libmp3lame.c 10.2 KB
Newer Older
1 2 3 4
/*
 * Interface to libmp3lame for mp3 encoding
 * Copyright (c) 2002 Lennert Buytenhek <buytenh@gnu.org>
 *
5 6 7
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
8 9
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
10
 * version 2.1 of the License, or (at your option) any later version.
11
 *
12
 * FFmpeg is distributed in the hope that it will be useful,
13
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 15
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
16
 *
17
 * You should have received a copy of the GNU Lesser General Public
18
 * License along with FFmpeg; if not, write to the Free Software
19
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20
 */
21

Michael Niedermayer's avatar
Michael Niedermayer committed
22
/**
23
 * @file
Michael Niedermayer's avatar
Michael Niedermayer committed
24 25
 * Interface to libmp3lame for mp3 encoding.
 */
26

27 28
#include <lame/lame.h>

29
#include "libavutil/channel_layout.h"
30
#include "libavutil/common.h"
31
#include "libavutil/float_dsp.h"
32
#include "libavutil/intreadwrite.h"
33 34
#include "libavutil/log.h"
#include "libavutil/opt.h"
35
#include "avcodec.h"
36
#include "audio_frame_queue.h"
37
#include "internal.h"
38
#include "mpegaudio.h"
39
#include "mpegaudiodecheader.h"
40

41
#define BUFFER_SIZE (7200 + 2 * MPA_FRAME_SIZE + MPA_FRAME_SIZE / 4+1000) // FIXME: Buffer size to small? Adding 1000 to make up for it.
42 43

typedef struct LAMEContext {
44
    AVClass *class;
45
    AVCodecContext *avctx;
46
    lame_global_flags *gfp;
47
    uint8_t *buffer;
48
    int buffer_index;
49
    int buffer_size;
50
    int reservoir;
51
    float *samples_flt[2];
52
    AudioFrameQueue afq;
53
    AVFloatDSPContext fdsp;
54
} LAMEContext;
55 56


57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76
static int realloc_buffer(LAMEContext *s)
{
    if (!s->buffer || s->buffer_size - s->buffer_index < BUFFER_SIZE) {
        uint8_t *tmp;
        int new_size = s->buffer_index + 2 * BUFFER_SIZE;

        av_dlog(s->avctx, "resizing output buffer: %d -> %d\n", s->buffer_size,
                new_size);
        tmp = av_realloc(s->buffer, new_size);
        if (!tmp) {
            av_freep(&s->buffer);
            s->buffer_size = s->buffer_index = 0;
            return AVERROR(ENOMEM);
        }
        s->buffer      = tmp;
        s->buffer_size = new_size;
    }
    return 0;
}

77
static av_cold int mp3lame_encode_close(AVCodecContext *avctx)
78
{
79
    LAMEContext *s = avctx->priv_data;
80

81
#if FF_API_OLD_ENCODE_AUDIO
82
    av_freep(&avctx->coded_frame);
83
#endif
84 85
    av_freep(&s->samples_flt[0]);
    av_freep(&s->samples_flt[1]);
86
    av_freep(&s->buffer);
87

88 89
    ff_af_queue_close(&s->afq);

90 91 92 93
    lame_close(s->gfp);
    return 0;
}

94
static av_cold int mp3lame_encode_init(AVCodecContext *avctx)
95
{
96
    LAMEContext *s = avctx->priv_data;
97
    int ret;
98

99 100
    s->avctx = avctx;

101
    /* initialize LAME and get defaults */
102
    if ((s->gfp = lame_init()) == NULL)
103
        return AVERROR(ENOMEM);
104

105

106
    lame_set_num_channels(s->gfp, avctx->channels);
107
    lame_set_mode(s->gfp, avctx->channels > 1 ? JOINT_STEREO : MONO);
108

109 110
    /* sample rate */
    lame_set_in_samplerate (s->gfp, avctx->sample_rate);
111
    lame_set_out_samplerate(s->gfp, avctx->sample_rate);
112 113 114

    /* algorithmic quality */
    if (avctx->compression_level == FF_COMPRESSION_DEFAULT)
115
        lame_set_quality(s->gfp, 5);
116
    else
117
        lame_set_quality(s->gfp, avctx->compression_level);
118 119

