noise-functions 0.8.5

A collection of fast and lightweight noise functions.
Documentation
use crate::{
    base::impl_noise,
    cellular::{CellIndex, DistanceFn, DistanceReturnType},
    from_fast_noise_2::cell::MAX_DISTANCE_COUNT,
};

#[cfg(feature = "nightly-simd")]
use core::simd::{f32x2, f32x4};

/// 2/3/4 dimensional noise of the distance to a cell.
#[derive(Debug, Clone, Copy, PartialEq)]
pub struct CustomCellDistance {
    pub jitter: f32,
    pub distance_fn: DistanceFn,
    pub distance_indices: [CellIndex; 2],
    pub return_type: DistanceReturnType,
}

impl CustomCellDistance {
    pub const fn jitter(mut self, jitter: f32) -> Self {
        self.jitter = jitter;
        self
    }

    pub const fn distance_fn(mut self, distance_fn: DistanceFn) -> Self {
        self.distance_fn = distance_fn;
        self
    }

    #[expect(dead_code)]
    pub const fn distance_indices(mut self, distance_indices: [CellIndex; 2]) -> Self {
        self.distance_indices = distance_indices;
        self
    }

    #[expect(dead_code)]
    pub const fn return_type(mut self, return_type: DistanceReturnType) -> Self {
        self.return_type = return_type;
        self
    }
}

impl Default for CustomCellDistance {
    fn default() -> Self {
        Self {
            jitter: 1.0,
            distance_fn: Default::default(),
            distance_indices: Default::default(),
            return_type: Default::default(),
        }
    }
}

impl_noise!(234 CustomCellDistance);

impl CustomCellDistance {
    fn get_return(self, distance: &mut [f32; MAX_DISTANCE_COUNT]) -> f32 {
        use crate::from_fast_noise_2::{inv_sqrt, reciprocal};

        let i0 = self.distance_indices[0] as usize;
        let i1 = self.distance_indices[1] as usize;

        if self.distance_fn == DistanceFn::Euclidean {
            distance[i0] *= inv_sqrt(distance[i0]);
            distance[i1] *= inv_sqrt(distance[i1]);
        }

        match self.return_type {
            DistanceReturnType::Index0 => distance[i0],
            DistanceReturnType::Index0Add1 => distance[i0] + distance[i1],
            DistanceReturnType::Index0Sub1 => distance[i0] - distance[i1],
            DistanceReturnType::Index0Mul1 => distance[i0] * distance[i1],
            DistanceReturnType::Index0Div1 => distance[i0] * reciprocal(distance[i1]),
        }
    }

    #[inline]
    fn gen2(self, [x, y]: [f32; 2], seed: i32) -> f32 {
        // implementation from FastNoise2
        use crate::from_fast_noise_2::{
            cell::{calc_distance2, JITTER_2D, MAX_DISTANCE_COUNT},
            hash_primes2_hb, inv_sqrt, max, min, mul_add, primes,
        };

        let jitter = self.jitter * JITTER_2D;

        let mut distance = [f32::INFINITY; MAX_DISTANCE_COUNT];

        let mut xc = (x as i32).wrapping_sub(1);
        let mut yc_base = (y as i32).wrapping_sub(1);

        let mut xcf = (xc as f32) - x;
        let ycf_base = (yc_base as f32) - y;

        xc = xc.wrapping_mul(primes::X);
        yc_base = yc_base.wrapping_mul(primes::Y);

        for _xi in 0..3 {
            let mut ycf = ycf_base;
            let mut yc = yc_base;

            for _yi in 0..3 {
                let hash = hash_primes2_hb(seed, xc, yc);
                let mut xd = ((hash & 0xffff) as f32) - (0xffff as f32 / 2.0);
                let mut yd = (((hash >> 16) & 0xffff) as f32) - (0xffff as f32 / 2.0);

                let inv_mag = jitter * inv_sqrt(mul_add(xd, xd, yd * yd));
                xd = mul_add(xd, inv_mag, xcf);
                yd = mul_add(yd, inv_mag, ycf);

                let new_distance = calc_distance2(self.distance_fn, xd, yd);

                let mut i = MAX_DISTANCE_COUNT - 1;

                loop {
                    distance[i] = max(min(distance[i], new_distance), distance[i.wrapping_sub(1)]);

                    i = i.wrapping_sub(1);

                    if i == 0 {
                        break;
                    }
                }

                distance[0] = min(distance[0], new_distance);

                ycf += 1.0;
                yc = yc.wrapping_add(primes::Y);
            }

            xcf += 1.0;
            xc = xc.wrapping_add(primes::X);
        }

        self.get_return(&mut distance)
    }

    #[inline]
    fn gen3(self, [x, y, z]: [f32; 3], seed: i32) -> f32 {
        // implementation from FastNoise2
        use crate::from_fast_noise_2::{
            cell::{calc_distance3, JITTER_3D, MAX_DISTANCE_COUNT},
            hash_primes3_hb, inv_sqrt, max, min, mul_add, primes,
        };

        let jitter = self.jitter * JITTER_3D;

        let mut distance = [f32::INFINITY; MAX_DISTANCE_COUNT];

        let mut xc = (x as i32).wrapping_sub(1);
        let mut yc_base = (y as i32).wrapping_sub(1);
        let mut zc_base = (z as i32).wrapping_sub(1);

        let mut xcf = (xc as f32) - x;
        let ycf_base = (yc_base as f32) - y;
        let zcf_base = (zc_base as f32) - z;

        xc = xc.wrapping_mul(primes::X);
        yc_base = yc_base.wrapping_mul(primes::Y);
        zc_base = zc_base.wrapping_mul(primes::Z);

        for _xi in 0..3 {
            let mut ycf = ycf_base;
            let mut yc = yc_base;

            for _yi in 0..3 {
                let mut zcf = zcf_base;
                let mut zc = zc_base;

                for _zi in 0..3 {
                    let hash = hash_primes3_hb(seed, xc, yc, zc);
                    let mut xd = ((hash & 0x3ff) as f32) - (0x3ff as f32 / 2.0);
                    let mut yd = (((hash >> 10) & 0x3ff) as f32) - (0x3ff as f32 / 2.0);
                    let mut zd = (((hash >> 20) & 0x3ff) as f32) - (0x3ff as f32 / 2.0);

                    let inv_mag = jitter * inv_sqrt(mul_add(xd, xd, mul_add(yd, yd, zd * zd)));
                    xd = mul_add(xd, inv_mag, xcf);
                    yd = mul_add(yd, inv_mag, ycf);
                    zd = mul_add(zd, inv_mag, zcf);

                    let new_distance = calc_distance3(self.distance_fn, xd, yd, zd);

                    let mut i = MAX_DISTANCE_COUNT - 1;

                    loop {
                        distance[i] = max(min(distance[i], new_distance), distance[i.wrapping_sub(1)]);

                        i = i.wrapping_sub(1);

                        if i == 0 {
                            break;
                        }
                    }

                    distance[0] = min(distance[0], new_distance);

                    zcf += 1.0;
                    zc = zc.wrapping_add(primes::Z);
                }

                ycf += 1.0;
                yc = yc.wrapping_add(primes::Y);
            }

            xcf += 1.0;
            xc = xc.wrapping_add(primes::X);
        }

        self.get_return(&mut distance)
    }

    #[inline]
    pub(super) fn gen4(self, [x, y, z, w]: [f32; 4], seed: i32) -> f32 {
        // implementation from FastNoise2
        use crate::from_fast_noise_2::{
            cell::{calc_distance4, JITTER_4D, MAX_DISTANCE_COUNT},
            hash_primes4_hb, inv_sqrt, max, min, mul_add, primes,
        };

        let jitter = self.jitter * JITTER_4D;

        let mut distance = [f32::INFINITY; MAX_DISTANCE_COUNT];

        let mut xc = (x as i32).wrapping_sub(1);
        let mut yc_base = (y as i32).wrapping_sub(1);
        let mut zc_base = (z as i32).wrapping_sub(1);
        let mut wc_base = (w as i32).wrapping_sub(1);

        let mut xcf = (xc as f32) - x;
        let ycf_base = (yc_base as f32) - y;
        let zcf_base = (zc_base as f32) - z;
        let wcf_base = (wc_base as f32) - w;

        xc = xc.wrapping_mul(primes::X);
        yc_base = yc_base.wrapping_mul(primes::Y);
        zc_base = zc_base.wrapping_mul(primes::Z);
        wc_base = wc_base.wrapping_mul(primes::W);

        for _xi in 0..3 {
            let mut ycf = ycf_base;
            let mut yc = yc_base;

            for _yi in 0..3 {
                let mut zcf = zcf_base;
                let mut zc = zc_base;

                for _zi in 0..3 {
                    let mut wcf = wcf_base;
                    let mut wc = wc_base;

                    for _wi in 0..3 {
                        let hash = hash_primes4_hb(seed, xc, yc, zc, wc);
                        let mut xd = ((hash & 0xff) as f32) - (0xff as f32 / 2.0);
                        let mut yd = (((hash >> 8) & 0xff) as f32) - (0xff as f32 / 2.0);
                        let mut zd = (((hash >> 16) & 0xff) as f32) - (0xff as f32 / 2.0);
                        let mut wd = (((hash >> 24) & 0xff) as f32) - (0xff as f32 / 2.0);

                        let inv_mag = jitter * inv_sqrt(mul_add(xd, xd, mul_add(yd, yd, mul_add(zd, zd, wd * wd))));
                        xd = mul_add(xd, inv_mag, xcf);
                        yd = mul_add(yd, inv_mag, ycf);
                        zd = mul_add(zd, inv_mag, zcf);
                        wd = mul_add(wd, inv_mag, wcf);

                        let new_distance = calc_distance4(self.distance_fn, xd, yd, zd, wd);

                        let mut i = MAX_DISTANCE_COUNT - 1;

                        loop {
                            distance[i] = max(min(distance[i], new_distance), distance[i.wrapping_sub(1)]);

                            i = i.wrapping_sub(1);

                            if i == 0 {
                                break;
                            }
                        }

                        distance[0] = min(distance[0], new_distance);

                        wcf += 1.0;
                        wc = wc.wrapping_add(primes::W);
                    }

                    zcf += 1.0;
                    zc = zc.wrapping_add(primes::Z);
                }

                ycf += 1.0;
                yc = yc.wrapping_add(primes::Y);
            }

            xcf += 1.0;
            xc = xc.wrapping_add(primes::X);
        }

        self.get_return(&mut distance)
    }

    #[inline]
    #[cfg(feature = "nightly-simd")]
    fn gen2a(self, point: f32x2, seed: i32) -> f32 {
        self.gen2(point.into(), seed)
    }

    #[inline]
    #[cfg(feature = "nightly-simd")]
    fn gen3a(self, point: f32x4, seed: i32) -> f32 {
        self.gen3(*crate::array_4_take_3(point.as_array()), seed)
    }

    #[inline]
    #[cfg(feature = "nightly-simd")]
    pub(super) fn gen4a(self, point: f32x4, seed: i32) -> f32 {
        self.gen4(point.into(), seed)
    }
}