ffmpeg/libavcodec/libvorbis.c

561 lines
19 KiB
C

/*
* copyright (c) 2002 Mark Hills <mark@pogo.org.uk>
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
/**
* @file
* Vorbis encoding support via libvorbisenc.
* @author Mark Hills <mark@pogo.org.uk>
*/
#include <vorbis/vorbisenc.h>
#include "libavutil/fifo.h"
#include "libavutil/opt.h"
#include "avcodec.h"
#include "audio_frame_queue.h"
#include "bytestream.h"
#include "internal.h"
#include "vorbis.h"
#include "vorbis_parser.h"
#undef NDEBUG
#include <assert.h>
/* Number of samples the user should send in each call.
* This value is used because it is the LCD of all possible frame sizes, so
* an output packet will always start at the same point as one of the input
* packets.
*/
#define OGGVORBIS_FRAME_SIZE 64
#define BUFFER_SIZE (1024 * 64)
typedef struct OggVorbisContext {
AVClass *av_class; /**< class for AVOptions */
AVFrame frame;
vorbis_info vi; /**< vorbis_info used during init */
vorbis_dsp_state vd; /**< DSP state used for analysis */
vorbis_block vb; /**< vorbis_block used for analysis */
AVFifoBuffer *pkt_fifo; /**< output packet buffer */
int eof; /**< end-of-file flag */
int dsp_initialized; /**< vd has been initialized */
vorbis_comment vc; /**< VorbisComment info */
ogg_packet op; /**< ogg packet */
double iblock; /**< impulse block bias option */
VorbisParseContext vp; /**< parse context to get durations */
AudioFrameQueue afq; /**< frame queue for timestamps */
} OggVorbisContext;
static const AVOption options[] = {
{ "iblock", "Sets the impulse block bias", offsetof(OggVorbisContext, iblock), AV_OPT_TYPE_DOUBLE, { .dbl = 0 }, -15, 0, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_ENCODING_PARAM },
{ NULL }
};
static const AVCodecDefault defaults[] = {
{ "b", "0" },
{ NULL },
};
static const AVClass class = { "libvorbis", av_default_item_name, options, LIBAVUTIL_VERSION_INT };
static int vorbis_error_to_averror(int ov_err)
{
switch (ov_err) {
case OV_EFAULT: return AVERROR_BUG;
case OV_EINVAL: return AVERROR(EINVAL);
case OV_EIMPL: return AVERROR(EINVAL);
default: return AVERROR_UNKNOWN;
}
}
static av_cold int oggvorbis_init_encoder(vorbis_info *vi,
AVCodecContext *avctx)
{
OggVorbisContext *s = avctx->priv_data;
double cfreq;
int ret;
if (avctx->flags & CODEC_FLAG_QSCALE || !avctx->bit_rate) {
/* variable bitrate
* NOTE: we use the oggenc range of -1 to 10 for global_quality for
* user convenience, but libvorbis uses -0.1 to 1.0.
