Compare commits

..

43 Commits

Author SHA1 Message Date
99faa4cd3d use av1 for encoding 2025-09-17 10:33:11 -04:00
61f0408bad blur: boxes_for_gaussian changes 2025-03-31 15:41:57 -04:00
9199791f51 blur: usage of width_sub_radius 2025-03-31 15:32:24 -04:00
b6fbc99dac update 2025-03-30 15:44:40 -04:00
00e91a709f add benchmarks 2025-03-28 19:37:43 -04:00
47e09571fc collect grids instead 2025-03-28 17:41:24 -04:00
16887c9712 syntax change 2025-03-28 17:37:00 -04:00
effe506b45 repulstion_distr -> repulsion_distr 2025-03-28 17:35:21 -04:00
492c527498 combine: improve pointer handling 2025-03-28 17:33:26 -04:00
b8f1e28eed tick: simplify parameter passing 2025-03-28 17:27:53 -04:00
ec7cce80b4 util: test improvements 2025-03-28 14:27:44 -04:00
0b3abe71ae update deps 2025-03-28 10:41:23 -04:00
e6cfab4a02 imgdata: make new_from_grid not take a reference 2025-03-28 10:31:07 -04:00
e3fff76792 Grid: inline deposit 2025-03-28 10:22:28 -04:00
ff769df97b replace wrap function with rem_euclid 2025-03-28 10:15:36 -04:00
4330101b68 Agent: angle -> heading 2025-03-28 10:06:13 -04:00
b4e2390690 PopulationConfig: make fields private 2025-03-28 00:17:31 -04:00
b0c9d3888e inline decay_factor 2025-03-28 00:17:05 -04:00
ab226026c3 make fields private 2025-03-28 00:00:19 -04:00
e973404c82 clippy 2025-03-27 23:58:26 -04:00
a0c07364d1 cleanup agent structs 2025-03-27 23:54:02 -04:00
f32315cb5d cleanup imports 2025-03-27 23:52:23 -04:00
6d6794456e improve wrap function 2025-03-27 23:50:01 -04:00
a60847ad6f test + cleanup agent direction code 2025-03-27 23:47:46 -04:00
8dd01ab105 Blur: update run test 2025-03-27 16:14:42 -04:00
50640efb17 Blur: improve tests 2025-03-27 15:53:52 -04:00
d7284fcd37 Grid: remove unneeded test 2025-03-27 14:58:42 -04:00
eee266979c proper doc comments 2025-03-27 14:53:36 -04:00
50e85dec90 Grid: remove unneeded power of two restriction 2025-03-27 14:51:14 -04:00
ab70ce7f53 Grid: move around unsafe block 2025-03-27 14:45:14 -04:00
985fb73042 Grid: make buf and blur private 2025-03-27 14:40:22 -04:00
a8fc644d6c replace common index code 2025-03-27 14:37:36 -04:00
d1f515b17d cleanup Buf struct 2025-03-27 14:35:54 -04:00
75fab93907 cleanup 2025-03-27 14:25:42 -04:00
9881502002 grid: avoid clone of buffer 2025-03-27 14:19:01 -04:00
8d54fa1eb1 settings update 2025-03-27 14:17:38 -04:00
278ccafb11 ffmpeg stuff 2025-03-27 13:27:21 -04:00
68e5d9fc3a remove unneeded functions 2025-03-24 17:04:58 -04:00
70354a4111 ThinGridData: avoid clones 2025-03-24 17:04:21 -04:00
56f3eae156 Model: rename grids field 2025-03-24 17:03:26 -04:00
5d574c9674 Model: rely on draining image data 2025-03-24 17:01:18 -04:00
b62745adec nits 2025-03-24 16:59:39 -04:00
b7f44d9ac0 simplify Model::run 2025-03-24 16:46:42 -04:00
14 changed files with 1631 additions and 561 deletions

3
.gitignore vendored
View File

@@ -1,6 +1,5 @@
/target
/tmp
/test.mp4
/output.mp4
/.vscode
# CLion

1137
Cargo.lock generated

File diff suppressed because it is too large Load Diff

View File

@@ -5,14 +5,21 @@ authors = ["Simon Gardling <titaniumtown@gmail.com>", "mindv0rtex <mindv0rtex@us
edition = "2021"
[dependencies]
image = "0.23"
indicatif = { version = "0.15", features = [ "rayon" ] }
itertools = "0.10"
rand = "0.8"
rand_distr = "0.4"
image = "0.25"
indicatif = { version = "0.17", features = [ "rayon" ] }
itertools = "0.14"
rand = "0.9"
rand_distr = "0.5"
rayon = "1.10"
fastapprox = "0.3"
[dev-dependencies]
criterion = { version = "0.5", features = ["html_reports"] }
[[bench]]
name = "benchmark"
harness = false
[profile.release]
codegen-units = 1
opt-level = 3

