Avoid global variable reference

[falsandtru]

ref: https://github.com/nodejs/node/search?l=JavaScript&q=primordials+Object

[falsandtru]

🤔

Error: You must provide `username` and `accessKey` as options or set "SAUCE_USERNAME" and "SAUCE_ACCESS_KEY" in your environment variables.
    at runSauceLabs (/home/travis/build/primus/eventemitter3/node_modules/sauce-test/lib/run-sauce-labs.js:46:29)
    at runTestsAtLocation (/home/travis/build/primus/eventemitter3/node_modules/sauce-test/index.js:68:12)
    at publishCode.then.location (/home/travis/build/primus/eventemitter3/node_modules/sauce-test/index.js:140:12)
    at /home/travis/build/primus/eventemitter3/node_modules/promise/lib/core.js:33:15
    at flush (/home/travis/build/primus/eventemitter3/node_modules/promise/node_modules/asap/asap.js:27:13)
    at process._tickCallback (internal/process/next_tick.js:61:11)

[lpinca]

I don't understand this PR. It doesn't use Node.js primordials so what is this protecting against?

[falsandtru]

Global variable reference greatly degrades the runtime performance as follows.

[BAD 0.7%]
Global variable reference
x 3,803,038 ops/sec ±1.03% (57 runs sampled)

[BAD 0%]
Global variable reference x10
x 331,805 ops/sec ±20.80% (52 runs sampled)

[BAD 0.5%]
Instantiation of global variable
x 2,894,146 ops/sec ±2.00% (57 runs sampled)

[Fastest]
Local variable reference
x 527,495,329 ops/sec ±0.58% (59 runs sampled)

Instantiation of local variable
x 473,611,915 ops/sec ±20.21% (52 runs sampled)

https://falsandtru.github.io/benchmark/suites/10/

This PR prevents the performance degradation.

[lpinca]

Can you show the difference by running eventemitter3 benchmarks before and after the change?

[lpinca]

The use of primordials caused performance regressions on Nodes.js not boost nodejs/node#29766.

[lpinca]

$ node -v
v13.7.0
$ cat index.js
'use strict';

const Benchmark = require('benchmark');

const suite = new Benchmark.Suite();
const Arr = Array;

let ret;

suite.add('local', function() {
  ret = new Arr();
});

suite.add('global', function() {
  ret = new Array();
});

suite.on('cycle', function(event) {
  console.log(event.target.toString());
});

suite.on('complete', function() {
  console.log('Fastest is ' + this.filter('fastest').map('name'));
});

suite.run({ async: true });
$ node index.js
local x 75,396,008 ops/sec ±0.35% (94 runs sampled)
global x 76,764,603 ops/sec ±0.10% (98 runs sampled)
Fastest is global

[falsandtru]

That issue lists two causes but this PR doesn't use either.

The regressions seem to come from two types of code patterns:

1. Calling functions through Function.prototype.{call, apply} instead of just calling them directly. There is an issue opened in the upstream by @\bmeck
   https://bugs.chromium.org/p/v8/issues/detail?id=9702
2. Looking up properties from fronzen objects - objects from our primordials namespace are frozen so lookups like const { Reflect } = primordials; Reflect.apply(...); is slower than just Reflect.apply when Reflect comes from the global object. This can be mitigated by caching the lookup results upfront, e.g. #29633 by @\mcollina There is also a fairly odd tracking issue for this in the upstream: https://bugs.chromium.org/p/v8/issues/detail?id=6831

[lpinca]

Yes, the point is that the use of primordials did not cause any performance improvement in Node.js

[falsandtru]

Confirmed. Looks like it is difficult to improve the performance on browsers without decreasing the performance 1-2% on Node.js.

[kmashint]

Note Function('return this')() breaks a typical Content-Security-Policy, so it shouldn't resort to that.