From 01b5769dc0b3f99ef087269c30f6fb34c8969b0f Mon Sep 17 00:00:00 2001
From: Quaternions <krakow20@gmail.com>
Date: Thu, 12 Oct 2023 03:36:42 -0700
Subject: [PATCH] fix TODOs
---
src/integer.rs | 27 +++++++++++++++++++++------
1 file changed, 21 insertions(+), 6 deletions(-)
diff --git a/src/integer.rs b/src/integer.rs
index 391ca7c..d435753 100644
--- a/src/integer.rs
+++ b/src/integer.rs
@@ -202,23 +202,38 @@ impl Angle32{
pub const PI:Self=Self(-1<<32);
#[inline]
pub fn wrap_from_i64(theta:i64)->Self{
- //TODO: make this good
- Self((theta.wrapping_add(1<<31).rem_euclid(1<<32)-(1<<31)) as i32)
+ //take lower bits
+ //note: this was checked on compiler explorer and compiles to 1 instruction!
+ Self(i32::from_ne_bytes(((theta&((1<<32)-1)) as u32).to_ne_bytes()))
}
#[inline]
pub fn clamp_from_i64(theta:i64)->Self{
- //TODO: make this good
+ //the assembly is a bit confusing for this, I thought it was checking the same thing twice
+ //but it's just checking and then overwriting the value for both upper and lower bounds.
Self(theta.clamp(i32::MIN as i64,i32::MAX as i64) as i32)
}
#[inline]
pub fn get(&self)->i32{
self.0
}
- /// Note that theta_min can be larger than theta_max and it will clamp the other way
+ /// Clamps the value towards the midpoint of the range.
+ /// Note that theta_min can be larger than theta_max and it will wrap clamp the other way around
#[inline]
pub fn clamp(&self,theta_min:Self,theta_max:Self)->Self{
- //TODO: make this good
- theta_min+Self::wrap_from_i64((self.0.wrapping_sub(theta_min.0) as i64).rem_euclid(1<<32).clamp(0,(theta_max.0.wrapping_sub(theta_min.0) as i64).rem_euclid(1<<32)))
+ //((max-min as u32)/2 as i32)+min
+ let midpoint=((
+ u32::from_ne_bytes(theta_max.0.to_ne_bytes())
+ .wrapping_sub(u32::from_ne_bytes(theta_min.0.to_ne_bytes()))
+ /2
+ ) as i32)//(u32::MAX/2) as i32 ALWAYS works
+ .wrapping_add(theta_min.0);
+ //(theta-mid).clamp(max-mid,min-mid)+mid
+ Self(
+ self.0.wrapping_sub(midpoint)
+ .max(theta_min.0.wrapping_sub(midpoint))
+ .min(theta_max.0.wrapping_sub(midpoint))
+ .wrapping_add(midpoint)
+ )
}
#[inline]
pub fn cos(&self)->Unit32{