107
benches/benchmark.rs Normal file
View File

@@ -0,0 +1,107 @@
use criterion::{criterion_group, criterion_main, BatchSize, BenchmarkId, Criterion};
use physarum::{
agent::Agent,
grid::{combine, Grid},
model,
};
use rand::{rngs::StdRng, SeedableRng};
// Benchmark agent movement and deposition
fn agent_benchmark(c: &mut Criterion) {
let mut group = c.benchmark_group("Agent Tick");
let n_agents = [1_000, 10_000, 100_000];
for &n in &n_agents {
group.bench_with_input(BenchmarkId::from_parameter(n), &n, |b, &n| {
let mut rng = StdRng::seed_from_u64(42);
let agents = (0..n).map(|_| Agent::new(256, 256, &mut rng)).collect();
let mut grid = Grid::new(256, 256, &mut rng, agents);
b.iter(|| {
grid.tick();
});
});
}
group.finish();
}
// Benchmark grid diffusion (blur)
fn diffusion_benchmark(c: &mut Criterion) {
let mut group = c.benchmark_group("Grid Diffusion");
let sizes = [(256, 256), (512, 512)];
let radii = [1, 3];
for &(w, h) in &sizes {
for &r in &radii {
group.bench_with_input(
BenchmarkId::new("diffuse", format!("{}x{}_r{}", w, h, r)),
&(w, h, r),
|b, &(w, h, r)| {
b.iter_batched(
|| {
let mut rng = StdRng::seed_from_u64(42);
Grid::new(w, h, &mut rng, vec![])
},
|mut grid| grid.diffuse(r),
BatchSize::SmallInput,
);
},
);
}
}
group.finish();
}
// Benchmark grid combining
fn combine_benchmark(c: &mut Criterion) {
let mut group = c.benchmark_group("Combine Grids");
let populations = [2, 4];
for &np in &populations {
group.bench_with_input(BenchmarkId::from_parameter(np), &np, |b, &np| {
b.iter_batched(
|| {
let mut rng = StdRng::seed_from_u64(42);
let grids = (0..np)
.map(|_| Grid::new(256, 256, &mut rng, vec![]))
.collect::<Vec<_>>();
let attraction_table = vec![vec![1.0; np]; np];
(grids, attraction_table)
},
|(mut grids, table)| combine(&mut grids, &table),
BatchSize::SmallInput,
);
});
}
group.finish();
}
// Benchmark full model step
fn model_step_benchmark(c: &mut Criterion) {
let mut group = c.benchmark_group("Model Step");
let params = [(256, 256, 2), (512, 512, 4)];
for &(w, h, np) in &params {
group.bench_with_input(
BenchmarkId::new("step", format!("{}x{}_p{}", w, h, np)),
&(w, h, np),
|b, &(w, h, np)| {
b.iter_batched(
|| model::Model::new(w, h, 1 << 16, np, 1),
|mut model| model.step(),
BatchSize::SmallInput,
);
},
);
}
group.finish();
}
criterion_group!(
benches,
agent_benchmark,
diffusion_benchmark,
combine_benchmark,
model_step_benchmark
);
criterion_main!(benches);

View File

@@ -1,18 +1,17 @@
use crate::grid::PopulationConfig;
use crate::{buffer::Buf, util::wrap};
use fastapprox::faster::{cos, sin};
use rand::{seq::SliceRandom, Rng};
use rand::prelude::IndexedRandom;
use rand::Rng;
use std::f32::consts::TAU;
use std::fmt::{Display, Formatter};
// A single Physarum agent. The x and y positions are continuous, hence we use floating point numbers instead of integers.
/// A single Physarum agent. The x and y positions are continuous, hence we use floating point numbers instead of integers.
#[derive(Debug, Clone, PartialEq)]
pub struct Agent {
pub x: f32,
pub y: f32,
pub angle: f32,
pub population_id: usize,
pub i: usize,
heading: f32,
}
impl Display for Agent {
@@ -22,46 +21,28 @@ impl Display for Agent {
}
impl Agent {
// Construct a new agent with random parameters.
pub fn new<R: Rng + ?Sized>(
width: usize,
height: usize,
id: usize,
rng: &mut R,
i: usize,
) -> Self {
let (x, y, angle) = rng.gen::<(f32, f32, f32)>();
/// Construct a new agent with random parameters.
pub fn new<R: Rng + ?Sized>(width: usize, height: usize, rng: &mut R) -> Self {
let (x, y, angle) = rng.random::<(f32, f32, f32)>();
Agent {
x: x * width as f32,
y: y * height as f32,
angle: angle * TAU,
population_id: id,
i,
heading: angle * TAU,
}
}
// Tick an agent
#[inline]
pub fn tick(
&mut self,
buf: &Buf,
sensor_distance: f32,
sensor_angle: f32,
rotation_angle: f32,
step_distance: f32,
width: usize,
height: usize,
) {
let xc = self.x + cos(self.angle) * sensor_distance;
let yc = self.y + sin(self.angle) * sensor_distance;
/// Tick an agent
pub fn tick(&mut self, buf: &Buf, pop_config: PopulationConfig, width: usize, height: usize) {
let xc = self.x + cos(self.heading) * pop_config.sensor_distance;
let yc = self.y + sin(self.heading) * pop_config.sensor_distance;
let agent_add_sens = self.angle + sensor_angle;
let agent_sub_sens = self.angle - sensor_angle;
let agent_add_sens = self.heading + pop_config.sensor_angle;
let agent_sub_sens = self.heading - pop_config.sensor_angle;
let xl = self.x + cos(agent_sub_sens) * sensor_distance;
let yl = self.y + sin(agent_sub_sens) * sensor_distance;
let xr = self.x + cos(agent_add_sens) * sensor_distance;
let yr = self.y + sin(agent_add_sens) * sensor_distance;
let xl = self.x + cos(agent_sub_sens) * pop_config.sensor_distance;
let yl = self.y + sin(agent_sub_sens) * pop_config.sensor_distance;
let xr = self.x + cos(agent_add_sens) * pop_config.sensor_distance;
let yr = self.y + sin(agent_add_sens) * pop_config.sensor_distance;
// We sense from the buffer because this is where we previously combined data from all the grid.
let center = buf.get_buf(xc, yc);
@@ -69,23 +50,30 @@ impl Agent {
let right = buf.get_buf(xr, yr);
// Rotate and move logic
let mut rng = rand::thread_rng();
let mut direction: f32 = 0.0;
if (center > left) && (center > right) {
direction = 0.0;
let direction = if (center > left) && (center > right) {
0.0
} else if (center < left) && (center < right) {
direction = *[-1.0, 1.0].choose(&mut rng).unwrap();
*[-1.0, 1.0]
.choose(&mut rand::rng())
.expect("unable to choose random direction")
} else if left < right {
direction = 1.0;
1.0
} else if right < left {
direction = -1.0;
}
-1.0
} else {
0.0
};
let delta_angle = rotation_angle * direction;
let delta_angle = pop_config.rotation_angle * direction;
self.angle = wrap(self.angle + delta_angle, TAU);
self.x = wrap(self.x + step_distance * cos(self.angle), width as f32);
self.y = wrap(self.y + step_distance * sin(self.angle), height as f32);
self.heading = wrap(self.heading + delta_angle, TAU);
self.x = wrap(
self.x + pop_config.step_distance * cos(self.heading),
width as f32,
);
self.y = wrap(
self.y + pop_config.step_distance * sin(self.heading),
height as f32,
);
}
}

View File

@@ -1,19 +1,11 @@
use itertools::multizip;
use rayon::prelude::*;
#[derive(Debug)]
#[derive(Debug, Clone)]
pub struct Blur {
row_buffer: Vec<f32>,
}
impl Clone for Blur {
fn clone(&self) -> Blur {
Blur {
row_buffer: self.row_buffer.clone(),
}
}
}
impl Blur {
pub fn new(width: usize) -> Self {
Blur {
@@ -21,7 +13,7 @@ impl Blur {
}
}
// Blur an image with 2 box filter passes. The result will be written to the src slice, while the buf slice is used as a scratch space.
/// Blur an image with 2 box filter passes. The result will be written to the src slice, while the buf slice is used as a scratch space.
pub fn run(
&mut self,
src: &mut [f32],
@@ -36,23 +28,22 @@ impl Blur {
self.box_blur(src, buf, width, height, boxes[1], decay);
}
// Approximate 1D Gaussian filter of standard deviation sigma with N box filter passes. Each element in the output array contains the radius of the box filter for the corresponding pass.
/// Approximate 1D Gaussian filter of standard deviation sigma with N box filter passes. Each element in the output array contains the radius of the box filter for the corresponding pass.
fn boxes_for_gaussian<const N: usize>(sigma: f32) -> [usize; N] {
let w_ideal = (12.0 * sigma * sigma / N as f32 + 1.0).sqrt();
let sigma_sq = sigma.powi(2);
let w_ideal = (12.0 * sigma_sq / N as f32 + 1.0).sqrt();
let mut w = w_ideal as usize;
w -= 1 - (w & 1);
let mut m = 0.25 * (N * (w + 3)) as f32;
m -= 3.0 * sigma * sigma / (w + 1) as f32;
let m = m.round() as usize;
let m = (0.25 * (N * (w + 3)) as f32 - 3.0 * sigma_sq / (w + 1) as f32).round() as usize;
let mut result = [0; N];
for (i, value) in result.iter_mut().enumerate() {
*value = (if i < m { w - 1 } else { w + 1 }) / 2;
}
result
(0..N)
.map(|i| (w + 1 - 2 * (i < m) as usize) / 2)
.collect::<Vec<_>>()
.try_into()
.unwrap()
}
// Perform one pass of the 2D box filter of the given radius. The result will be written to the src slice, while the buf slice is used as a scratch space.
/// Perform one pass of the 2D box filter of the given radius. The result will be written to the src slice, while the buf slice is used as a scratch space.
fn box_blur(
&mut self,
src: &mut [f32],
@@ -66,7 +57,7 @@ impl Blur {
self.box_blur_v(buf, src, width, height, radius, decay);
}
// Perform one pass of the 1D box filter of the given radius along x axis.
/// Perform one pass of the 1D box filter of the given radius along x axis.
fn box_blur_h(&mut self, src: &[f32], dst: &mut [f32], width: usize, radius: usize) {
let weight = 1.0 / (2 * radius + 1) as f32;
@@ -75,7 +66,7 @@ impl Blur {
.for_each(|(src_row, dst_row)| {
// First we build a value for the beginning of each row. We assume periodic boundary conditions, so we need to push the left index to the opposite side of the row.
let width_sub_radius = width - radius;
let mut value = src_row[width - radius - 1];
let mut value = src_row[width_sub_radius - 1];
for j in 0..radius {
value += src_row[width_sub_radius + j] + src_row[j];
}
@@ -89,7 +80,7 @@ impl Blur {
})
}
// Perform one pass of the 1D box filter of the given radius along y axis. Applies the decay factor to the destination buffer.
/// Perform one pass of the 1D box filter of the given radius along y axis. Applies the decay factor to the destination buffer.
fn box_blur_v(
&mut self,
src: &[f32],
@@ -285,7 +276,7 @@ mod tests {
0.494_753_96,
];
for (v1, v2) in dst.iter().zip(sol) {
assert!((v1 - v2).abs() < 1e-6);
assert!((v1 - v2).abs() < 1e-6, "box_blur_h failure");
}
blur.box_blur_v(&src, &mut dst, width, height, 1, 1.0);
@@ -356,7 +347,7 @@ mod tests {
0.672_591_45,
];
for (v1, v2) in dst.iter().zip(sol) {
assert!((v1 - v2).abs() < 1e-6);
assert!((v1 - v2).abs() < 1e-6, "box_blur_v failure");
}
blur.box_blur(&mut src, &mut dst, width, height, 1, 1.0);
@@ -427,7 +418,7 @@ mod tests {
0.538_112_16,
];
for (v1, v2) in src.iter().zip(sol) {
assert!((v1 - v2).abs() < 1e-6);
assert!((v1 - v2).abs() < 1e-6, "box_blur failure");
}
}
@@ -442,4 +433,129 @@ mod tests {
let boxes = Blur::boxes_for_gaussian::<3>(2.5);
assert_eq!(boxes, [2, 2, 2]);
}
#[test]
fn total_blur_test() {
let height = 10;
let width = 10;
let mut src = (1..=(height * width))
.map(|i| (i as f32).recip())
.collect::<Vec<_>>();
let mut blur = Blur::new(width);
blur.run(
&mut src,
&mut vec![0.0; width * height],
width,
height,
2_f32,
0.1,
);
let sol = vec![
0.050528992,
0.044103604,
0.038919702,
0.032494307,
0.027310405,
0.020885015,
0.023104476,
0.020885015,
0.023104476,
0.028288381,
0.043704934,
0.038152207,
0.033674292,
0.028121557,
0.023643643,
0.018090911,
0.020009955,
0.018090911,
0.020009955,
0.024487872,
0.03968891,
0.03461781,
0.03053501,
0.025463907,
0.021381106,
0.016310005,
0.018066125,
0.016310005,
0.018066125,
0.022148928,
0.032864854,
0.028666414,
0.025289603,
0.021091158,
0.017714344,
0.013515903,
0.014971604,
0.013515901,
0.014971604,
0.01834842,
0.02884883,
0.025132021,
0.022150321,
0.018433508,
0.015451807,
0.011734996,
0.013027772,
0.011734993,
0.013027772,
0.016009476,
0.022024775,
0.019180624,
0.016904911,
0.014060758,
0.011785044,
0.008940893,
0.009933252,
0.00894089,
0.009933252,
0.012208968,
0.02513346,
0.021875666,
0.019268055,
0.016010256,
0.013402643,
0.010144847,
0.011281048,
0.010144845,
0.011281048,
0.013888664,
0.022024775,
0.019180622,
0.016904911,
0.014060758,
0.011785044,
0.008940893,
0.009933252,
0.00894089,
0.009933252,
0.012208967,
0.02513346,
0.021875666,
0.019268055,
0.016010256,
0.013402643,
0.010144847,
0.011281048,
0.010144845,
0.011281048,
0.013888664,
0.029149484,
0.02541006,
0.022407336,
0.018667907,
0.015665181,
0.011925754,
0.013224879,
0.011925753,
0.013224879,
0.016227607,
];
for (v1, v2) in src.iter().zip(sol) {
assert!((v1 - v2).abs() < 1e-6, "run failure {} vs {}", v1, v2);
}
}
}

