simbleau/nbody-wasm-sim

An N-body WebAssembly simulation using Web GPU

Spencer C. Imbleau (he/him)

🏠 Website | 📝 Blog | 📋 Résumé
📧 Contact methods are available on my website.

nbody-wasm-sim's Issues

Pausing

Pausing appears broken.

Canvas doesn't dynamically resize

When you resize the browser, you get distortion of the viewport.

Normally, and on page load, we get a ResizeEvent which allows us to set the canvas size:

However, on browser resize, we don't get these events anymore, leaving me unsure of how to resize the canvas and aspect ratio itself. For now I've added CSS rules on index.html to keep it fullscreen, but I'm not entirely sure how to fix this without ResizeEvents from winit.

GpuPrimitives -> GpuTypes

Let's see a GpuPrimitive and GpuUniform, which would organize our current prim folder.
For uniform, use bind() function as the interface
-- Move bind logic from render/camera
-- Move bind logic from sim/world

Anti-aliasing

Some basic anti-aliasing through WebGPU isn't a difficult ask, and should probably be done at some point.

Pretty sure the SVG renderer from Lyon uses anti-aliasing, incase anyone needs a quick reference.

Renove x2

There are two renovate files. They may be conflicting.

Compute gravity under a compute shader

A big potential speedup is waiting to happen: Calculate the gravity vector under a compute shader. Currently we are using CPU based math for the gravitational effects, which is pretty slow.

The goal: replace this with compute shader logic.. Might require significant refactoring.

Enable WebGPU over WebGL

WebGPU is unstable right now and locked behind a feature flag on Canary/Nightly.

When the API is stable, we want to replace the WebGL feature and use WebGPU.

Right now this is blocked.

Use trunk as build tool

Wasm-pack is super slow and deprecated. Let's favor trunk over wasm-pack.

Package organization

Gpu_types does not exactly supersede render, so it makes sense to organize the packages.

"Alphamode" error when compiling to WASM

error[E0063]: missing field alpha_mode in initializer of SurfaceConfiguration
--> src\render\wgpu_context.rs:57:22
|
57 | let config = wgpu::SurfaceConfiguration {
| ^^^^^^^^^^^^^^^^^^^^^^^^^^ missing alpha_mode

I got this running by adding the following line (/src/render/wgpu_context, line #63):

        alpha_mode: wgpu::CompositeAlphaMode::Opaque

Pull request?

GpuTransform should be named consistently

GpuTransform is sometimes referred to as an instance in certain places. This should be consistently one or the other.

Shaders seem to deconstruct mat4x4 unnecessarily?

In the shaders, we seem to deconstruct the mat4x4 into 4 vec4s.

Take the following example:

struct InstanceInput {
    @location(2) model_matrix_0: vec4<f32>,
    @location(3) model_matrix_1: vec4<f32>,
    @location(4) model_matrix_2: vec4<f32>,
    @location(5) model_matrix_3: vec4<f32>,
};

Could have been represented as such:

struct InstanceInput {
    @location(2) model_matrix: mat4x4<f32>,
};

But this would require we pass the blob as [f32; 4]; 4]

Try particular

You're welcome! It's a fun exercise for me as I'm currently working on a similar project using bevy and rapier so if my work can contribute to other projects than it should!

Originally posted by @Canleskis in #59 (comment)

How about maybe using Particular in this next :D

Migrate sim logic into a sim.rs

The logic for the simulation currently exists in state.rs

Seems irregular, or surprising. We should get around to migrating that appropriately.

e.g.

//sim.rs
pub fn run(&mut state) {
   ...
}

World constants

World constants are disorganized and need to be tweaked for better simulation

MIME Types not loading: GitHub Pages issue

JS/WASM do not appear to be loading correctly through GitHub Pages. This is likely how it is built with Trunk.

Frame Descriptor cleanup

Yeah this one speaks for itself. It's gotten hairy as a point of code dumping for a while. Needs love.

Wireframe broken

Introduced in #6, wireframe causes a validation error due to the vertex blob now having incorrect memory order.

Radial distribution of bodies is not uniform

We made the same mistake as https://twitter.com/keenanisalive/status/1529490555893428226

Loading modal

Since the simulation is around ~1.75MB currently, it would be nice to have a loading modal that goes away on load. This could be plopped into the index.html with a simple hook to display:none it on load.

AWESOME

This project is totally Awesome!

Physics organization

We should probably see a trait RigidBody make its way into scope.

E.g.

trait RigidBody {
    fn collides_with(other: RigidBody) -> bool;
    fn get_collision(other: RigidBody) -> Collision;
}

struct Collision {
    impact_position: Vec2,
    impact_magnitude: f32,
}

Dependency Dashboard

This issue lists Renovate updates and detected dependencies. Read the Dependency Dashboard docs to learn more.

Open

These updates have all been created already. Click a checkbox below to force a retry/rebase of any.

• Update Rust crate particular to 0.3.1
• Update Rust crate wgpu to 0.14.2

Detected dependencies

cargo

Cargo.toml

• console_error_panic_hook 0.1.7
• wasm-bindgen 0.2.83
• wasm-bindgen-futures 0.4.33
• raw-window-handle 0.5.0
• winit 0.27.5
• gloo-console 0.2.3
• instant 0.1.12
• wgpu 0.14.0
• glam 0.22.0
• bytemuck 1.12.3
• js-sys 0.3.60
• nalgebra 0.31.4
• particular 0.3.0
• rapier2d 0.16.1
• web-sys 0.3.60
• image 0.24.5

github-actions

.github/workflows/build.yml

• actions/checkout v3
• actions/cache v3
• actions/cache v3

.github/workflows/release.yml

• actions/checkout v3
• actions/cache v3
• actions/cache v3
• peaceiris/actions-gh-pages v3

---

• Check this box to trigger a request for Renovate to run again on this repository

Stream assets in as textures

We'll need a way to put in textures for the simulation.