    /* rate control */
120
    if (avctx->flags & CODEC_FLAG_QSCALE) {
121
        lame_set_VBR(s->gfp, vbr_default);
122
        lame_set_VBR_quality(s->gfp, avctx->global_quality / (float)FF_QP2LAMBDA);
123 124 125
    } else {
        if (avctx->bit_rate)
            lame_set_brate(s->gfp, avctx->bit_rate / 1000);
126
    }
127

128
    /* do not get a Xing VBR header frame from LAME */
129
    lame_set_bWriteVbrTag(s->gfp,0);
130

131
    /* bit reservoir usage */
132
    lame_set_disable_reservoir(s->gfp, !s->reservoir);
133

134
    /* set specified parameters */
135 136 137
    if (lame_init_params(s->gfp) < 0) {
        ret = -1;
        goto error;
138
    }
139

140 141 142 143
    /* get encoder delay */
    avctx->delay = lame_get_encoder_delay(s->gfp) + 528 + 1;
    ff_af_queue_init(avctx, &s->afq);

144
    avctx->frame_size  = lame_get_framesize(s->gfp);
145 146

#if FF_API_OLD_ENCODE_AUDIO
147
    avctx->coded_frame = avcodec_alloc_frame();
148 149 150 151
    if (!avctx->coded_frame) {
        ret = AVERROR(ENOMEM);
        goto error;
    }
152
#endif
153

154 155
    /* allocate float sample buffers */
    if (avctx->sample_fmt == AV_SAMPLE_FMT_FLTP) {
156 157
        int ch;
        for (ch = 0; ch < avctx->channels; ch++) {
158 159 160
            s->samples_flt[ch] = av_malloc(avctx->frame_size *
                                           sizeof(*s->samples_flt[ch]));
            if (!s->samples_flt[ch]) {
161 162 163
                ret = AVERROR(ENOMEM);
                goto error;
            }
164 165 166
        }
    }

167 168 169 170
    ret = realloc_buffer(s);
    if (ret < 0)
        goto error;

171
    avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
172

173
    return 0;
174
error:
175
    mp3lame_encode_close(avctx);
176
    return ret;
177 178
}

179 180 181 182 183
#define ENCODE_BUFFER(func, buf_type, buf_name) do {                        \
    lame_result = func(s->gfp,                                              \
                       (const buf_type *)buf_name[0],                       \
                       (const buf_type *)buf_name[1], frame->nb_samples,    \
                       s->buffer + s->buffer_index,                         \
184
                       s->buffer_size - s->buffer_index);                   \
185 186
} while (0)

187 188
static int mp3lame_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
                                const AVFrame *frame, int *got_packet_ptr)
189
{
190
    LAMEContext *s = avctx->priv_data;
191
    MPADecodeHeader hdr;
192
    int len, ret, ch;
193
    int lame_result;
194

195
    if (frame) {
196
        switch (avctx->sample_fmt) {
197 198
        case AV_SAMPLE_FMT_S16P:
            ENCODE_BUFFER(lame_encode_buffer, int16_t, frame->data);
199
            break;
200 201
        case AV_SAMPLE_FMT_S32P:
            ENCODE_BUFFER(lame_encode_buffer_int, int32_t, frame->data);
202
            break;
203 204 205 206 207 208
        case AV_SAMPLE_FMT_FLTP:
            if (frame->linesize[0] < 4 * FFALIGN(frame->nb_samples, 8)) {
                av_log(avctx, AV_LOG_ERROR, "inadequate AVFrame plane padding\n");
                return AVERROR(EINVAL);
            }
            for (ch = 0; ch < avctx->channels; ch++) {
209 210 211 212
                s->fdsp.vector_fmul_scalar(s->samples_flt[ch],
                                           (const float *)frame->data[ch],
                                           32768.0f,
                                           FFALIGN(frame->nb_samples, 8));
213 214
            }
            ENCODE_BUFFER(lame_encode_buffer_float, float, s->samples_flt);
215 216 217
            break;
        default:
            return AVERROR_BUG;
218
        }
219 220 221
    } else {
        lame_result = lame_encode_flush(s->gfp, s->buffer + s->buffer_index,
                                        BUFFER_SIZE - s->buffer_index);
222
    }
223 224 225 226
    if (lame_result < 0) {
        if (lame_result == -1) {
            av_log(avctx, AV_LOG_ERROR,
                   "lame: output buffer too small (buffer index: %d, free bytes: %d)\n",
227
                   s->buffer_index, s->buffer_size - s->buffer_index);
Michael Niedermayer's avatar
Michael Niedermayer committed
228
        }
229
        return -1;
230 231
    }
    s->buffer_index += lame_result;
232 233 234 235 236
    ret = realloc_buffer(s);
    if (ret < 0) {
        av_log(avctx, AV_LOG_ERROR, "error reallocating output buffer\n");
        return ret;
    }
237