*/
float q = avctx->global_quality / (float)FF_QP2LAMBDA;
/* default to 3 if the user did not set quality or bitrate */
if (!(avctx->flags & CODEC_FLAG_QSCALE))
q = 3.0;
if ((ret = vorbis_encode_setup_vbr(vi, avctx->channels,
avctx->sample_rate,
q / 10.0)))
goto error;
} else {
int minrate = avctx->rc_min_rate > 0 ? avctx->rc_min_rate : -1;
int maxrate = avctx->rc_max_rate > 0 ? avctx->rc_max_rate : -1;
/* average bitrate */
if ((ret = vorbis_encode_setup_managed(vi, avctx->channels,
avctx->sample_rate, maxrate,
avctx->bit_rate, minrate)))
goto error;
/* variable bitrate by estimate, disable slow rate management */
if (minrate == -1 && maxrate == -1)
if ((ret = vorbis_encode_ctl(vi, OV_ECTL_RATEMANAGE2_SET, NULL)))
goto error; /* should not happen */
}
/* cutoff frequency */
if (avctx->cutoff > 0) {
cfreq = avctx->cutoff / 1000.0;
if ((ret = vorbis_encode_ctl(vi, OV_ECTL_LOWPASS_SET, &cfreq)))
goto error; /* should not happen */
}
/* impulse block bias */
if (s->iblock) {
if ((ret = vorbis_encode_ctl(vi, OV_ECTL_IBLOCK_SET, &s->iblock)))
goto error;
}
if (avctx->channels == 3 &&
avctx->channel_layout != (AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER) ||
avctx->channels == 4 &&
avctx->channel_layout != AV_CH_LAYOUT_2_2 &&
avctx->channel_layout != AV_CH_LAYOUT_QUAD ||
avctx->channels == 5 &&
avctx->channel_layout != AV_CH_LAYOUT_5POINT0 &&
avctx->channel_layout != AV_CH_LAYOUT_5POINT0_BACK ||
avctx->channels == 6 &&
avctx->channel_layout != AV_CH_LAYOUT_5POINT1 &&
avctx->channel_layout != AV_CH_LAYOUT_5POINT1_BACK ||
avctx->channels == 7 &&
avctx->channel_layout != (AV_CH_LAYOUT_5POINT1|AV_CH_BACK_CENTER) ||
avctx->channels == 8 &&
avctx->channel_layout != AV_CH_LAYOUT_7POINT1) {
if (avctx->channel_layout) {
char name[32];
av_get_channel_layout_string(name, sizeof(name), avctx->channels,
avctx->channel_layout);
av_log(avctx, AV_LOG_ERROR, "%s not supported by Vorbis: "
"output stream will have incorrect "
"channel layout.\n", name);
} else {
av_log(avctx, AV_LOG_WARNING, "No channel layout specified. The encoder "
"will use Vorbis channel layout for "
"%d channels.\n", avctx->channels);
}
}
if ((ret = vorbis_encode_setup_init(vi)))
goto error;
return 0;
error:
return vorbis_error_to_averror(ret);
}
/* How many bytes are needed for a buffer of length 'l' */
static int xiph_len(int l)
{
return 1 + l / 255 + l;
}
static av_cold int oggvorbis_encode_close(AVCodecContext *avctx)
{
OggVorbisContext *s = avctx->priv_data;
/* notify vorbisenc this is EOF */
if (s->dsp_initialized)
vorbis_analysis_wrote(&s->vd, 0);
vorbis_block_clear(&s->vb);
vorbis_dsp_clear(&s->vd);
vorbis_info_clear(&s->vi);
av_fifo_free(s->pkt_fifo);
ff_af_queue_close(&s->afq);
#if FF_API_OLD_ENCODE_AUDIO
av_freep(&avctx->coded_frame);
#endif
av_freep(&avctx->extradata);
return 0;
}
static av_cold int oggvorbis_encode_init(AVCodecContext *avctx)
{
OggVorbisContext *s = avctx->priv_data;
ogg_packet header, header_comm, header_code;
uint8_t *p;
unsigned int offset;
int ret;
vorbis_info_init(&s->vi);
