Skip to content

legraphista/isolated-threads

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

21 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Isolated Threads

Yes, nodejs has real threads too! Many thanks to isolated-vm!

Getting Started

Installation

npm i isolated-threads

Example JS

const { Thread, ThreadPool } = require('isolated-threads');

const context = () => {

  const random = () => Math.round(Math.random() * 255);

  return ({ w, h, c }) => {

    const sharedArray = new SharedArrayBuffer(w * h * c);
    const image = new Uint8Array(sharedArray);

    console.log('generating noise image with params', { w, h, c });

    const stride = w * c;

    for (let y = 0; y < h; y++) {
      for (let x = 0; x < w; x++) {
        for (let k = 0; k < c; k++) {
          image[y * stride + x * c + k] = random();
        }
      }
    }

    return sharedArray;
  }
};

const t = new Thread(context);

(async () => {
  const interval = setInterval(() => console.error('event loop'), 100);

  const image = new Uint8Array(await t.run({ w: 3840, h: 2160, c: 3 }));
  console.log('image output buffer len', image.length);

  clearInterval(interval);
})();

Example TS

import Thread = require("./src/thread");

const context = () => {

  const random = () => Math.round(Math.random() * 255);

  return ({ w, h, c }: { w: number, h: number, c: number }): SharedArrayBuffer => {

    const sharedArray = new SharedArrayBuffer(w * h * c);
    const image = new Uint8Array(sharedArray);

    console.log('generating noise image with params', { w, h, c });

    const stride = w * c;

    for (let y = 0; y < h; y++) {
      for (let x = 0; x < w; x++) {
        for (let k = 0; k < c; k++) {
          image[y * stride + x * c + k] = random();
        }
      }
    }

    return sharedArray;
  }
};

// inferred arguments and result from context 
const t = new Thread(context);
// or
// enforce arguments and result to context
const t = new Thread<[{ w: number, h: number, c: number }], SharedArrayBuffer>(context);

(async () => {
  const interval = setInterval(() => console.error('event loop'), 100);

  const shared = await t.run({ w: 3840, h: 2160, c: 3 });
  const image = new Uint8Array(shared);
  console.log('image output buffer len', image.length);

  clearInterval(interval);
})();

Class Thread

new Thread(context, {memoryLimit})

  • context [Function]
    • Holds a context with constants and/or functions required
    • Returns the function that will be called ar tuntime on a separate thread.
  • options [object]
    • memoryLimit: number - the maximum memory (in MB) at which to trigger an error and disposal (for more info see here)

thread.run(...args): Promise<*>

  • Will run the context returned function in a separate thread with the given parameters and the returned promise will resolve with the function's return data.

Class ThreadPool

new ThreadPool(context, poolSize, {memoryLimit})

  • context [Function]
    • Holds a context with constants and/or functions required
    • Returns the function that will be called ar tuntime on a separate thread.
  • poolSize [number] - the number of threads to be spawned
  • options [object]
    • memoryLimit: number - the maximum memory (in MB) at which to trigger an error and disposal (for more info see here)

threadPool.run(...args): Promise<*>

  • Will run the context returned function in a separate thread with the given parameters and the returned promise will resolve with the function's return data.
  • Avilable schedulers as of now are only round-robin, more to come.

A word about Buffers and other node features

Limitations or running in a v8 only isolate

Since the different thread is executed in an isolated instance of v8, you will not be able to use Buffer or require and you will not have access to any kind of event loop functions, like setTimeout or setInterval.

Moving large amounts of data

When you are passing data between isolates, the data is serialized and copied between threads (read more here).
If you with to copy a large amount of data, you should use Typed Arrays (like UInt8Array or Float64Array) with a SharedArrayBuffer backend (if available).
In the above example we allocate a w * h * c sized shared buffer, and then use it as a backend for Uint8Array.
We do all data manipulation using the uint8 array and then return back the shared one.
Upon receiving the shared array from the thread, we reinterpret it into a uint8 array and extract our data.

Debugging

To debug other threads, constructor's options must contains inspector as a port number and optioally filename as a file name for the thread. If filename is not passed, then the file:line where the thread is spawned will be used. Upon spawning the thread, you will receive a connection URI in the stderr stream. Navigate to the URI using Google Chrome or use it to configure your IDE of choice.

Caveats

Passing functions

As of now, you cannot pass function callback to the isolate. If requested, it can be added, but the callback function will run on the main thread, defeating the purposes of isolating threads.

Mixing shared buffers in objects

You cannot pass a shared buffer inside an object. You must pass it in a different argument.

  • ({sharedBuffer, w, h, c}) => sharedBuffer will not work
  • (sharedBuffer, {w, h, c}) => sharedBuffer will work
  • ({a, b, c}) => ({sum: a+b+c}) will work
  • (...args) => Math.max(...args) will work

About

NodeJS real threads with shared memory

Resources

License

Stars

Watchers

Forks

Packages

No packages published