2nd/10_Wiki/Topics/AI_and_ML/브라우저 메인 스레드 최적화 및 타임 슬라이싱.md

id, title, category, status, canonical_id, aliases, duplicate_of, source_trust_level, confidence_score, verification_status, tags, raw_sources, last_reinforced, github_commit, tech_stack

id	title	category	status	canonical_id	aliases	duplicate_of	source_trust_level	confidence_score	verification_status	tags	raw_sources	last_reinforced	github_commit	tech_stack
wiki-2026-0508-브라우저-메인-스레드-최적화-및-타임-슬라이싱	브라우저 메인 스레드 최적화 및 타임 슬라이싱	10_Wiki/Topics	verified	self	Main Thread Optimization Time Slicing INP Optimization Long Tasks Scheduler API	none	A	0.9	applied	web-performance main-thread scheduler inp long-tasks web-workers		2026-05-10	pending	language framework typescript vanilla-web-platform

브라우저 메인 스레드 최적화 및 타임 슬라이싱

매 한 줄

"매 long task 를 매 50ms 미만 chunk 로 split, scheduler.yield() 로 매 input 에 양보". 매 INP (Interaction to Next Paint, 2024 부터 Core Web Vital) 의 핵심 metric — 매 200ms 이하 가 Good. 매 2026 의 Scheduler API + isInputPending + Web Worker 의 매 standard toolkit.

매 핵심

매 Main thread 의 책임

• JS execution, Style + Layout + Paint, Compositing prep, Event dispatch.
• Long task: 50ms 초과 task — INP 의 적.
• Render-blocking: long task 동안 rAF, input handler 모두 starve.

매 Time slicing 전략

• Yield to event loop: scheduler.yield() (2026 baseline), 매 fallback 으로 setTimeout(0).
• Priority-aware scheduling: scheduler.postTask({priority: 'background'}).
• isInputPending(): input 의 pending 시 매 즉시 yield.
• Web Worker offload: pure compute (parsing, encoding, crypto) 는 main thread 떠남.

매 응용

1. Large list rendering: virtualized + chunked.
2. JSON parsing / encoding: Worker + Transferable.
3. Hydration: priority-based React 19 / Astro / Qwik resumability.

💻 패턴

scheduler.yield() (2026 baseline)

async function processChunks(items: Item[]) {
  for (const it of items) {
    process(it);
    if (navigator.scheduling?.isInputPending?.()) {
      await scheduler.yield(); // resume after input is handled
    }
  }
}

scheduler.postTask with priority

// User-blocking: must run ASAP
scheduler.postTask(updateUserCriticalUI, { priority: 'user-blocking' });
// User-visible: default
scheduler.postTask(refreshSidebar, { priority: 'user-visible' });
// Background: prefetch, analytics flush
scheduler.postTask(prefetchNext, { priority: 'background' });

Manual time slicer (fallback)

async function timeSlice<T>(items: T[], work: (item: T) => void, budgetMs = 5) {
  let start = performance.now();
  for (const it of items) {
    work(it);
    if (performance.now() - start > budgetMs) {
      await new Promise<void>((r) => setTimeout(r, 0));
      start = performance.now();
    }
  }
}

Web Worker offload (Comlink pattern)

// worker.ts
import * as Comlink from 'comlink';
const api = {
  parseLargeJSON: (text: string) => JSON.parse(text),
  hashFile: async (buf: ArrayBuffer) => {
    const h = await crypto.subtle.digest('SHA-256', buf);
    return new Uint8Array(h);
  },
};
Comlink.expose(api);

// main.ts
const worker = new Worker(new URL('./worker.ts', import.meta.url), { type: 'module' });
const api = Comlink.wrap<typeof api>(worker);
const parsed = await api.parseLargeJSON(hugeText);

Transferable ArrayBuffer (zero-copy)

const buf = new ArrayBuffer(1024 * 1024 * 10);
worker.postMessage({ buf }, [buf]); // ownership transfers; main no longer holds

INP measurement (Performance Observer)

const observer = new PerformanceObserver((list) => {
  for (const entry of list.getEntries() as PerformanceEventTiming[]) {
    const inp = entry.processingEnd - entry.startTime;
    if (inp > 200) console.warn('Slow interaction', entry.name, inp);
  }
});
observer.observe({ type: 'event', buffered: true, durationThreshold: 16 });

React 19 — useTransition for non-urgent updates

const [isPending, startTransition] = useTransition();

function handleSearch(q: string) {
  setQuery(q); // urgent — input echo
  startTransition(() => {
    setSearchResults(filterHugeList(allItems, q)); // non-urgent — yields to input
  });
}

IntersectionObserver — lazy heavy work

const io = new IntersectionObserver((entries) => {
  for (const e of entries) {
    if (e.isIntersecting) {
      scheduler.postTask(() => mountHeavyComponent(e.target), { priority: 'background' });
      io.unobserve(e.target);
    }
  }
}, { rootMargin: '200px' });
document.querySelectorAll('.lazy-section').forEach(el => io.observe(el));

Long Animation Frames (LoAF) API

new PerformanceObserver((list) => {
  for (const e of list.getEntries() as any[]) {
    console.warn('LoAF', e.duration, e.scripts);
  }
}).observe({ type: 'long-animation-frame', buffered: true });

매 결정 기준

상황	Approach
> 50ms synchronous work	Time slice with scheduler.yield()
Pure compute (parse, hash, image)	Web Worker + Transferable
Non-urgent state update (React)	useTransition
Off-screen heavy work	IntersectionObserver + postTask
Background prefetch	scheduler.postTask priority 'background'
Hot UI update with input	isInputPending checkpoint

기본값: 매 task > 5ms 면 yield consider. 매 task > 50ms 면 split mandatory. 매 pure compute 는 Worker 의 first choice.

🔗 Graph

• 부모: Web-Performance · Core Web Vitals Optimization (INP, LCP 개선) · INP-Optimization
• 변형: Scheduler-API · Web Worker (웹 워커) · OffscreenCanvas
• 응용: 성능 중심의 웹 애니메이션 및 인터랙션 구현
• Adjacent: Long Tasks · Long-Animation-Frames · Comlink

🤖 LLM 활용

언제: time-slice boilerplate, Worker setup scaffold, scheduler API migration plan. 언제 X: 매 actual INP 측정 — RUM (CrUX, Web Vitals JS) 만 truth. LLM 의 perf claim 의 hallucination.

❌ 안티패턴

• Sync heavy work in input handler: 매 INP 폭증.
• setTimeout(0) busy-loop: scheduler.yield() preferred (priority-aware).
• Worker per task spawn: 매 startup overhead. Pool 또는 long-lived worker.
• structuredClone of huge data to Worker: Transferable 사용.
• Ignoring isInputPending: 매 yield 의 timing 의 너무 늦음.

🧪 검증 / 중복

• Verified (web.dev/inp 2026, Scheduler API spec, LoAF spec, Chrome DevTools Performance docs).
• 신뢰도 A.

🕓 Changelog

날짜	변경
2026-05-08	Phase 1
2026-05-10	Manual cleanup — Scheduler API, time slicing, Worker offload, INP patterns