if ((ret = oggvorbis_init_encoder(&s->vi, avctx))) {
av_log(avctx, AV_LOG_ERROR, "encoder setup failed\n");
goto error;
}
if ((ret = vorbis_analysis_init(&s->vd, &s->vi))) {
av_log(avctx, AV_LOG_ERROR, "analysis init failed\n");
ret = vorbis_error_to_averror(ret);
goto error;
}
s->dsp_initialized = 1;
if ((ret = vorbis_block_init(&s->vd, &s->vb))) {
av_log(avctx, AV_LOG_ERROR, "dsp init failed\n");
ret = vorbis_error_to_averror(ret);
goto error;
}
vorbis_comment_init(&s->vc);
vorbis_comment_add_tag(&s->vc, "encoder", LIBAVCODEC_IDENT);
if ((ret = vorbis_analysis_headerout(&s->vd, &s->vc, &header, &header_comm,
&header_code))) {
ret = vorbis_error_to_averror(ret);
goto error;
}
avctx->extradata_size = 1 + xiph_len(header.bytes) +
xiph_len(header_comm.bytes) +
header_code.bytes;
p = avctx->extradata = av_malloc(avctx->extradata_size +
FF_INPUT_BUFFER_PADDING_SIZE);
if (!p) {
ret = AVERROR(ENOMEM);
goto error;
}
p[0] = 2;
offset = 1;
offset += av_xiphlacing(&p[offset], header.bytes);
offset += av_xiphlacing(&p[offset], header_comm.bytes);
memcpy(&p[offset], header.packet, header.bytes);
offset += header.bytes;
memcpy(&p[offset], header_comm.packet, header_comm.bytes);
offset += header_comm.bytes;
memcpy(&p[offset], header_code.packet, header_code.bytes);
offset += header_code.bytes;
assert(offset == avctx->extradata_size);
if ((ret = avpriv_vorbis_parse_extradata(avctx, &s->vp)) < 0) {
av_log(avctx, AV_LOG_ERROR, "invalid extradata\n");
return ret;
}
vorbis_comment_clear(&s->vc);
avctx->frame_size = OGGVORBIS_FRAME_SIZE;
ff_af_queue_init(avctx, &s->afq);
s->pkt_fifo = av_fifo_alloc(BUFFER_SIZE);
if (!s->pkt_fifo) {
ret = AVERROR(ENOMEM);
goto error;
}
#if FF_API_OLD_ENCODE_AUDIO
avctx->coded_frame = avcodec_alloc_frame();
if (!avctx->coded_frame) {
ret = AVERROR(ENOMEM);
goto error;
}
#endif
return 0;
error:
oggvorbis_encode_close(avctx);
return ret;
}
static int oggvorbis_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
const AVFrame *frame, int *got_packet_ptr)
{
OggVorbisContext *s = avctx->priv_data;
ogg_packet op;
int ret, duration;
/* send samples to libvorbis */
if (frame) {
const float *audio = (const float *)frame->data[0];
const int samples = frame->nb_samples;
float **buffer;
int c, channels = s->vi.channels;
buffer = vorbis_analysis_buffer(&s->vd, samples);
for (c = 0; c < channels; c++) {
int i;
int co = (channels > 8) ? c :
ff_vorbis_encoding_channel_layout_offsets[channels - 1][c];
for (i = 0; i < samples; i++)
buffer[c][i] = audio[i * channels + co];
}
if ((ret = vorbis_analysis_wrote(&s->vd, samples)) < 0) {
av_log(avctx, AV_LOG_ERROR, "error in vorbis_analysis_wrote()\n");
return vorbis_error_to_averror(ret);
}
if ((ret = ff_af_queue_add(&s->afq, frame) < 0))
return ret;
} else {
if (!s->eof)
if ((ret = vorbis_analysis_wrote(&s->vd, 0)) < 0) {
av_log(avctx, AV_LOG_ERROR, "error in vorbis_analysis_wrote()\n");
return vorbis_error_to_averror(ret);
}
s->eof = 1;
}
/* retrieve available packets from libvorbis */
while ((ret = vorbis_analysis_blockout(&s->vd, &s->vb)) == 1) {
