camaro v6

Posted on May 19, 2020  •  2 minutes  • 245 words

I recently discover piscina project. It’s a very fast and convenient Node.js worker thread pool implementation.

Remember when worker_threads first introduced, the worker startup is rather slow and pool implementation is generally advised. However, there wasn’t any good enough implementation yet until piscina.

Since v4 when I move to WebAssembly, camaro performance took a huge hit (3 folds) and I was still trying to find a way to fix this perf regression.

Well, piscina (worker_threads) seems to be the answer to that.

Take a look at piscina example:

const Piscina = require('piscina');

const piscina = new Piscina({
  filename: path.resolve(__dirname, 'worker.js')
});

(async function() {
  const result = await piscina.runTask({ a: 4, b: 6 });
  console.log(result);  // Prints 10
})();

and worker.js

module.exports = ({ a, b }) => {
  return a + b;
};

Sure it looks simple enough so I wrote a quick script to wrap camaro with piscina. And the performance improvement is sweet: it’s about five times faster (ops/sec) and the CPU on my laptop is stressed nicely.

camaro v6: 1,395.6 ops/sec
fast-xml-parser: 153 ops/sec
xml2js: 47.6 ops/sec
xml-js: 51 ops/sec

More importantly, it scales nicely with CPU core counts, which camaro v4 with WebAssembly isn’t.

In order to use this, I would have to drop support for Node version 11 and older but the performance improvement of this magnitude should guarantee such breaking changes right?

I published the first alpha build to npm if anyone want to give it a try.