238 239
    /* add current frame to the queue */
    if (frame) {
240
        if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
241 242 243
            return ret;
    }

244 245 246
    /* Move 1 frame from the LAME buffer to the output packet, if available.
       We have to parse the first frame header in the output buffer to
       determine the frame size. */
247
    if (s->buffer_index < 4)
248
        return 0;
249 250 251 252 253
    if (avpriv_mpegaudio_decode_header(&hdr, AV_RB32(s->buffer))) {
        av_log(avctx, AV_LOG_ERROR, "free format output not supported\n");
        return -1;
    }
    len = hdr.frame_size;
254 255
    av_dlog(avctx, "in:%d packet-len:%d index:%d\n", avctx->frame_size, len,
            s->buffer_index);
256
    if (len <= s->buffer_index) {
257
        if ((ret = ff_alloc_packet2(avctx, avpkt, len)))
258 259
            return ret;
        memcpy(avpkt->data, s->buffer, len);
Michael Niedermayer's avatar
Michael Niedermayer committed
260
        s->buffer_index -= len;
261
        memmove(s->buffer, s->buffer + len, s->buffer_index);
262 263 264 265 266 267 268 269 270

        /* Get the next frame pts/duration */
        ff_af_queue_remove(&s->afq, avctx->frame_size, &avpkt->pts,
                           &avpkt->duration);

        avpkt->size = len;
        *got_packet_ptr = 1;
    }
    return 0;
271 272
}

273
#define OFFSET(x) offsetof(LAMEContext, x)
274 275
#define AE AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
static const AVOption options[] = {
276
    { "reservoir", "Use bit reservoir.", OFFSET(reservoir), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, AE },
277 278 279 280 281 282 283 284 285
    { NULL },
};

static const AVClass libmp3lame_class = {
    .class_name = "libmp3lame encoder",
    .item_name  = av_default_item_name,
    .option     = options,
    .version    = LIBAVUTIL_VERSION_INT,
};
286

287 288 289 290 291
static const AVCodecDefault libmp3lame_defaults[] = {
    { "b",          "0" },
    { NULL },
};

292 293 294 295
static const int libmp3lame_sample_rates[] = {
    44100, 48000,  32000, 22050, 24000, 16000, 11025, 12000, 8000, 0
};

296
AVCodec ff_libmp3lame_encoder = {
297 298
    .name                  = "libmp3lame",
    .type                  = AVMEDIA_TYPE_AUDIO,
299
    .id                    = AV_CODEC_ID_MP3,
300 301
    .priv_data_size        = sizeof(LAMEContext),
    .init                  = mp3lame_encode_init,
302
    .encode2               = mp3lame_encode_frame,
303
    .close                 = mp3lame_encode_close,
304
    .capabilities          = CODEC_CAP_DELAY | CODEC_CAP_SMALL_LAST_FRAME,
305 306 307
    .sample_fmts           = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S32P,
                                                             AV_SAMPLE_FMT_FLTP,
                                                             AV_SAMPLE_FMT_S16P,
308
                                                             AV_SAMPLE_FMT_NONE },
309
    .supported_samplerates = libmp3lame_sample_rates,
310
    .channel_layouts       = (const uint64_t[]) { AV_CH_LAYOUT_MONO,
311
                                                  AV_CH_LAYOUT_STEREO,
312
                                                  0 },
313 314
    .long_name             = NULL_IF_CONFIG_SMALL("libmp3lame MP3 (MPEG audio layer 3)"),
    .priv_class            = &libmp3lame_class,
315
    .defaults              = libmp3lame_defaults,
316
};