if ((ret = vorbis_analysis(&s->vb, NULL)) < 0)
break;
if ((ret = vorbis_bitrate_addblock(&s->vb)) < 0)
break;
/* add any available packets to the output packet buffer */
while ((ret = vorbis_bitrate_flushpacket(&s->vd, &op)) == 1) {
if (av_fifo_space(s->pkt_fifo) < sizeof(ogg_packet) + op.bytes) {
av_log(avctx, AV_LOG_ERROR, "packet buffer is too small\n");
return AVERROR_BUG;
}
av_fifo_generic_write(s->pkt_fifo, &op, sizeof(ogg_packet), NULL);
av_fifo_generic_write(s->pkt_fifo, op.packet, op.bytes, NULL);
}
if (ret < 0) {
av_log(avctx, AV_LOG_ERROR, "error getting available packets\n");
break;
}
}
if (ret < 0) {
av_log(avctx, AV_LOG_ERROR, "error getting available packets\n");
return vorbis_error_to_averror(ret);
}
/* check for available packets */
if (av_fifo_size(s->pkt_fifo) < sizeof(ogg_packet))
return 0;
av_fifo_generic_read(s->pkt_fifo, &op, sizeof(ogg_packet), NULL);
if ((ret = ff_alloc_packet2(avctx, avpkt, op.bytes)))
return ret;
av_fifo_generic_read(s->pkt_fifo, avpkt->data, op.bytes, NULL);
avpkt->pts = ff_samples_to_time_base(avctx, op.granulepos);
duration = avpriv_vorbis_parse_frame(&s->vp, avpkt->data, avpkt->size);
if (duration > 0) {
/* we do not know encoder delay until we get the first packet from
* libvorbis, so we have to update the AudioFrameQueue counts */
if (!avctx->delay) {
avctx->delay = duration;
s->afq.remaining_delay += duration;
s->afq.remaining_samples += duration;
}
ff_af_queue_remove(&s->afq, duration, &avpkt->pts, &avpkt->duration);
}
*got_packet_ptr = 1;
return 0;
}
AVCodec ff_libvorbis_encoder = {
.name = "libvorbis",
.type = AVMEDIA_TYPE_AUDIO,
.id = CODEC_ID_VORBIS,
.priv_data_size = sizeof(OggVorbisContext),
.init = oggvorbis_encode_init,
.encode2 = oggvorbis_encode_frame,
.close = oggvorbis_encode_close,
.capabilities = CODEC_CAP_DELAY,
.sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLT,
AV_SAMPLE_FMT_NONE },
.long_name = NULL_IF_CONFIG_SMALL("libvorbis Vorbis"),
.priv_class = &class,
.defaults = defaults,
};
static int oggvorbis_decode_init(AVCodecContext *avccontext) {
OggVorbisContext *context = avccontext->priv_data ;
uint8_t *p= avccontext->extradata;
int i, hsizes[3];
unsigned char *headers[3], *extradata = avccontext->extradata;
vorbis_info_init(&context->vi) ;
vorbis_comment_init(&context->vc) ;
if(! avccontext->extradata_size || ! p) {
av_log(avccontext, AV_LOG_ERROR, "vorbis extradata absent\n");
return -1;
}
if(p[0] == 0 && p[1] == 30) {
for(i = 0; i < 3; i++){
hsizes[i] = bytestream_get_be16(&p);
headers[i] = p;
p += hsizes[i];
}
} else if(*p == 2) {
unsigned int offset = 1;
p++;
for(i=0; i<2; i++) {
hsizes[i] = 0;
while((*p == 0xFF) && (offset < avccontext->extradata_size)) {
hsizes[i] += 0xFF;
offset++;
p++;
}
if(offset >= avccontext->extradata_size - 1) {
av_log(avccontext, AV_LOG_ERROR,
"vorbis header sizes damaged\n");
return -1;
}
hsizes[i] += *p;
offset++;
p++;
}
hsizes[2] = avccontext->extradata_size - hsizes[0]-hsizes[1]-offset;
#if 0
av_log(avccontext, AV_LOG_DEBUG,
"vorbis header sizes: %d, %d, %d, / extradata_len is %d \n",
