Use SetPosition_hard instead of SetPosition_toc for the mp3 position, remove dependency on accurate sync boolean

Resolves #610.

`RageSoundReader_MP3` has three different methods of setting the position. The default choice of `SetPosition_toc` is unable to provide an accurate result. The reason Edit Mode has such a severe desync when working with MP3's is due to the preference of using `SetPosition_toc`.

We can prevent this by using `SetPosition_hard` which works very well.  Judging by comments in the file, it seems like `SetPosition_hard` was too resource intensive to use as the default method, 20+ years ago.

The only downside is, if you start from an arbitrary position (which is **only** possible in Edit Mode), you have a few milliseconds of corrupted audio, but then the sync is perfect.  You can hear this in the demonstration video below.

When we start from the beginning of a file, such as playing a song, or using `Play whole song` in edit mode, `MADLIB_rewind` will be used.

`SetPosition_hard` provides an accurate result, and is consistent with the offset provided by `MADLIB_rewind`. The result is a consistent sync experience whether the MP3 is started from the beginning, or from an arbitrary position midway thru the file.

The `m_bAccurateSync` boolean dependency has been removed from SetPosition since it does not seem to provide any benefit on modern hardware. It seems fine to remove it so that we can call `MADLIB_rewind` or `SetPosition_hard` as fast as possible.

In my testing I never got `SetPosition_estimate` to be used, but I've left it in as a backup option.
This commit is contained in:
sukibaby
2025-03-02 22:21:30 -08:00
committed by teejusb
parent 5326808f38
commit 6f497f70e7
2 changed files with 18 additions and 100 deletions
+18 -99
View File
@@ -755,89 +755,18 @@ bool RageSoundReader_MP3::MADLIB_rewind()
/* Methods of seeking:
*
* 1. We can jump based on a TOC. We potentially have two; the Xing TOC and our
* own index. The Xing TOC is only accurate to 1/256th of the file size,
* so it's unsuitable for precise seeks. Our own TOC is byte-accurate.
* (SetPosition_toc)
* 1. We can seek from any position to any higher position by decoding headers.
* (SetPosition_hard)
*
* 2. We can jump based on the bitrate. This is fast, but not accurate.
* (SetPosition_estimate)
*
* 3. We can seek from any position to any higher position by decoding headers.
* (SetPosition_hard)
* 3. MADLIB_rewind will rewind the stream to the beginning. This is always accurate.
*
* Both 1 and 2 will leave the position behind the actual requested position;
* combine them with 3 to catch up. Never do 3 alone in "fast" mode, since it's
* slow if it ends up seeking from the beginning of the file. Never do 2 in
* "precise" mode.
* SetPosition_estimate will leave the position behind the actual requested position;
* combine them with MADLIB_rewind or SetPosition_hard to catch up.
*/
/* Returns actual position on success, 0 if past EOF, -1 on error. */
int RageSoundReader_MP3::SetPosition_toc( int iFrame, bool Xing )
{
ASSERT( !Xing || mad->has_xing );
ASSERT( mad->length != -1 );
/* This leaves our timer accurate if we're using our own TOC, and inaccurate
* if we're using Xing. */
mad->timer_accurate = !Xing;
int bytepos = -1;
if( Xing )
{
/* We can speed up the seek using the XING tag. First, figure
* out what percentage the requested position falls in. */
ASSERT( SampleRate != 0 );
int ms = int( (iFrame * 1000LL) / SampleRate );
ASSERT( mad->length != 0 );
int percent = ms * 100 / mad->length;
if( percent < 100 )
{
int jump = mad->xingtag.toc[percent];
bytepos = mad->filesize * jump / 256;
}
else
bytepos = 2000000000; /* force EOF */
mad_timer_set( &mad->Timer, 0, percent * mad->length, 100000 );
}
else
{
mad_timer_t desired;
mad_timer_set( &desired, 0, iFrame, SampleRate );
if( mad->tocmap.empty() )
return 1; /* don't have any info */
/* Find the last entry <= iFrame that we actually have an entry for;
* this will get us as close as possible. */
madlib_t::tocmap_t::iterator it = mad->tocmap.upper_bound( desired );
if( it == mad->tocmap.begin() )
return 1; /* don't have any info */
--it;
mad->Timer = it->first;
bytepos = it->second;
}
if( bytepos != -1 )
{
/* Seek backwards up to 4k. */
const int seekpos = std::max( 0, bytepos - 1024*4 );
seek_stream_to_byte( seekpos );
do
{
int ret = do_mad_frame_decode();
if( ret <= 0 )
return ret; /* it set the error */
} while( get_this_frame_byte(mad) < bytepos );
synth_output();
}
return 1;
}
int RageSoundReader_MP3::SetPosition_hard( int iFrame )
{
mad_timer_t desired;
@@ -960,33 +889,23 @@ int RageSoundReader_MP3::SetPosition_estimate( int iFrame )
int RageSoundReader_MP3::SetPosition( int iFrame )
{
if( m_bAccurateSync )
/* Rewinding is always fast and accurate, and SetPosition_estimate is bad at 0. */
if (!iFrame)
{
/* Seek using our own internal (accurate) TOC. */
int ret = SetPosition_toc( iFrame, false );
if( ret <= 0 )
return ret; /* it set the error */
/* Align exactly. */
return SetPosition_hard( iFrame );
MADLIB_rewind();
return 1; /* ok */
}
else
/* We can do a fast jump in VBR with Xing with more accuracy than without Xing. */
if (mad->has_xing)
{
/* Rewinding is always fast and accurate, and SetPosition_estimate is bad at 0. */
if( !iFrame )
{
MADLIB_rewind();
return 1; /* ok */
}
/* We can do a fast jump in VBR with Xing with more accuracy than without Xing. */
if( mad->has_xing )
return SetPosition_toc( iFrame, true );
/* Guess. This is only remotely accurate when we're not VBR, but also
* do it if we have no Xing tag. */
return SetPosition_estimate( iFrame );
return SetPosition_hard(iFrame);
}
LOG->Trace("MP3 decoder had to revert to guessing the song position");
/* Guess. This is only remotely accurate when we're not VBR, but also
* do it if we have no Xing tag. */
return SetPosition_estimate(iFrame);
}
bool RageSoundReader_MP3::SetProperty( const RString &sProperty, float fValue )
-1
View File
@@ -41,7 +41,6 @@ private:
bool MADLIB_rewind();
int SetPosition_toc( int iSample, bool Xing );
int SetPosition_hard( int iSample );
int SetPosition_estimate( int iSample );