WebGPGPU, or WGU for short, is a WebGL2 based library enabling general purpose computation on the GPU.
const WGU = WebGPGPU(context);
new WGU.VertexFeedbackComputer({
units: 8e5,
struct: {
position: "vec2",
velocity: "vec2",
mass: "float",
color: "vec3"
},
initialize: (i, buffer) => {...}
updateStep: {
glsl: `
void main() {
o_position = i_position + i_velocity;
o_velocity = i_velocity - 0.001 * i_position / i_mass;
o_mass = i_mass;
o_color = i_color;
}
`
},
renderStep: {
glsl: `
void main() {
gl_Position = vec4(i_position, 0.0, 1.0);
vertexColor = vec4(i_color, .5);
gl_PointSize = i_mass/2.0;
}
`
}
});See this example in action here
The code above uses a VertexFeedbackComputer (a particular way of computing stuff provided by the library, relying on Vertex Buffers and OpenGL ES 3.0's Transform Feedback) to simulate a simple 800,000 particle system. Each particle has its own position, velocity, mass, and color, initially random, and is represented by a coloured point as it falls toward the origin.
Documentation is available online here or in markdown form at master/build/docs
Include the library from rawgit.com or download it locally
<script src="https://rawgit.com/npny/webgpgpu/master/src/webgpgpu.js"></script>
<script src="src/webgpgu.js"></script>In order to use WebGPGPU in your page, you then need to initialize it with an existing, valid WebGL2 context:
<canvas id="canvas"></canvas>
<script>
const canvas = document.getElementById("canvas");
const context = canvas.getContext("webgl2");
const WGU = WebGPGPU(context);
// Let's get rolling
new WGU.VertexFeedbackComputer();
</script>WebGPGPU is released under the MIT license. Pierre Boyer, 2017.