nbody_barnes_hut

nbody_barnes_hut is designed to facilitate the simulation of N-body systems in O(nlogn) time. This is useful for many applications: common ones are gravitational simulations and electrostatic simulations. Simulations in 2D and 3D are both supported.

lib.rs:

Nbody barnes hut

nbody_barnes_hut is designed to facilitate the simulation of N-body systems in O(nlogn) time. This is useful for many applications: common ones are gravitational simulations and electrostatic simulations. Simulations in 2D and 3D are both supported.

This crate is not multithreaded. Rather, call calc_forces_on_particle in a multithreaded loop (for example, with rayon). The time to create the tree is negligible in comparison to the time used calculating forces.

Example

Here is a basic 3D gravitational simulator:

use rand::Rng;
use nbody_barnes_hut::particle_3d::Particle3D;
use nbody_barnes_hut::vector_3d::Vector3D;
use nbody_barnes_hut::barnes_hut_3d::OctTree;

const G: f64 = 6.67E-11; // Newton's Gravitational Constant

// Create 10 000 random points
let mut rng = rand::thread_rng();
let points: Vec<Particle3D> = (0..10_000)
    .map(|_| {
        let pos = Vector3D::new(
            rng.gen_range(-1000.0, 1000.0),
            rng.gen_range(-1000.0, 1000.0),
            rng.gen_range(-1000.0, 1000.0),
        );
        Particle3D::new(pos, 30.0)
    })
    .collect();

// This is pretty hacky
let points_ref = &points.iter().collect::<Vec<&Particle3D>>()[..];

let tree = OctTree::new(points_ref, 0.5);

for p in &points {
    // Do something with this value
    let acceleration_on_particle = tree.calc_forces_on_particle(
        p.position,
        (),
        |d_squared, mass, dis_vec, _| {
            // dis_vec is not normalized, so we have to normalize it here
            G * mass * dis_vec / (d_squared * d_squared.sqrt())
        },
    );
}