hashtags/lmms
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases 1 Wiki Activity

InstrumentTrack: skip processing for IPH-driven instruments when silent

Browse Source 
InstrumentPlayHandle-driven instruments (i.e. instruments producing only
one sound stream for all notes) are running all the time even if no notes
are running. The plugin itself usually does not consume much CPU time
while silent but all effects afterwards inside LMMS unneccessarily
consume lots of CPU time for processing silent buffer.

Typical case: a song with lots of instruments like ZynAddSubFX.

With this change, all processing is skipped on silent buffers which
notably decreases CPU load for many projects. All effects in following
effect chains continue to run until they're silent as well.

Closes #267.
This commit is contained in:
Tobias Doerffel
2014-02-06 22:59:14 +01:00
parent be5e17c29c
commit 1266278229
1 changed files with 28 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files

46
src/tracks/InstrumentTrack.cpp
View File

@@ -65,6 +65,7 @@
#include "MainWindow.h"
#include "MidiClient.h"
#include "MidiPortMenu.h"
#include "MixHelpers.h"
#include "DataFile.h"
#include "NotePlayHandle.h"
#include "pattern.h"
@@ -150,16 +151,24 @@ InstrumentTrack::~InstrumentTrack()
void InstrumentTrack::processAudioBuffer( sampleFrame * _buf,
const fpp_t _frames,
NotePlayHandle * _n )
void InstrumentTrack::processAudioBuffer( sampleFrame* buf, const fpp_t frames, NotePlayHandle* n )
{
// we must not play the sound if this InstrumentTrack is muted...
if( isMuted() || ( _n && _n->isBbTrackMuted() ) )
if( isMuted() || ( n && n->isBbTrackMuted() ) )
{
return;
}
// Test for silent input data if instrument provides a single stream only (i.e. driven by InstrumentPlayHandle)
// We could do that in all other cases as well but the overhead for silence test is bigger than
// what we potentially save. While playing a note, a NotePlayHandle-driven instrument will produce sound in
// 99 of 100 cases so that test would be a waste of time.
if( n == NULL && MixHelpers::isSilent( buf, frames ) )
{
// skip further processing
return;
}
// if effects "went to sleep" because there was no input, wake them up
// now
m_audioPort.effects()->startRunning();
@@ -167,17 +176,16 @@ void InstrumentTrack::processAudioBuffer( sampleFrame * _buf,
float v_scale = (float) getVolume() / DefaultVolume;
// instruments using instrument-play-handles will call this method
// without any knowledge about notes, so they pass NULL for _n, which
// without any knowledge about notes, so they pass NULL for n, which
// is no problem for us since we just bypass the envelopes+LFOs
if( _n != NULL )
if( n )
{
m_soundShaping.processAudioBuffer( _buf, _frames, _n );
v_scale *= ( (float) _n->getVolume() / DefaultVolume );
m_soundShaping.processAudioBuffer( buf, frames, n );
v_scale *= ( (float) n->getVolume() / DefaultVolume );
}
else
{
if( getVolume() < DefaultVolume &&
m_instrument->isMidiBased() )
if( getVolume() < DefaultVolume && m_instrument->isMidiBased() )
{
v_scale = 1;
}
@@ -185,18 +193,20 @@ void InstrumentTrack::processAudioBuffer( sampleFrame * _buf,
m_audioPort.setNextFxChannel( m_effectChannelModel.value() );
int framesToMix = frames;
int offset = 0;
int panning = m_panningModel.value();
if( _n != NULL )
if( n )
{
panning += _n->getPanning();
framesToMix = qMin<f_cnt_t>( n->framesLeftForCurrentPeriod(), framesToMix );
offset = n->offset();
panning += n->getPanning();
panning = tLimit<int>( panning, PanningLeft, PanningRight );
}
engine::mixer()->bufferToPort( _buf, ( _n != NULL ) ?
qMin<f_cnt_t>(_n->framesLeftForCurrentPeriod(), _frames ) :
_frames,
( _n != NULL ) ? _n->offset() : 0,
panningToVolumeVector( panning, v_scale ),
&m_audioPort );
engine::mixer()->bufferToPort( buf, framesToMix, offset, panningToVolumeVector( panning, v_scale ), &m_audioPort );
}