View File

@@ -1,24 +1,25 @@
#[derive(Debug, Clone)]
pub struct Buf {
pub width: usize,
pub height: usize,
width: usize,
height: usize,
pub buf: Vec<f32>,
}
impl Buf {
pub const fn new(width: usize, height: usize, buf: Vec<f32>) -> Self {
Buf { width, height, buf }
pub fn new(width: usize, height: usize) -> Self {
Buf {
width,
height,
buf: vec![0.0; height * width],
}
}
// Truncate x and y and return a corresponding index into the data slice.
fn index(&self, x: f32, y: f32) -> usize {
// x/y can come in negative, hence we shift them by width/height.
let i = (x + self.width as f32) as usize & (self.width - 1);
let j = (y + self.height as f32) as usize & (self.height - 1);
j * self.width + i
/// Truncate x and y and return a corresponding index into the data slice.
const fn index(&self, x: f32, y: f32) -> usize {
crate::util::index(self.width, self.height, x, y)
}
// Get the buffer value at a given position. The implementation effectively treats data as periodic, hence any finite position will produce a value.
/// Get the buffer value at a given position. The implementation effectively treats data as periodic, hence any finite position will produce a value.
pub fn get_buf(&self, x: f32, y: f32) -> f32 {
self.buf[self.index(x, y)]
}

