Sure:

const { Worker, isMainThread, parentPort } = require("worker_threads");
const fs = require("fs");

function work() {
  const fib = n => (n <= 1 ? 1 : fib(n - 1) + fib(n - 2));
  let x;
  // Just do some expensive calculations
  for (let i = 0; i < 1e4; i++) {
    x = fib(1 + (i % 24));
  }
  if (parentPort) parentPort.postMessage(x);
}
const proftitle = "fibonacci";

if (isMainThread) {
  const v8Profiler = require("v8-profiler-next");
  v8Profiler.setSamplingInterval(100);
  v8Profiler.startProfiling(proftitle, true);
  new Worker(__filename).on("message", result => {
    const profiler = v8Profiler.stopProfiling(proftitle);
    profiler.export(function(error, result) {
      fs.writeFileSync("profile.cpuprofile", result);
      profiler.delete();
    });
    console.log(result);
  });
} else {
  work(); // worker thread
}

This is the profile:

(The whole thing takes ~500ms. It looks like only the orchestration on the main thread is captured and then it's idling, until the worker posts message back)

In comparison this is how profile looks when profiling the fibonnaci calculation itself, without using worker_threads:

Flamecharts generated from the cpuprofiles (using http://github.com/brendangregg/FlameGraph):

VS standalone fibonacci function, without worker threads:

Worker threads #9

Description

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions