alex/nih-plug
1
0
Fork 0
Code Activity

Use double precision for exponential step sizes

Browse source 
Otherwise you will get issues with rounding errors if someone ever tries
to have smoothing times over a second or so.
This commit is contained in:
Robbert van der Helm 2022-05-12 13:50:45 +02:00
parent 6a40cc9916
commit e8764e7327
1 changed files with 2 additions and 2 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

4
src/param/smoothing.rs
View file

@ -171,13 +171,13 @@ impl<T: Smoothable> Smoother<T> {
// We need to solve `current * (step_size ^ steps_left) = target` for
// `step_size`
nih_debug_assert_ne!(current, 0.0);
(self.target.to_f32() / current).powf((steps_left as f32).recip())
((self.target.to_f32() / current) as f64).powf((steps_left as f64).recip()) as f32
}
// In this case the step size value is the coefficient the current value will be
// multiplied by, while the target value is multipled by one minus the coefficient. This
// reaches 99.99% of the target value after `steps_left`. The smoother will snap to the
// target value after that point.
SmoothingStyle::Exponential(_) => 0.0001f32.powf(1.0 / steps_left as f32),
SmoothingStyle::Exponential(_) => 0.0001f64.powf(1.0 / steps_left as f64) as f32,
};
}