View File

@@ -1,10 +1,10 @@
use crate::{agent::Agent, blur::Blur, buffer::Buf};
use rand::{distributions::Uniform, Rng};
use rand::Rng;
use rand_distr::Uniform;
use rayon::{iter::ParallelIterator, prelude::*};
use std::fmt::{Display, Formatter};
// A population configuration.
/// A population configuration.
#[derive(Debug, Clone, Copy)]
pub struct PopulationConfig {
pub sensor_distance: f32,
@@ -12,7 +12,6 @@ pub struct PopulationConfig {
pub sensor_angle: f32,
pub rotation_angle: f32,
decay_factor: f32,
deposition_amount: f32,
}
@@ -23,33 +22,14 @@ impl Display for PopulationConfig {
}
impl PopulationConfig {
const SENSOR_ANGLE_MIN: f32 = 0.0;
const SENSOR_ANGLE_MAX: f32 = 120.0;
const SENSOR_DISTANCE_MIN: f32 = 0.0;
const SENSOR_DISTANCE_MAX: f32 = 64.0;
const ROTATION_ANGLE_MIN: f32 = 0.0;
const ROTATION_ANGLE_MAX: f32 = 120.0;
const STEP_DISTANCE_MIN: f32 = 0.2;
const STEP_DISTANCE_MAX: f32 = 2.0;
const DEPOSITION_AMOUNT_MIN: f32 = 5.0;
const DEPOSITION_AMOUNT_MAX: f32 = 5.0;
const DECAY_FACTOR_MIN: f32 = 0.1;
const DECAY_FACTOR_MAX: f32 = 0.1;
// Construct a random configuration.
/// Construct a random configuration.
pub fn new<R: Rng + ?Sized>(rng: &mut R) -> Self {
PopulationConfig {
sensor_distance: rng.gen_range(Self::SENSOR_DISTANCE_MIN..=Self::SENSOR_DISTANCE_MAX),
step_distance: rng.gen_range(Self::STEP_DISTANCE_MIN..=Self::STEP_DISTANCE_MAX),
decay_factor: rng.gen_range(Self::DECAY_FACTOR_MIN..=Self::DECAY_FACTOR_MAX),
sensor_angle: rng
.gen_range(Self::SENSOR_ANGLE_MIN..=Self::SENSOR_ANGLE_MAX)
.to_radians(),
rotation_angle: rng
.gen_range(Self::ROTATION_ANGLE_MIN..=Self::ROTATION_ANGLE_MAX)
.to_radians(),
deposition_amount: rng
.gen_range(Self::DEPOSITION_AMOUNT_MIN..=Self::DEPOSITION_AMOUNT_MAX),
sensor_distance: rng.random_range(0.0..=64.0),
step_distance: rng.random_range(0.2..=2.0),
sensor_angle: rng.random_range(0.0_f32..=120.0).to_radians(),
rotation_angle: rng.random_range(0.0_f32..=120.0).to_radians(),
deposition_amount: rng.random_range(5.0..=5.0),
}
}
}
@@ -64,24 +44,21 @@ pub struct Grid {
pub data: Vec<f32>,
// Scratch space for the blur operation.
// pub buf: Vec<f32>,
pub buf: Buf,
pub blur: Blur,
buf: Buf,
blur: Blur,
pub agents: Vec<Agent>,
}
impl Grid {
// Create a new grid filled with random floats in the [0.0..1.0) range.
/// Create a new grid filled with random floats in the [0.0..1.0) range.
pub fn new<R: Rng + ?Sized>(
width: usize,
height: usize,
rng: &mut R,
agents: Vec<Agent>,
) -> Self {
if !width.is_power_of_two() || !height.is_power_of_two() {
panic!("Grid dimensions must be a power of two.");
}
let range = Uniform::from(0.0..1.0);
let range = Uniform::new(0.0, 1.0).expect("unable to create uniform distr");
let data = rng.sample_iter(range).take(width * height).collect();
Grid {
@@ -89,34 +66,18 @@ impl Grid {
height,
data,
config: PopulationConfig::new(rng),
buf: Buf::new(width, height, vec![0.0; width * height]),
buf: Buf::new(width, height),
blur: Blur::new(width),
agents,
}
}
// Truncate x and y and return a corresponding index into the data slice.
fn index(&self, x: f32, y: f32) -> usize {
// x/y can come in negative, hence we shift them by width/height.
let i = (x + self.width as f32) as usize & (self.width - 1);
let j = (y + self.height as f32) as usize & (self.height - 1);
j * self.width + i
/// Truncate x and y and return a corresponding index into the data slice.
const fn index(&self, x: f32, y: f32) -> usize {
crate::util::index(self.width, self.height, x, y)
}
/*
// Get the buffer value at a given position. The implementation effectively treats data as periodic, hence any finite position will produce a value.
pub fn get_buf(&self, x: f32, y: f32) -> f32 {
self.buf.buf[self.index(x, y)]
}
*/
// Add a value to the grid data at a given position.
pub fn deposit(&mut self, x: f32, y: f32) {
let idx = self.index(x, y);
self.data[idx] += self.config.deposition_amount;
}
// Diffuse grid data and apply a decay multiplier.
/// Diffuse grid data and apply a decay multiplier.
pub fn diffuse(&mut self, radius: usize) {
self.blur.run(
&mut self.data,
@@ -124,64 +85,23 @@ impl Grid {
self.width,
self.height,
radius as f32,
self.config.decay_factor,
0.1, // decay is always 0.1
);
}
#[inline]
pub fn tick(&mut self) {
let (width, height) = (self.width, self.height);
let PopulationConfig {
sensor_distance,
sensor_angle,
rotation_angle,
step_distance,
..
} = self.config;
let buf = self.buf.clone();
self.agents.par_iter_mut().for_each(|agent| {
agent.tick(
&buf,
sensor_distance,
sensor_angle,
rotation_angle,
step_distance,
width,
height,
);
agent.tick(&self.buf, self.config, self.width, self.height);
});
self.deposit_all();
}
#[inline]
pub fn deposit_all(&mut self) {
let agent_list = self.agents.clone();
for agent in agent_list.iter() {
self.deposit(agent.x, agent.y);
for agent in self.agents.iter() {
let idx = self.index(agent.x, agent.y);
self.data[idx] += self.config.deposition_amount;
}
}
// No longer needed (moved to imgdata.rs)
/*
pub fn quantile(&self, fraction: f32) -> f32 {
let index = if (fraction - 1.0_f32).abs() < f32::EPSILON {
self.data.len() - 1
} else {
(self.data.len() as f32 * fraction) as usize
};
let mut sorted = self.data.clone();
sorted
.as_mut_slice()
.select_nth_unstable_by(index, |a, b| a.partial_cmp(b).unwrap());
sorted[index]
}
pub fn data(&self) -> &[f32] {
&self.data
}
*/
}
pub fn combine<T>(grids: &mut [Grid], attraction_table: &[T])
@@ -192,14 +112,16 @@ where
let bufs: Vec<_> = grids.iter().map(|grid| &grid.buf.buf).collect();
// We mutate grid buffers and read grid data. We use unsafe because we need shared/unique borrows on different fields of the same Grid struct.
bufs.iter().enumerate().for_each(|(i, buf)| unsafe {
bufs.iter().enumerate().for_each(|(i, buf)| {
let buf_ptr = *buf as *const Vec<f32> as *mut Vec<f32>;
buf_ptr.as_mut().unwrap().fill(0.0);
// SAFETY! we can take these are raw pointers because we are
// getting it from a `&mut [Grid]`
let buf_ptr_mut = unsafe { buf_ptr.as_mut().unwrap_unchecked() };
buf_ptr_mut.fill(0.0);
datas.iter().enumerate().for_each(|(j, other)| {
let multiplier = attraction_table[i].as_ref()[j];
buf_ptr
.as_mut()
.unwrap()
buf_ptr_mut
.iter_mut()
.zip(*other)
.for_each(|(to, from)| *to += from * multiplier)
@@ -211,16 +133,9 @@ where
mod tests {
use super::*;
#[test]
#[should_panic]
fn test_grid_new_panics() {
let mut rng = rand::thread_rng();
let _ = Grid::new(5, 5, &mut rng, vec![]);
}
#[test]
fn test_grid_new() {
let mut rng = rand::thread_rng();
let mut rng = rand::rng();
let grid = Grid::new(8, 8, &mut rng, vec![]);
assert_eq!(grid.index(0.5, 0.6), 0);
assert_eq!(grid.index(1.5, 0.6), 1);