hsizes[0], hsizes[1], hsizes[2], avccontext->extradata_size);
#endif
headers[0] = extradata + offset;
headers[1] = extradata + offset + hsizes[0];
headers[2] = extradata + offset + hsizes[0] + hsizes[1];
} else {
av_log(avccontext, AV_LOG_ERROR,
"vorbis initial header len is wrong: %d\n", *p);
return -1;
}
for(i=0; i<3; i++){
context->op.b_o_s= i==0;
context->op.bytes = hsizes[i];
context->op.packet = headers[i];
if(vorbis_synthesis_headerin(&context->vi, &context->vc, &context->op)<0){
av_log(avccontext, AV_LOG_ERROR, "%d. vorbis header damaged\n", i+1);
return -1;
}
}
avccontext->channels = context->vi.channels;
avccontext->sample_rate = context->vi.rate;
avccontext->time_base= (AVRational){1, avccontext->sample_rate};
vorbis_synthesis_init(&context->vd, &context->vi);
vorbis_block_init(&context->vd, &context->vb);
return 0 ;
}
static inline int conv(int samples, float **pcm, char *buf, int channels) {
int i, j;
ogg_int16_t *ptr, *data = (ogg_int16_t*)buf ;
float *mono ;
for(i = 0 ; i < channels ; i++){
ptr = &data[i];
mono = pcm[i] ;
for(j = 0 ; j < samples ; j++) {
*ptr = av_clip_int16(mono[j] * 32767.f);
ptr += channels;
}
}
return 0 ;
}
static int oggvorbis_decode_frame(AVCodecContext *avccontext, void *data,
int *got_frame_ptr, AVPacket *avpkt)
{
OggVorbisContext *context = avccontext->priv_data ;
float **pcm ;
ogg_packet *op= &context->op;
int samples, total_samples, total_bytes;
int ret;
int16_t *output;
if(!avpkt->size){
//FIXME flush
return 0;
}
context->frame.nb_samples = 8192*4;
if ((ret = avccontext->get_buffer(avccontext, &context->frame)) < 0) {
av_log(avccontext, AV_LOG_ERROR, "get_buffer() failed\n");
return ret;
}
output = (int16_t *)context->frame.data[0];
op->packet = avpkt->data;
op->bytes = avpkt->size;
// av_log(avccontext, AV_LOG_DEBUG, "%d %d %d %"PRId64" %"PRId64" %d %d\n", op->bytes, op->b_o_s, op->e_o_s, op->granulepos, op->packetno, buf_size, context->vi.rate);
/* for(i=0; i<op->bytes; i++)
av_log(avccontext, AV_LOG_DEBUG, "%02X ", op->packet[i]);
av_log(avccontext, AV_LOG_DEBUG, "\n");*/
if(vorbis_synthesis(&context->vb, op) == 0)
vorbis_synthesis_blockin(&context->vd, &context->vb) ;
total_samples = 0 ;
total_bytes = 0 ;
while((samples = vorbis_synthesis_pcmout(&context->vd, &pcm)) > 0) {
conv(samples, pcm, (char*)output + total_bytes, context->vi.channels) ;
total_bytes += samples * 2 * context->vi.channels ;
total_samples += samples ;
vorbis_synthesis_read(&context->vd, samples) ;
}
context->frame.nb_samples = total_samples;
*got_frame_ptr = 1;
*(AVFrame *)data = context->frame;
return avpkt->size;
}
static int oggvorbis_decode_close(AVCodecContext *avccontext) {
OggVorbisContext *context = avccontext->priv_data ;
vorbis_info_clear(&context->vi) ;
vorbis_comment_clear(&context->vc) ;
return 0 ;
}
AVCodec ff_libvorbis_decoder = {
.name = "libvorbis",
.type = AVMEDIA_TYPE_AUDIO,
.id = CODEC_ID_VORBIS,
.priv_data_size = sizeof(OggVorbisContext),
.init = oggvorbis_decode_init,
.decode = oggvorbis_decode_frame,
.close = oggvorbis_decode_close,
.capabilities = CODEC_CAP_DELAY,
} ;