Remove v8-compile-cache

[ybiquitous]

Split from #6898
See also zertosh/v8-compile-cache#30

We've used v8-compile-cache for startup performance:

stylelint/bin/stylelint.js

Lines 5 to 6 in 2413e8f

	// to use V8's code cache to speed up instantiation time
	require('v8-compile-cache');

But #6898 exposed a problem with ESM and dynamic imports.

If there are no big performance penalties or concerns, we'd like to remove v8-compile-cache to prevent future issues.

[ybiquitous]

Ref: #6901

[ybiquitous]

I'm looking into this. v8-comple-cache was added by:

• feat: add v8-compile-cache #4272

This was inspired by ESLint:

• Chore: add v8-compile-cache to speed up instantiation time eslint/eslint#11921

However, ESLint removed v8-comple-cache by:

• feat: Enable eslint.config.js lookup from CLI eslint/eslint#16235

[ybiquitous]

v8-compile-cache is a blocker towards the migration to ESM:

• Move to ESM #5291 (comment)

[ybiquitous]

I tried a benchmark with the following script (saving scripts/benchmark-v8-compile-cache.js):

'use strict';

/* eslint-disable no-console */

const cp = require('child_process');
const Benchmark = require('benchmark');

const cmdArgs = ['../bin/stylelint.js', 'visual.css', '-c', 'visual-config.json'];
const cwd = __dirname;

const suite = new Benchmark.Suite('test v8-compile-cache', {
	onStart() {
		try {
			cp.execFileSync('node', cmdArgs, { cwd });
		} catch (e) {
			if (e.status === 2) {
				// ignore
			} else {
				throw e;
			}
		}
	},
	onComplete(event) {
		const s = event.currentTarget;
		const fastest = s.filter('fastest').map('name');

		console.log(`Fastest is "${fastest}"`);
	},
});

suite.add('enable cache', {
	defer: true,
	fn(deferred) {
		const env = { ...process.env };

		cp.execFile('node', cmdArgs, { cwd, env }, () => deferred.resolve());
	},
	onComplete(event) {
		console.log(event.target.toString());
	},
});

suite.add('disable cache', {
	defer: true,
	fn(deferred) {
		const env = {
			...process.env,
			DISABLE_V8_COMPILE_CACHE: '1',
		};

		cp.execFile('node', cmdArgs, { cwd, env }, () => deferred.resolve());
	},
	onComplete(event) {
		console.log(event.target.toString());
	},
});

suite.run({ async: true });

Environment (M1 MacBook Pro):

$ sw_vers
ProductName:		macOS
ProductVersion:		13.3.1
ProductVersionExtra:	(a)
BuildVersion:		22E772610a

$ npm version
{
  stylelint: '15.7.0',
  npm: '9.6.7',
  node: '20.2.0',
  acorn: '8.8.2',
  ada: '2.4.0',
  ares: '1.19.0',
  base64: '0.5.0',
  brotli: '1.0.9',
  cjs_module_lexer: '1.2.2',
  cldr: '43.0',
  icu: '73.1',
  llhttp: '8.1.0',
  modules: '115',
  napi: '8',
  nghttp2: '1.52.0',
  nghttp3: '0.7.0',
  ngtcp2: '0.8.1',
  openssl: '3.0.8+quic',
  simdutf: '3.2.9',
  tz: '2023c',
  undici: '5.22.0',
  unicode: '15.0',
  uv: '1.44.2',
  uvwasi: '0.0.16',
  v8: '11.3.244.8-node.9',
  zlib: '1.2.13'
}

Run:

$ node scripts/benchmark-v8-compile-cache.js
enable cache x 3.82 ops/sec ±2.97% (23 runs sampled)
disable cache x 3.73 ops/sec ±2.67% (23 runs sampled)
Fastest is "enable cache"

---

As a result, I found out that there are almost no differences on the latest Node.js 20.2.0.

EDIT: This benchmark uses the DISABLE_V8_COMPILE_CACHE environment variable.