View File

@@ -1,41 +1,30 @@
use crate::{grid::Grid, palette::Palette};
use image::RgbImage;
use itertools::multizip;
/// Stores data that is located in grids that is used for image generation
#[derive(Clone)]
pub struct ThinGridData {
pub width: usize,
pub height: usize,
pub data: Vec<f32>,
width: usize,
height: usize,
data: Vec<f32>,
}
impl ThinGridData {
// Convert Grid to ThinGridData
pub fn new_from_grid(in_grid: &Grid) -> Self {
/// Convert Grid to ThinGridData
pub fn new_from_grid(in_grid: Grid) -> Self {
ThinGridData {
width: in_grid.width,
height: in_grid.height,
data: in_grid.data.clone(),
data: in_grid.data,
}
}
#[allow(dead_code)]
pub fn new_from_grid_vec(in_grids: Vec<Grid>) -> Vec<Self> {
in_grids
.iter()
.map(|grid| Self::new_from_grid(grid))
.collect()
pub fn new_from_grid_vec(in_grids: &[Grid]) -> Vec<Self> {
in_grids.iter().cloned().map(Self::new_from_grid).collect()
}
// from grid.rs (needed in image gen)
#[allow(dead_code)]
pub fn data(&self) -> &[f32] {
&self.data
}
// from grid.rs (needed in image gen)
#[allow(dead_code)]
/// from grid.rs (needed in image gen)
pub fn quantile(&self, fraction: f32) -> f32 {
let index = if (fraction - 1.0_f32).abs() < f32::EPSILON {
self.data.len() - 1
@@ -48,46 +37,24 @@ impl ThinGridData {
.select_nth_unstable_by(index, |a, b| a.partial_cmp(b).unwrap());
sorted[index]
}
pub fn size_of(&self) -> usize {
let mut output: usize = 0;
output += std::mem::size_of_val(&self.width);
output += std::mem::size_of_val(&self.height);
for i in self.data.iter() {
output += std::mem::size_of_val(&i);
}
output
}
}
/// Class for storing data that will be used to create images
#[derive(Clone)]
pub struct ImgData {
pub grids: Vec<ThinGridData>,
pub palette: Palette,
pub iteration: i32,
grids: Vec<ThinGridData>,
palette: Palette,
}
impl ImgData {
pub fn new(in_grids: Vec<ThinGridData>, in_palette: Palette, in_iteration: i32) -> Self {
pub const fn new(in_grids: Vec<ThinGridData>, in_palette: Palette) -> Self {
ImgData {
grids: in_grids,
palette: in_palette,
iteration: in_iteration,
}
}
pub fn size_of(&self) -> usize {
let mut output: usize = 0;
output += std::mem::size_of_val(&self.iteration);
output += std::mem::size_of_val(&self.palette);
for grid in self.grids.iter() {
output += grid.size_of();
}
output
}
pub fn save_to_image(&self) {
pub fn to_image(&self) -> RgbImage {
let (width, height) = (self.grids[0].width, self.grids[0].height);
let mut img = image::RgbImage::new(width as u32, height as u32);
@@ -104,7 +71,7 @@ impl ImgData {
for (grid, max_value, color) in
multizip((&self.grids, &max_values, &self.palette.colors))
{
let mut t = (grid.data()[i] / max_value).clamp(0.0, 1.0);
let mut t = (grid.data[i] / max_value).clamp(0.0, 1.0);
t = t.powf(1.0 / 2.2); // gamma correction
r += color.0[0] as f32 * t;
g += color.0[1] as f32 * t;
@@ -116,8 +83,6 @@ impl ImgData {
img.put_pixel(x as u32, y as u32, image::Rgb([r as u8, g as u8, b as u8]));
}
}
img.save(format!("./tmp/out_{}.png", self.iteration).as_str())
.unwrap();
img
}
}

View File

@@ -1,8 +1,8 @@
mod agent;
pub mod agent;
mod blur;
mod buffer;
mod grid;
mod imgdata; // for storing image data
pub mod grid;
pub mod imgdata; // for storing image data
pub mod model;
mod palette;
mod util; // for math things

View File

@@ -1,39 +1,57 @@
use physarum::model;
use physarum::{
imgdata::{ImgData, ThinGridData},
model,
};
use std::io::Write;
fn main() {
// # of iterations to go through
let n_iterations = 1024;
// let n_iterations = 2048;
// let n_iterations = 1 << 14;
// Size of grid and pictures
// let (width, height) = (256, 256);
// let (width, height) = (512, 512);
let (width, height) = (1024, 1024);
// # of agents
// let n_particles = 1 << 10;
// let n_particles = 1 << 16;
// let n_particles = 1 << 20;
let n_particles = 1 << 24;
println!("n_particles: {}", n_particles);
let n_particles = 1 << 22;
let diffusivity = 1;
let n_populations = 3;
// `n_populations` is the # of types of agents
// let n_populations = 4;
let n_populations = 1;
// let n_populations = 1 + rng.gen_range(1..4); // make # of populations between 2 and 5
let mut model = model::Model::new(width, height, n_particles, n_populations, diffusivity);
model.print_configurations();
let mut model = model::Model::new(width, height, n_particles, n_populations, diffusivity); // Create the model
// Setup ffmpeg
let mut ffmpeg = std::process::Command::new("ffmpeg")
.args([
"-y",
"-f",
"rawvideo",
"-pix_fmt",
"rgb24",
"-s",
&format!("{}x{}", width, height),
"-r",
"30",
"-i",
"-",
"-c:v",
"libsvtav1",
"output.mp4",
])
.stdin(std::process::Stdio::piped())
.spawn()
.expect("Failed to start ffmpeg");
let mut stdin = ffmpeg.stdin.take().unwrap();
model.print_configurations(); // Print config for model
for _ in 0..n_iterations {
model.step();
model.run(n_iterations); // Actually run the model
// Generate image
let grids = ThinGridData::new_from_grid_vec(model.population_grids());
let img_data = ImgData::new(grids, model.palette());
let img = img_data.to_image();
let raw_data = img.into_raw();
// export saved image data
println!("Rendering all saved image data....");
model.render_all_imgdata();
model.flush_image_data();
// Write to ffmpeg
stdin.write_all(&raw_data).unwrap();
}
// Cleanup
drop(stdin);
ffmpeg.wait().unwrap();
println!("Done!");
}

