*/
#include "hb.h"
-
-#include "libavcodec/avcodec.h"
-#include "libavformat/avformat.h"
-#include "libswscale/swscale.h"
+#include "hbffmpeg.h"
+#include "libavcodec/audioconvert.h"
static int decavcodecInit( hb_work_object_t *, hb_job_t * );
static int decavcodecWork( hb_work_object_t *, hb_buffer_t **, hb_buffer_t ** );
pv->parser = av_parser_init( codec_id );
pv->context = avcodec_alloc_context();
- avcodec_open( pv->context, codec );
+ hb_avcodec_open( pv->context, codec );
return 0;
}
}
if ( pv->context && pv->context->codec )
{
- avcodec_close( pv->context );
+ hb_avcodec_close( pv->context );
}
if ( pv->list )
{
*buf_out = NULL;
+ if ( in->start < -1 && pv->pts_next <= 0 )
+ {
+ // discard buffers that start before video time 0
+ return HB_WORK_OK;
+ }
+
cur = ( in->start < 0 )? pv->pts_next : in->start;
pos = 0;
return 0;
}
+static const int chan2layout[] = {
+ HB_INPUT_CH_LAYOUT_MONO, // We should allow no audio really.
+ HB_INPUT_CH_LAYOUT_MONO,
+ HB_INPUT_CH_LAYOUT_STEREO,
+ HB_INPUT_CH_LAYOUT_2F1R,
+ HB_INPUT_CH_LAYOUT_2F2R,
+ HB_INPUT_CH_LAYOUT_3F2R,
+ HB_INPUT_CH_LAYOUT_4F2R,
+ HB_INPUT_CH_LAYOUT_STEREO,
+ HB_INPUT_CH_LAYOUT_STEREO,
+};
+
static int decavcodecBSInfo( hb_work_object_t *w, const hb_buffer_t *buf,
hb_work_info_t *info )
{
hb_work_private_t *pv = w->private_data;
+ int ret = 0;
memset( info, 0, sizeof(*info) );
// now we just return dummy values if there's a codec that will handle it.
AVCodec *codec = avcodec_find_decoder( w->codec_param? w->codec_param :
CODEC_ID_MP2 );
- if ( codec )
+ if ( ! codec )
{
- static char codec_name[64];
+ // there's no ffmpeg codec for this audio type - give up
+ return -1;
+ }
- info->name = strncpy( codec_name, codec->name, sizeof(codec_name)-1 );
- info->bitrate = 384000;
- info->rate = 48000;
- info->rate_base = 1;
- info->channel_layout = HB_INPUT_CH_LAYOUT_STEREO;
- return 1;
+ static char codec_name[64];
+ info->name = strncpy( codec_name, codec->name, sizeof(codec_name)-1 );
+
+ AVCodecParserContext *parser = av_parser_init( codec->id );
+ AVCodecContext *context = avcodec_alloc_context();
+ hb_avcodec_open( context, codec );
+#if defined( SYS_CYGWIN )
+ uint8_t *buffer = memalign(16, AVCODEC_MAX_AUDIO_FRAME_SIZE);
+#else
+ uint8_t *buffer = malloc( AVCODEC_MAX_AUDIO_FRAME_SIZE );
+#endif
+ int out_size = AVCODEC_MAX_AUDIO_FRAME_SIZE;
+ unsigned char *pbuffer;
+ int pos = 0, pbuffer_size;
+
+ while ( pos < buf->size )
+ {
+ int len = av_parser_parse( parser, context, &pbuffer, &pbuffer_size,
+ buf->data + pos, buf->size - pos,
+ buf->start, buf->start );
+ pos += len;
+ if ( pbuffer_size > 0 )
+ {
+ len = avcodec_decode_audio2( context, (int16_t*)buffer, &out_size,
+ pbuffer, pbuffer_size );
+ if ( len > 0 && context->sample_rate > 0 )
+ {
+ info->bitrate = context->bit_rate;
+ info->rate = context->sample_rate;
+ info->rate_base = 1;
+ info->channel_layout = chan2layout[context->channels & 7];
+ ret = 1;
+ break;
+ }
+ }
}
- return -1;
+ free( buffer );
+ av_parser_close( parser );
+ hb_avcodec_close( context );
+ return ret;
}
/* -------------------------------------------------------------
static int decodeFrame( hb_work_private_t *pv, uint8_t *data, int size )
{
- int got_picture;
+ int got_picture, oldlevel = 0;
AVFrame frame;
+ if ( global_verbosity_level <= 1 )
+ {
+ oldlevel = av_log_get_level();
+ av_log_set_level( AV_LOG_QUIET );
+ }
if ( avcodec_decode_video( pv->context, &frame, &got_picture, data, size ) < 0 )
{
++pv->decode_errors;
}
+ if ( global_verbosity_level <= 1 )
+ {
+ av_log_set_level( oldlevel );
+ }
if( got_picture )
{
// ffmpeg makes it hard to attach a pts to a frame. if the MPEG ES
// There's a mis-feature in ffmpeg that causes the context to be
// incorrectly initialized the 1st time avcodec_open is called.
