate float sample buffers */
140 if (avctx->sample_fmt == AV_SAMPLE_FMT_FLTP) {
e00959a9
JR
141 int ch;
142 for (ch = 0; ch < avctx->channels; ch++) {
473b297f
JR
143 s->samples_flt[ch] = av_malloc(avctx->frame_size *
144 sizeof(*s->samples_flt[ch]));
145 if (!s->samples_flt[ch]) {
e00959a9
JR
146 ret = AVERROR(ENOMEM);
147 goto error;
148 }
149 }
150 }
151
abd8b9e7
JR
152 ret = realloc_buffer(s);
153 if (ret < 0)
154 goto error;
155
284ea790 156 avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
473b297f 157
e344c1ea 158 return 0;
8dad25eb 159error:
e2322252 160 mp3lame_encode_close(avctx);
8dad25eb 161 return ret;
0716b577
J
162}
163
473b297f
JR
164#define ENCODE_BUFFER(func, buf_type, buf_name) do { \
165 lame_result = func(s->gfp, \
166 (const buf_type *)buf_name[0], \
167 (const buf_type *)buf_name[1], frame->nb_samples, \
168 s->buffer + s->buffer_index, \
abd8b9e7 169 s->buffer_size - s->buffer_index); \
e00959a9
JR
170} while (0)
171
3d853d7a
JR
172static int mp3lame_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
173 const AVFrame *frame, int *got_packet_ptr)
0716b577 174{
e2322252 175 LAMEContext *s = avctx->priv_data;
1f516c04 176 MPADecodeHeader hdr;
473b297f 177 int len, ret, ch;
e344c1ea 178 int lame_result;
0716b577 179
3d853d7a 180 if (frame) {
e00959a9 181 switch (avctx->sample_fmt) {
473b297f
JR
182 case AV_SAMPLE_FMT_S16P:
183 ENCODE_BUFFER(lame_encode_buffer, int16_t, frame->data);
e00959a9 184 break;
473b297f
JR
185 case AV_SAMPLE_FMT_S32P:
186 ENCODE_BUFFER(lame_encode_buffer_int, int32_t, frame->data);
e00959a9 187 break;
473b297f
JR
188 case AV_SAMPLE_FMT_FLTP:
189 if (frame->linesize[0] < 4 * FFALIGN(frame->nb_samples, 8)) {
190 av_log(avctx, AV_LOG_ERROR, "inadequate AVFrame plane padding\n");
191 return AVERROR(EINVAL);
192 }
193 for (ch = 0; ch < avctx->channels; ch++) {
284ea790
JR
194 s->fdsp.vector_fmul_scalar(s->samples_flt[ch],
195 (const float *)frame->data[ch],
196 32768.0f,
197 FFALIGN(frame->nb_samples, 8));
473b297f
JR
198 }
199 ENCODE_BUFFER(lame_encode_buffer_float, float, s->samples_flt);
e00959a9
JR
200 break;
201 default:
202 return AVERROR_BUG;
6f824977 203 }
c4db3446
AD
204 } else {
205 lame_result = lame_encode_flush(s->gfp, s->buffer + s->buffer_index,
e984f478 206 s->buffer_size - s->buffer_index);
0f5c3f21 207 }
c4db3446
AD
208 if (lame_result < 0) {
209 if (lame_result == -1) {
c4db3446
AD
210 av_log(avctx, AV_LOG_ERROR,
211 "lame: output buffer too small (buffer index: %d, free bytes: %d)\n",
abd8b9e7 212 s->buffer_index, s->buffer_size - s->buffer_index);
203fa6b4 213 }
20836483 214 return -1;
0f5c3f21 215 }
0f5c3f21 216 s->buffer_index += lame_result;
abd8b9e7
JR
217 ret = realloc_buffer(s);
218 if (ret < 0) {
219 av_log(avctx, AV_LOG_ERROR, "error reallocating output buffer\n");
220 return ret;
221 }
0f5c3f21 222
3d853d7a
JR
223 /* add current frame to the queue */
224 if (frame) {
1d7ffd06 225 if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
3d853d7a
JR
226 return ret;
227 }
228
e2322252
JR
229 /* Move 1 frame from the LAME buffer to the output packet, if available.