View File

@@ -1,35 +1,32 @@
use crate::{
agent::Agent,
grid::{combine, Grid},
imgdata::{ImgData, ThinGridData},
palette::{random_palette, Palette},
};
use indicatif::{ParallelProgressIterator, ProgressBar, ProgressStyle};
// use rand::Rng;
use indicatif::{ProgressBar, ProgressStyle};
use rand_distr::{Distribution, Normal};
use rayon::{iter::ParallelIterator, prelude::*};
use std::{path::Path, time::Instant};
use std::time::Instant;
// Top-level simulation class.
/// Top-level simulation class.
pub struct Model {
// The grid they move on.
grids: Vec<Grid>,
/// per-population grid (one for each population)
population_grids: Vec<Grid>,
// Attraction table governs interaction across populations
/// Attraction table governs interaction across populations
attraction_table: Vec<Vec<f32>>,
// Global grid diffusivity.
/// Global grid diffusivity.
diffusivity: usize,
// Current model iteration.
iteration: i32,
/// Current model iteration.
iteration: usize,
// Color palette
/// Color palette
palette: Palette,
// List of ImgData to be processed post-simulation into images
img_data_vec: Vec<ImgData>,
time_per_agent_list: Vec<f64>,
time_per_step_list: Vec<f64>,
}
impl Model {
@@ -39,13 +36,13 @@ impl Model {
const REPULSION_FACTOR_STD: f32 = 0.1;
pub fn print_configurations(&self) {
for (i, grid) in self.grids.iter().enumerate() {
for (i, grid) in self.population_grids.iter().enumerate() {
println!("Grid {}: {}", i, grid.config);
}
println!("Attraction table: {:#?}", self.attraction_table);
}
// Construct a new model with random initial conditions and random configuration.
/// Construct a new model with random initial conditions and random configuration.
pub fn new(
width: usize,
height: usize,
@@ -56,195 +53,92 @@ impl Model {
let particles_per_grid = (n_particles as f64 / n_populations as f64).ceil() as usize;
let _n_particles = particles_per_grid * n_populations;
let mut rng = rand::thread_rng();
let mut rng = rand::rng();
let attraction_distr =
Normal::new(Self::ATTRACTION_FACTOR_MEAN, Self::ATTRACTION_FACTOR_STD).unwrap();
let repulstion_distr =
let repulsion_distr =
Normal::new(Self::REPULSION_FACTOR_MEAN, Self::REPULSION_FACTOR_STD).unwrap();
let mut attraction_table = Vec::with_capacity(n_populations);
for i in 0..n_populations {
attraction_table.push(Vec::with_capacity(n_populations));
for j in 0..n_populations {
attraction_table[i].push(if i == j {
attraction_distr.sample(&mut rng)
} else {
repulstion_distr.sample(&mut rng)
});
attraction_table[i].push(
if i == j {
&attraction_distr
} else {
&repulsion_distr
}
.sample(&mut rng),
);
}
}
let mut grids: Vec<Grid> = Vec::new();
for pop in 0..n_populations {
let agents = (0..particles_per_grid)
.map(|i| Agent::new(width, height, pop, &mut rng, i))
.collect();
grids.push(Grid::new(width, height, &mut rng, agents));
}
let grids = (0..n_populations)
.map(|_| {
let agents = (0..particles_per_grid)
.map(|_| Agent::new(width, height, &mut rng))
.collect();
Grid::new(width, height, &mut rng, agents)
})
.collect();
Model {
grids,
population_grids: grids,
attraction_table,
diffusivity,
iteration: 0,
palette: random_palette(),
img_data_vec: Vec::new(),
time_per_agent_list: Vec::new(),
time_per_step_list: Vec::new(),
}
}
// Simulates `steps` # of steps
#[inline]
pub fn step(&mut self) {
combine(&mut self.population_grids, &self.attraction_table);
let agents_tick_time = Instant::now();
self.population_grids.par_iter_mut().for_each(|grid| {
grid.tick();
grid.diffuse(self.diffusivity);
});
let agents_tick_elapsed = agents_tick_time.elapsed().as_millis() as f64;
let agents_num: usize = self.population_grids.iter().map(|g| g.agents.len()).sum();
let ms_per_agent = agents_tick_elapsed / agents_num as f64;
self.time_per_agent_list.push(ms_per_agent);
self.time_per_step_list.push(agents_tick_elapsed);
self.iteration += 1;
}
pub fn run(&mut self, steps: usize) {
let debug: bool = false;
let pb = ProgressBar::new(steps as u64);
pb.set_style(
ProgressStyle::default_bar()
.template(
"{spinner:.green} [{elapsed_precise}] [{bar:40.cyan/blue}] {pos}/{len} ({eta} {percent}%, {per_sec})",
)
.progress_chars("#>-"),
);
pb.set_style(ProgressStyle::default_bar()
.template("{spinner:.green} [{elapsed_precise}] [{bar:40.cyan/blue}] {pos}/{len} ({eta} {percent}%, {per_sec})").expect("invalid progresstyle template")
.progress_chars("#>-"));
let mut time_per_agent_list: Vec<f64> = Vec::new();
let mut time_per_step_list: Vec<f64> = Vec::new();
let agents_num: usize = self.grids.iter().map(|grid| grid.agents.len()).sum();
for i in 0..steps {
if debug {
println!("Starting tick for all agents...")
};
// Combine grids
let grids = &mut self.grids;
combine(grids, &self.attraction_table);
let agents_tick_time = Instant::now();
// Tick agents
let diffusivity = self.diffusivity;
self.grids.par_iter_mut().for_each(|grid| {
grid.tick();
grid.diffuse(diffusivity); // Diffuse + Decay
});
self.save_image_data();
let agents_tick_elapsed: f64 = agents_tick_time.elapsed().as_millis() as f64;
let ms_per_agent: f64 = agents_tick_elapsed / (agents_num as f64);
time_per_agent_list.push(ms_per_agent);
time_per_step_list.push(agents_tick_elapsed);
if debug {
println!(
"Finished tick for all agents. took {}ms\nTime per agent: {}ms\n",
agents_tick_elapsed, ms_per_agent
)
};
self.iteration += 1;
pb.set_position(i as u64);
for _ in 0..steps {
self.step();
pb.inc(1);
}
pb.finish();
let avg_per_step: f64 =
time_per_step_list.iter().sum::<f64>() / time_per_step_list.len() as f64;
self.time_per_step_list.iter().sum::<f64>() / self.time_per_step_list.len() as f64;
let avg_per_agent: f64 =
time_per_agent_list.iter().sum::<f64>() / time_per_agent_list.len() as f64;
self.time_per_agent_list.iter().sum::<f64>() / self.time_per_agent_list.len() as f64;
println!(
"Average time per step: {}ms\nAverage time per agent: {}ms",
avg_per_step, avg_per_agent
);
}
fn size_of_imgdata_vec(&self) -> usize {
self.img_data_vec[0].size_of() * self.img_data_vec.len()
pub fn population_grids(&self) -> &[Grid] {
&self.population_grids
}
fn save_image_data(&mut self) {
let grids = ThinGridData::new_from_grid_vec(self.grids.clone());
let img_data = ImgData::new(grids, self.palette, self.iteration);
self.img_data_vec.push(img_data);
let size: usize = self.size_of_imgdata_vec();
let mb = size / 1024 / 1024;
// println!("{} B | {} KB | {} MB", size, size/1024, size/1024/1024);
let max_mb = 6000;
if mb >= max_mb {
println!(
"ram usage is over {} MB (and len of {}), flushing to disk\n",
max_mb,
self.img_data_vec.len()
);
self.render_all_imgdata();
self.flush_image_data();
}
pub fn palette(&self) -> Palette {
self.palette
}
pub fn flush_image_data(&mut self) {
self.img_data_vec.clear();
}
pub fn render_all_imgdata(&self) {
if !Path::new("./tmp").exists() {
std::fs::create_dir("./tmp").expect("could create directory");
}
let pb = ProgressBar::new(self.img_data_vec.len() as u64);
pb.set_style(ProgressStyle::default_bar().template(
"{spinner:.green} [{elapsed_precise}] [{bar:40.cyan/blue}] ({pos}/{len}, {percent}%, {per_sec})",
));
/*
for img in &self.img_data_vec {
Self::save_to_image(img.to_owned());
pb.inc(1);
}
pb.finish();
*/
self.img_data_vec
.par_iter()
.progress_with(pb)
.for_each(|img| {
// Self::save_to_image(img.to_owned());
img.save_to_image();
});
}
/*
pub fn save_to_image(imgdata: ImgData) {
let (width, height) = (imgdata.grids[0].width, imgdata.grids[0].height);
let mut img = image::RgbImage::new(width as u32, height as u32);
let max_values: Vec<_> = imgdata
.grids
.iter()
.map(|grid| grid.quantile(0.999) * 1.5)
.collect();
for y in 0..height {
for x in 0..width {
let i = y * width + x;
let (mut r, mut g, mut b) = (0.0_f32, 0.0_f32, 0.0_f32);
for (grid, max_value, color) in
multizip((&imgdata.grids, &max_values, &imgdata.palette.colors))
{
let mut t = (grid.data()[i] / max_value).clamp(0.0, 1.0);
t = t.powf(1.0 / 2.2); // gamma correction
r += color.0[0] as f32 * t;
g += color.0[1] as f32 * t;
b += color.0[2] as f32 * t;
}
r = r.clamp(0.0, 255.0);
g = g.clamp(0.0, 255.0);
b = b.clamp(0.0, 255.0);
img.put_pixel(x as u32, y as u32, image::Rgb([r as u8, g as u8, b as u8]));
}
}
img.save(format!("./tmp/out_{}.png", imgdata.iteration).as_str())
.unwrap();
}
*/
}

