Reaction Diffusion
cellular automata reaction diffusion mouse
Move your mouse to explore the generation of different patterns.
// This work is licensed under CC BY 4.0
// https://creativecommons.org/licenses/by/4.0/
struct Sys {
time: f32,
resolution: vec2<f32>,
mouse: vec4<f32>,
aspect: vec2<f32>
};
struct Uni {
size: vec2<f32>
}
@group(0) @binding(0) var<uniform> uni: Uni;
@group(0) @binding(4) var<uniform> sys: Sys;
@group(0) @binding(1) var<storage> current: array<vec2<f32>>;
@group(0) @binding(2) var<storage, read_write> next: array<vec2<f32>>;
struct VertexInput {
@location(0) pos: vec2<f32>,
@builtin(instance_index) instance: u32
};
struct VertexOutput {
@builtin(position) pos: vec4f,
@location(0) state: vec2f
}
fn getIndex(cell : vec2<f32>) -> u32 {
let c = (cell + uni.size) % uni.size;
return u32( c.y * uni.size.x + c.x );
}
@vertex
fn vertexMain(input : VertexInput) -> VertexOutput {
let i = f32(input.instance);
let cell = vec2f(i % uni.size.x, floor(i / uni.size.x) );
// takes the avergave of the neighbours for this vertice in this state.
let c1 = current[getIndex( vec2(cell.x + input.pos.x , cell.y) )];
let c2 = current[getIndex( vec2(cell.x + input.pos.x, cell.y - input.pos.y) )];
let c3 = current[getIndex( vec2(cell.x , cell.y - input.pos.y) )];
let c4 = current[input.instance];
// multisample the state to reduce aliasing
let state = (c1 + c2 + c3 + c4) / 4.0;
let cellSize = 2. / uni.size.xy ;
// The cell(0,0) is a the top left corner of the screen.
// The cell(uni.size.x,uni.size.y) is a the bottom right corner of the screen.
let cellOffset = vec2(cell.x, uni.size.y - 1. - cell.y) * cellSize + (cellSize * .5) ;
// input.pos is in the range [-1,1]...[1,1] and it's the same coord system as the uv of the screen
let cellPos = (input.pos / uni.size.xy) + cellOffset - 1.;
var output: VertexOutput;
output.pos = vec4f(vec2(cellPos), 0., 1.); //[0.1,0.1]...[0.9,0.9] cell vertex positions
output.state = state; // the current state
return output;
}
@fragment
fn fragmentMain( input: VertexOutput) -> @location(0) vec4f {
// we use the state to control the color
let v = input.state;
let color = hsv2rgb(vec3f( abs(v.y-v.x) , 1., pow(v.y + 0.001 ,.5) ));
return vec4f( tosRGB(color), 1.);
}
@compute @workgroup_size(8, 8)
fn computeMain(@builtin(global_invocation_id) cell: vec3u) {
// keep the simulation in the range [0,size]
if (cell.x >= u32(uni.size.x) || cell.y >= u32(uni.size.y)) { return; }
// a non exhaustive list of the type of patterns we can create
// the first controls the activation, the second the inhibition
let params = array<vec2<f32>,9>(
vec2(0.046 , 0.063), // worms
vec2(0.082, 0.06), // worms and loops
vec2(0.042, 0.059), // Turing patterns
vec2(0.014, 0.054), // moving spot
vec2(0.037, 0.06), // fingerprint
vec2(0.014, 0.047), // spots and loops
vec2(0.062, 0.0609), // U-SKATE
vec2(0.039, 0.058), // holes
vec2(0.022, 0.0610) // microbes
);
// we divide the screen in 9 areas and we use
// the mouse position to select the pattern
let m = floor(sys.mouse.xy * 3);
let ai = params[u32(m.y) * 3u + u32(m.x)];
// calculate the value and store it in the next buffer
let v = rd( vec2i(cell.xy), vec2i(uni.size), ai.x, ai.y);
next[cell.y * u32(uni.size.x) + cell.x] = clamp(v, vec2(0.), vec2(1.));
// we add a small amount of B in the mouse position
let pos = vec2u(floor(sys.mouse.xy * uni.size));
let index = pos.y * u32(uni.size.x) + pos.x;
next[index].y = 1.;
}
// 9 point stencil laplace operator
const K_LAPLACE9 = array<f32,9>(.25, .5, .25, .5, -3., .5, .25, .5, .25);
// gray-scott reaction-diffusion system
fn rd( cell: vec2<i32>, size: vec2<i32>, feed:f32, decay:f32) -> vec2f {
var laplacian = vec2(0.);
for(var i = 0; i < 9; i++) {
let offset = (vec2i( (i / 3) - 1 , (i % 3) - 1 ) + cell + size) % size;
laplacian += (K_LAPLACE9[i] * current[offset.y * size.x + offset.x]);
}
// chemicals A and B are stored in the x and y component of buffer
let ab = current[ cell.y * size.x + cell.x];
// calculate the dA and dB
let da = ((.2097 * laplacian.x) - (ab.x * ab.y * ab.y)) + (feed * (1. - ab.x));
let db = ((.1050 * laplacian.y) + (ab.x * ab.y * ab.y)) - ((feed + decay) * ab.y);
// we use the 1 as the dt time step
let dt = 1.;
return ab + dt * vec2(da,db);
}
// convert a color in hsv color space to rgb
fn hsv2rgb(c :vec3f) -> vec3f {
let k = vec4f(1.0, 2.0 / 3.0, 1.0 / 3.0, 3.0);
let p = abs(fract(c.xxx + k.xyz) * 6.0 - k.www);
return c.z * mix(k.xxx, clamp(p - k.xxx, vec3(0.0), vec3(1.0)), c.y);
}
// Converts a color from linear light gamma to sRGB gamma
fn tosRGB(linearRGB: vec3f) -> vec3f {
let cutoff = vec3<bool>(linearRGB.x < 0.0031308, linearRGB.y < 0.0031308, linearRGB.z < 0.0031308);
let higher = vec3(1.055) * pow(linearRGB.rgb, vec3(1.0/2.4)) - vec3(0.055);
let lower = linearRGB.rgb * vec3(12.92);
return vec3<f32>(mix(higher, lower, vec3<f32>(cutoff)));
}
import { PSpec, Definitions, square, scaleAspect } from "../../lib/poiesis/index.ts";
export const rd = (code:string, defs:Definitions) => {
const spec = (w:number,h:number) : PSpec => {
const size = scaleAspect(w,h,512);
const current = Array.from( { length: size.x * size.y } , () => [ 1 , (Math.random() > 0.001 ? 0 : 1) ] );
return {
code: code,
defs: defs,
geometry: {
vertex: {
data: square(1.),
attributes: ["pos"],
instances: size.x * size.y
}
},
uniforms: () => ({
uni: {
size: [size.x, size.y]
}
}),
storages: [
{ name: "current", size: size.x * size.y, data: current } ,
{ name: "next", size: size.x * size.y}
],
computes: [
{ name: "computeMain", workgroups: [Math.ceil(size.x / 8), Math.ceil(size.y / 8), 1] },
],
computeGroupCount: 32,
bindings: [ [0,4,1,2], [0,4,2,1] ]
}
}
return spec
}