230 We have to parse the first frame header in the output buffer to
231 determine the frame size. */
c4db3446 232 if (s->buffer_index < 4)
0f5c3f21 233 return 0;
1f516c04
JR
234 if (avpriv_mpegaudio_decode_header(&hdr, AV_RB32(s->buffer))) {
235 av_log(avctx, AV_LOG_ERROR, "free format output not supported\n");
236 return -1;
237 }
238 len = hdr.frame_size;
bf909fc4
JR
239 av_dlog(avctx, "in:%d packet-len:%d index:%d\n", avctx->frame_size, len,
240 s->buffer_index);
c4db3446 241 if (len <= s->buffer_index) {
3d853d7a
JR
242 if ((ret = ff_alloc_packet(avpkt, len))) {
243 av_log(avctx, AV_LOG_ERROR, "Error getting output packet\n");
244 return ret;
245 }
246 memcpy(avpkt->data, s->buffer, len);
203fa6b4 247 s->buffer_index -= len;
c4db3446 248 memmove(s->buffer, s->buffer + len, s->buffer_index);
3d853d7a
JR
249
250 /* Get the next frame pts/duration */
251 ff_af_queue_remove(&s->afq, avctx->frame_size, &avpkt->pts,
252 &avpkt->duration);
253
254 avpkt->size = len;
255 *got_packet_ptr = 1;
256 }
257 return 0;
0716b577
J
258}
259
e2322252 260#define OFFSET(x) offsetof(LAMEContext, x)
a7cec3a0
AK
261#define AE AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
262static const AVOption options[] = {
e6153f17 263 { "reservoir", "Use bit reservoir.", OFFSET(reservoir), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, AE },
729d821f 264 { "joint_stereo", "Use joint stereo.", OFFSET(joint_stereo), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, AE },
a7cec3a0
AK
265 { NULL },
266};
267
268static const AVClass libmp3lame_class = {
269 .class_name = "libmp3lame encoder",
270 .item_name = av_default_item_name,
271 .option = options,
272 .version = LIBAVUTIL_VERSION_INT,
273};
0716b577 274
232e16dd
JR
275static const AVCodecDefault libmp3lame_defaults[] = {
276 { "b", "0" },
277 { NULL },
278};
279
e2322252
JR
280static const int libmp3lame_sample_rates[] = {
281 44100, 48000, 32000, 22050, 24000, 16000, 11025, 12000, 8000, 0
282};
283
d36beb3f 284AVCodec ff_libmp3lame_encoder = {
c4db3446 285 .name = "libmp3lame",
b2bed932 286 .long_name = NULL_IF_CONFIG_SMALL("libmp3lame MP3 (MPEG audio layer 3)"),
c4db3446 287 .type = AVMEDIA_TYPE_AUDIO,
36ef5369 288 .id = AV_CODEC_ID_MP3,
e2322252
JR
289 .priv_data_size = sizeof(LAMEContext),
290 .init = mp3lame_encode_init,
3d853d7a 291 .encode2 = mp3lame_encode_frame,
e2322252 292 .close = mp3lame_encode_close,
3d853d7a 293 .capabilities = CODEC_CAP_DELAY | CODEC_CAP_SMALL_LAST_FRAME,
473b297f
JR
294 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S32P,
295 AV_SAMPLE_FMT_FLTP,
296 AV_SAMPLE_FMT_S16P,
c4db3446 297 AV_SAMPLE_FMT_NONE },
e2322252 298 .supported_samplerates = libmp3lame_sample_rates,
e6694659 299 .channel_layouts = (const uint64_t[]) { AV_CH_LAYOUT_MONO,
bcbb30e2
MN
300 AV_CH_LAYOUT_STEREO,
301 0 },
c4db3446 302 .priv_class = &libmp3lame_class,
232e16dd 303 .defaults = libmp3lame_defaults,
0716b577 304};