libmp3lame.c 10.4 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
    int joint_stereo;
52
    int abr;
53
    float *samples_flt[2];
54
    AudioFrameQueue afq;
55
    AVFloatDSPContext fdsp;
56
} LAMEContext;
57 58


59 60 61
static int realloc_buffer(LAMEContext *s)
{
    if (!s->buffer || s->buffer_size - s->buffer_index < BUFFER_SIZE) {
62
        int new_size = s->buffer_index + 2 * BUFFER_SIZE, err;
63 64 65

        av_dlog(s->avctx, "resizing output buffer: %d -> %d\n", s->buffer_size,
                new_size);
66
        if ((err = av_reallocp(&s->buffer, new_size)) < 0) {
67
            s->buffer_size = s->buffer_index = 0;
68
            return err;
69 70 71 72 73 74
        }
        s->buffer_size = new_size;
    }
    return 0;
}

75
static av_cold int mp3lame_encode_close(AVCodecContext *avctx)
76
{
77
    LAMEContext *s = avctx->priv_data;
78

79 80
    av_freep(&s->samples_flt[0]);
    av_freep(&s->samples_flt[1]);
81
    av_freep(&s->buffer);
82

83 84
    ff_af_queue_close(&s->afq);

85 86 87 88
    lame_close(s->gfp);
    return 0;
}

89
static av_cold int mp3lame_encode_init(AVCodecContext *avctx)
90
{
91
    LAMEContext *s = avctx->priv_data;
92
    int ret;
93

94 95
    s->avctx = avctx;

96
    /* initialize LAME and get defaults */
97
    if ((s->gfp = lame_init()) == NULL)
98
        return AVERROR(ENOMEM);
99

100

101
    lame_set_num_channels(s->gfp, avctx->channels);
102
    lame_set_mode(s->gfp, avctx->channels > 1 ? s->joint_stereo ? JOINT_STEREO : STEREO : MONO);
103

104 105
    /* sample rate */
    lame_set_in_samplerate (s->gfp, avctx->sample_rate);
106
    lame_set_out_samplerate(s->gfp, avctx->sample_rate);
107 108 109

    /* algorithmic quality */
    if (avctx->compression_level == FF_COMPRESSION_DEFAULT)
110
        lame_set_quality(s->gfp, 5);
111
    else
112
        lame_set_quality(s->gfp, avctx->compression_level);
113 114

    /* rate control */
115
    if (avctx->flags & CODEC_FLAG_QSCALE) { // VBR
116
        lame_set_VBR(s->gfp, vbr_default);
117
        lame_set_VBR_quality(s->gfp, avctx->global_quality / (float)FF_QP2LAMBDA);
118
    } else {
119 120 121 122 123 124 125
        if (avctx->bit_rate) {
            if (s->abr) {                   // ABR
                lame_set_VBR(s->gfp, vbr_abr);
                lame_set_VBR_mean_bitrate_kbps(s->gfp, avctx->bit_rate / 1000);
            } else                          // CBR
                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

146 147
    /* allocate float sample buffers */
    if (avctx->sample_fmt == AV_SAMPLE_FMT_FLTP) {
148 149
        int ch;
        for (ch = 0; ch < avctx->channels; ch++) {
150 151 152
            s->samples_flt[ch] = av_malloc(avctx->frame_size *
                                           sizeof(*s->samples_flt[ch]));
            if (!s->samples_flt[ch]) {
153 154 155
                ret = AVERROR(ENOMEM);
                goto error;
            }
156 157 158
        }
    }

159 160 161 162
    ret = realloc_buffer(s);
    if (ret < 0)
        goto error;

163
    avpriv_float_dsp_init(&s->fdsp, avctx->flags & CODEC_FLAG_BITEXACT);
164

165
    return 0;
166
error:
167
    mp3lame_encode_close(avctx);
168
    return ret;
169 170
}

171 172 173 174 175
#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,                         \
176
                       s->buffer_size - s->buffer_index);                   \
177 178
} while (0)

179 180
static int mp3lame_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
                                const AVFrame *frame, int *got_packet_ptr)
181
{
182
    LAMEContext *s = avctx->priv_data;
183
    MPADecodeHeader hdr;
184
    int len, ret, ch;
185
    int lame_result;
186

