Faster sample format conversion (#618)

* Remove branch in `u16::to_i16` and `i16::to_u16`

If optimizations are at least at level 1, the conversions compile down to the same code as the old one, but if they are turned off, the new code is faster.

* Faster conversion from `f32` to `u16`

The new code is faster in release mode.
This commit is contained in:
Ram.Type-0 2022-01-14 20:23:30 +09:00 • committed by GitHub
parent f8b1ab53b4
commit 1b522fb3db
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23

View file

@ -51,11 +51,7 @@ unsafe impl Sample for u16 {
#[inline]
fn to_i16(&self) -> i16 {
if *self >= 32768 {
(*self - 32768) as i16
} else {
(*self as i16) - 32767 - 1
}
(*self as i16).wrapping_add(i16::MIN)
}
#[inline]
@ -91,11 +87,7 @@ unsafe impl Sample for i16 {
#[inline]
fn to_u16(&self) -> u16 {
if *self < 0 {
(*self - i16::MIN) as u16
} else {
(*self as u16) + 32768
}
self.wrapping_add(i16::MIN) as u16
}
#[inline]
@ -106,7 +98,7 @@ unsafe impl Sample for i16 {
sample.to_i16()
}
}
const F32_TO_16BIT_INT_MULTIPLIER: f32 = u16::MAX as f32 * 0.5;
unsafe impl Sample for f32 {
const FORMAT: SampleFormat = SampleFormat::F32;
@ -126,7 +118,8 @@ unsafe impl Sample for f32 {
#[inline]
fn to_u16(&self) -> u16 {
(((*self + 1.0) * 0.5) * u16::MAX as f32).round() as u16
self.mul_add(F32_TO_16BIT_INT_MULTIPLIER, F32_TO_16BIT_INT_MULTIPLIER)
.round() as u16
}
#[inline]