View File

@@ -1,4 +1,4 @@
use rand::{seq::SliceRandom, thread_rng, Rng};
use rand::{seq::SliceRandom, Rng};
#[derive(Clone, Copy)]
pub struct Palette {
@@ -6,8 +6,8 @@ pub struct Palette {
}
pub fn random_palette() -> Palette {
let mut rng = thread_rng();
let mut palette = PALETTES[rng.gen_range(0..PALETTES.len())];
let mut rng = rand::rng();
let mut palette = PALETTES[rng.random_range(0..PALETTES.len())];
palette.colors.shuffle(&mut rng);
palette
}

View File

@@ -1,4 +1,111 @@
#[inline]
pub fn wrap(x: f32, max: f32) -> f32 {
x - max * ((x > max) as i32 as f32 - (x < 0.0_f32) as i32 as f32)
// x - max * ((x > max) as i32 - x.is_sign_negative() as i32) as f32
x.rem_euclid(max)
}
/// Truncate x and y and return a corresponding index into the data slice.
#[inline]
pub const fn index(width: usize, height: usize, x: f32, y: f32) -> usize {
// x/y can come in negative, hence we shift them by width/height.
let i = (x + width as f32) as usize & (width - 1);
let j = (y + height as f32) as usize & (height - 1);
j * width + i
}
#[cfg(test)]
mod test {
use super::*;
mod wrap {
use super::*;
#[test]
fn over() {
assert_eq!(wrap(1.1, 1.0), 0.100000024); // floating point weirdness
}
#[test]
fn middle() {
assert_eq!(wrap(0.5, 1.0), 0.5);
}
#[test]
fn under() {
assert_eq!(wrap(-1.0, 2.0), 1.0);
}
}
mod index {
use super::*;
#[test]
fn basic_positive_coordinates() {
let width = 4;
let height = 4;
assert_eq!(index(width, height, 1.5, 2.5), 9);
}
#[test]
fn negative_x_coordinate() {
let width = 8;
let height = 8;
assert_eq!(index(width, height, -3.2, 5.6), 44);
}
#[test]
fn exact_boundary_values() {
let width = 16;
let height = 16;
assert_eq!(index(width, height, 16.0, 0.0), 0);
}
#[test]
fn large_coordinates() {
let width = 2;
let height = 2;
assert_eq!(index(width, height, 1000.0, 2000.0), 0);
}
#[test]
fn negative_x_and_y() {
let width = 4;
let height = 4;
assert_eq!(index(width, height, -1.5, -0.5), 14);
}
#[test]
fn fractional_truncation() {
let width = 4;
let height = 4;
assert_eq!(index(width, height, 3.9, 3.999), 15);
}
#[test]
fn zero_coordinates() {
let width = 4;
let height = 4;
assert_eq!(index(width, height, 0.0, 0.0), 0);
}
#[test]
fn x_equals_width() {
let width = 8;
let height = 8;
assert_eq!(index(width, height, 8.0, 0.0), 0);
}
#[test]
fn y_negative_beyond_height() {
let width = 4;
let height = 4;
assert_eq!(index(width, height, 0.0, -4.5), 0);
}
#[test]
fn width_and_height_one() {
let width = 1;
let height = 1;
assert_eq!(index(width, height, 123.4, -56.7), 0);
}
}
}