glmw

WebAssembly powered Matrix and Vector library

Usage no npm install needed!

<script type="module">
  import glmw from 'https://cdn.skypack.dev/glmw';
</script>

README

GLMW - WebAssembly powered Matrix and Vector library


Description

This is an experimental near 1:1 port of gl-matrix v2.4.0 to WebAssembly.

Performance

In many cases glmw runs more than twice as fast as gl-matrix.

Some methods like *.str and *.equals are bridged and bring in some extra overhead.

Creating views with *.view is cheap, because they return a typed subarray of the WebAssembly module's memory buffer.

Benchmarks

Limitations

  • This library requires async instantiation, since WebAssembly has a synchronous buffer instantiation size limitation.
  • You need to manually free data, since there is no garbage collection yet (be careful! :p).
  • Methods like mat4.create and mat4.multiply return a numeric address. To get an view on your data you need to use e.g. mat4.view(address). This returns a Float32Array which is a direct view onto the allocated data in WebAssembly's memory. You can manually read/write from this view.
  • WebAssembly's memory cannot be directly shared with JavaScript's memory. This means that you cannot pass an JavaScript array into methods like vec3.sqrLength. You first have to convert it into the given module type (e.g. vec3.fromValues) which then gives you the memory address of the allocated data.
  • There is some overhead when calling from JavaScript->WebAssembly, but it seems acceptable. Slight performance drops are noticeable when calling a function more than ~15.000 times.

Bridged methods

  • *.str so a JavaScript String is returned.
  • *.equals so a JavaScript Boolean is returned.
  • *.exactEquals (see *.equals).

New methods

  • *.view lets you create views onto your data in WebAssembly's memory.
  • *.free to free data from WebAssembly's memory.

What is left

API modules

  • mat2
  • mat2d
  • mat3
  • vec2
  • quat

Installation

npm install glmw

or the browser distribution from here.

Instantiation

Before being able to use the library, you first have to call it's init method which then asynchronously compiles the WebAssembly module.

If you call a glmw function before it got instantiated somewhere, then a TypeError is thrown, because the function is simply not compiled yet.

Browser

This builds and compiles the WebAssembly module.

<script src="//rawgit.com/maierfelix/glmw/master/dist/glmw-browser.js"></script>
glmw.init().then(ready => {
  // glmw is now ready and can be used anywhere
});

ES

Import and call the init method in your main file. Afterwards you can use glmw anywhere.

index.js

import { init, vec3 } from "glmw";
import calc from "./calc";

init().then(ready => {
  // glmw is now ready and can be used anywhere
  calc();
});

calc.js

import { mat4 } from "glmw";
export default function() {
  return mat4.create();
};

Node

Require and call the init method in your main file. Afterwards you can use glmw anywhere.

index.js

const { init, vec3 } = require("glmw");
const calc = require("./calc");

init().then(ready => {
  // glmw is now ready and can be used anywhere
  calc();
});

calc.js

const { mat4 } = require("glmw");
module.exports = function() {
  return mat4.create();
};

Usage

Simple example

As you can see here, the API didn't really change.

let a = vec3.create();
let b = vec3.fromValues(1.0, 2.0, 3.0);
vec3.add(a, a, b);
console.log( vec3.view(a) ); // Float32Array(3) [1, 2, 3]

Creating views

First, this is what glmw returns to you. Instead of references, only the numeric addresses are returned:

a = mat4.create();          // 65688
b = mat4.create();          // 65760
c = mat4.multiply(a, a, b); // 65688

You can change data by hand this way:

vA = mat4.view(a); // Float32Array([1, 0, 0...])
vA[0] = 2;         // you can now read/write
vA;                // Float32Array([2, 0, 0...])

Freeing data

Since WebAssembly doesn't have garbage collection yet, you have to be careful when and where you allocate data. You can free data by calling *.free:

a = mat4.create(); // allocate data for a
mat4.free(a);      // a's data is now freed