// If you close it and open a 2nd time, it finishes the job.
- avcodec_open( pv->context, codec );
- avcodec_close( pv->context );
- avcodec_open( pv->context, codec );
+ hb_avcodec_open( pv->context, codec );
+ hb_avcodec_close( pv->context );
+ hb_avcodec_open( pv->context, codec );
}
if( in->start >= 0 )
info->rate = 27000000;
info->rate_base = (int64_t)context->time_base.num * 27000000LL /
context->time_base.den;
+ if ( context->ticks_per_frame > 1 )
+ {
+ // for ffmpeg 0.5 & later, the H.264 & MPEG-2 time base is
+ // field rate rather than frame rate so convert back to frames.
+ info->rate_base *= context->ticks_per_frame;
+ }
/* Sometimes there's no pixel aspect set in the source. In that case,
assume a 1:1 PAR. Otherwise, preserve the source PAR. */
if ( ! pv->context->codec )
{
AVCodec *codec = avcodec_find_decoder( pv->context->codec_id );
- avcodec_open( pv->context, codec );
+ hb_avcodec_open( pv->context, codec );
}
// set up our best guess at the frame duration.
// the frame rate in the codec is usually bogus but it's sometimes
// complicated, machine dependent alignment here we use the
// fact that malloc returns an aligned pointer on most architectures.
- #ifdef SYS_CYGWIN
+ #if defined( SYS_CYGWIN )
// Cygwin's malloc doesn't appear to return 16-byte aligned memory so use memalign instead.
pv->buffer = memalign(16, AVCODEC_MAX_AUDIO_FRAME_SIZE);
#else
pos += len;
if( out_size > 0 )
{
+ // We require signed 16-bit ints for the output format. If
+ // we got something different convert it.
+ if ( context->sample_fmt != SAMPLE_FMT_S16 )
+ {
+ // Note: av_audio_convert seems to be a work-in-progress but
+ // looks like it will eventually handle general audio
+ // mixdowns which would allow us much more flexibility
+ // in handling multichannel audio in HB. If we were doing
+ // anything more complicated than a one-for-one format
+ // conversion we'd probably want to cache the converter
+ // context in the pv.
+ int isamp = av_get_bits_per_sample_format( context->sample_fmt ) / 8;
+ AVAudioConvert *ctx = av_audio_convert_alloc( SAMPLE_FMT_S16, 1,
+ context->sample_fmt, 1,
+ NULL, 0 );
+ // get output buffer size (in 2-byte samples) then malloc a buffer
+ out_size = ( out_size * 2 ) / isamp;
+ buffer = malloc( out_size );
+
+ // we're doing straight sample format conversion which behaves as if
+ // there were only one channel.
+ const void * const ibuf[6] = { pv->buffer };
+ void * const obuf[6] = { buffer };
+ const int istride[6] = { isamp };
+ const int ostride[6] = { 2 };
+
+ av_audio_convert( ctx, obuf, ostride, ibuf, istride, out_size >> 1 );
+ av_audio_convert_free( ctx );
+ }
hb_buffer_t *buf = hb_buffer_init( 2 * out_size );
// convert from bytes to total samples
fl32[i] = buffer[i];
}
hb_list_add( pv->list, buf );
+
+ // if we allocated a buffer for sample format conversion, free it
+ if ( buffer != pv->buffer )
+ {
+ free( buffer );
+ }
}
}
}
}
hb_work_private_t *pv = w->private_data;
+
+ if ( (*buf_in)->start < -1 && pv->pts_next <= 0 )
+ {
+ // discard buffers that start before video time 0
+ *buf_out = NULL;
+ return HB_WORK_OK;
+ }
+
if ( ! pv->context )
{
init_ffmpeg_context( w );
OSDN Git Service
