Optimize control point interpolation.
This commit is contained in:
parent
f168b90e07
commit
1fae0b89d7
|
@ -319,7 +319,8 @@ Control_Sequence::handle ( int m )
|
|||
static inline float
|
||||
linear_interpolate ( float y1, float y2, float mu )
|
||||
{
|
||||
return y1 + mu * ( y2 - y1 );
|
||||
// return y1 + mu * ( y2 - y1 );
|
||||
return y1 * ( 1.0f - mu ) + y2 * mu;
|
||||
}
|
||||
|
||||
static inline float
|
||||
|
@ -328,23 +329,6 @@ sigmoid_interpolate ( float y1, float y2, float mu )
|
|||
return linear_interpolate( y1, y2, ( 1 - cos( mu * M_PI ) ) / 2 );
|
||||
}
|
||||
|
||||
static inline float
|
||||
quadratic_interpolate ( float y1, float y2, float mu )
|
||||
{
|
||||
return ( y1 * y1 ) * ( mu * mu ) + ( 2.0f * y1 * y2 ) * mu * ( y1 * y1 );
|
||||
}
|
||||
|
||||
/* static inline float */
|
||||
/* quadratic_interpolate ( float y1, float y2, float mu ) */
|
||||
/* { */
|
||||
/* } */
|
||||
|
||||
/* static inline float */
|
||||
/* exponential_interpolate ( float y1, float y2, float mu ) */
|
||||
/* { */
|
||||
/* // return y1 * pow( y2 / y1, mu ); */
|
||||
/* } */
|
||||
|
||||
/* THREAD: ?? */
|
||||
/** fill buf with /nframes/ of interpolated control curve values
|
||||
* starting at /frame/ */
|
||||
|
@ -363,22 +347,25 @@ Control_Sequence::play ( sample_t *buf, nframes_t frame, nframes_t nframes )
|
|||
if ( p2->when() < frame )
|
||||
continue;
|
||||
|
||||
nframes_t d = p2->when() - p1->when();
|
||||
/* do incremental linear interpolation */
|
||||
|
||||
for ( nframes_t i = frame - p1->when(); i < d; ++i )
|
||||
{
|
||||
// *(buf++) = 1.0f - ( 2 * sigmoid_interpolate( p1->control(), p2->control(), i / (float)d ) );
|
||||
*(buf++) = 1.0f - ( 2 * linear_interpolate( p1->control(), p2->control(), i / (float)d ) );
|
||||
// *(buf++) = 1.0f - ( 2 * quadratic_interpolate( p1->control(), p2->control(), i / (float)d ) );
|
||||
const nframes_t len = p2->when() - p1->when();
|
||||
|
||||
if ( ! n-- )
|
||||
return nframes;
|
||||
/* scale to -1.0 to 1.0 */
|
||||
const float y1 = 1.0f - ( 2.0f * p1->control() );
|
||||
const float y2 = 1.0f - ( 2.0f * p2->control() );
|
||||
|
||||
const nframes_t start = frame - p1->when();
|
||||
const float incr = ( y2 - y1 ) / (float)len;
|
||||
|
||||
float v = y1 + start * incr;
|
||||
|
||||
for ( nframes_t i = start; i < len && n--; ++i, v += incr )
|
||||
*(buf++) = v;
|
||||
|
||||
frame++;
|
||||
}
|
||||
}
|
||||
|
||||
return frame;
|
||||
return nframes - n;
|
||||
}
|
||||
|
||||
|
||||
|
|
|
@ -50,7 +50,7 @@
|
|||
/* TODO: read/write data from/to disk in larger chunks to avoid
|
||||
* excessive seeking. 256k is supposedly the sweetspot. */
|
||||
|
||||
//float Disk_Stream::seconds_to_buffer = 5.0f;
|
||||
// float Disk_Stream::seconds_to_buffer = 5.0f;
|
||||
float Disk_Stream::seconds_to_buffer = 2.0f;
|
||||
// size_t Disk_Stream::disk_block_frames = 2048;
|
||||
|
||||
|
|
|
@ -142,12 +142,13 @@ Playback_DS::disk_thread ( void )
|
|||
|
||||
for ( int i = channels(); i--; )
|
||||
{
|
||||
while ( jack_ringbuffer_write_space( _rb[ i ] ) < block_size )
|
||||
{
|
||||
printf( "IO: disk buffer overrun!\n" );
|
||||
/* FIXME: is this *really* the right thing to do? */
|
||||
usleep( 2000 );
|
||||
}
|
||||
|
||||
/* while ( jack_ringbuffer_write_space( _rb[ i ] ) < block_size ) */
|
||||
/* { */
|
||||
/* printf( "IO: disk buffer overrun!\n" ); */
|
||||
/* /\* FIXME: is this *really* the right thing to do? *\/ */
|
||||
/* usleep( 2000 ); */
|
||||
/* } */
|
||||
|
||||
/* deinterleave direcectly into the ringbuffer to avoid
|
||||
* unnecessary copying */
|
||||
|
|
Loading…
Reference in New Issue