187
    if (frame) {
188
        switch (avctx->sample_fmt) {
189 190
        case AV_SAMPLE_FMT_S16P:
            ENCODE_BUFFER(lame_encode_buffer, int16_t, frame->data);
191
            break;
192 193
        case AV_SAMPLE_FMT_S32P:
            ENCODE_BUFFER(lame_encode_buffer_int, int32_t, frame->data);
194
            break;
195 196 197 198 199 200
        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++) {
201 202 203 204
                s->fdsp.vector_fmul_scalar(s->samples_flt[ch],
                                           (const float *)frame->data[ch],
                                           32768.0f,
                                           FFALIGN(frame->nb_samples, 8));
205 206
            }
            ENCODE_BUFFER(lame_encode_buffer_float, float, s->samples_flt);
207 208 209
            break;
        default:
            return AVERROR_BUG;
210
        }
211 212
    } else {
        lame_result = lame_encode_flush(s->gfp, s->buffer + s->buffer_index,
213
                                        s->buffer_size - s->buffer_index);
214
    }
215 216 217 218
    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",
219
                   s->buffer_index, s->buffer_size - s->buffer_index);
Michael Niedermayer's avatar
Michael Niedermayer committed
220
        }
221
        return -1;
222 223
    }
    s->buffer_index += lame_result;
224 225 226 227 228
    ret = realloc_buffer(s);
    if (ret < 0) {
        av_log(avctx, AV_LOG_ERROR, "error reallocating output buffer\n");
        return ret;
    }
229

230 231
    /* add current frame to the queue */
    if (frame) {
232
        if ((ret = ff_af_queue_add(&s->afq, frame)) < 0)
233 234 235
            return ret;
    }

236 237 238
    /* 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. */
239
    if (s->buffer_index < 4)
240
        return 0;
241 242 243 244 245
    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;
246 247
    av_dlog(avctx, "in:%d packet-len:%d index:%d\n", avctx->frame_size, len,
            s->buffer_index);
248
    if (len <= s->buffer_index) {
249
        if ((ret = ff_alloc_packet2(avctx, avpkt, len)) < 0)
250 251
            return ret;
        memcpy(avpkt->data, s->buffer, len);
Michael Niedermayer's avatar
Michael Niedermayer committed
252
        s->buffer_index -= len;
253
        memmove(s->buffer, s->buffer + len, s->buffer_index);
254 255 256 257 258 259 260 261 262

        /* 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;
263 264
}

265
#define OFFSET(x) offsetof(LAMEContext, x)
266 267
#define AE AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
static const AVOption options[] = {
268 269
    { "reservoir",    "use bit reservoir", OFFSET(reservoir),    AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, AE },
    { "joint_stereo", "use joint stereo",  OFFSET(joint_stereo), AV_OPT_TYPE_INT, { .i64 = 1 }, 0, 1, AE },
270
    { "abr",          "use ABR",           OFFSET(abr),          AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, AE },
271 272 273 274 275 276 277 278 279
    { NULL },
};

static const AVClass libmp3lame_class = {
    .class_name = "libmp3lame encoder",
    .item_name  = av_default_item_name,
    .option     = options,
    .version    = LIBAVUTIL_VERSION_INT,
};
280

281 282 283 284 285
static const AVCodecDefault libmp3lame_defaults[] = {
    { "b",          "0" },
    { NULL },
};

286 287 288 289
static const int libmp3lame_sample_rates[] = {
    44100, 48000,  32000, 22050, 24000, 16000, 11025, 12000, 8000, 0
};

290
AVCodec ff_libmp3lame_encoder = {
291
    .name                  = "libmp3lame",
292
    .long_name             = NULL_IF_CONFIG_SMALL("libmp3lame MP3 (MPEG audio layer 3)"),
293
    .type                  = AVMEDIA_TYPE_AUDIO,
294
    .id                    = AV_CODEC_ID_MP3,
295 296
    .priv_data_size        = sizeof(LAMEContext),
    .init                  = mp3lame_encode_init,
297
    .encode2               = mp3lame_encode_frame,
298
    .close                 = mp3lame_encode_close,
299
    .capabilities          = CODEC_CAP_DELAY | CODEC_CAP_SMALL_LAST_FRAME,
300 301 302
    .sample_fmts           = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S32P,
                                                             AV_SAMPLE_FMT_FLTP,
                                                             AV_SAMPLE_FMT_S16P,
303
                                                             AV_SAMPLE_FMT_NONE },
304
    .supported_samplerates = libmp3lame_sample_rates,
305
    .channel_layouts       = (const uint64_t[]) { AV_CH_LAYOUT_MONO,
306
                                                  AV_CH_LAYOUT_STEREO,
307
                                                  0 },
308
    .priv_class            = &libmp3lame_class,
309
    .defaults              = libmp3lame_defaults,
310
};