A Quest to Guarantee Responsiveness: Scheduling On and Off the Main Thread
summary
Claim: achieving responsiveness needs scheduling plus selective multithreading, not just wording about "examples." Chrome explores a main-thread scheduler API with global and virtual task queues and an isInputPending signal, and a Task Worklet model to run named tasks on a sticky threadpool. Facts: target is ~10ms chunks vs 16ms frames, worker startup ≈10ms and a thread hop 1–15ms. Prototypes live in Chromium behind a flag, polyfills and benchmarks show smoother frames but some extra input latency when using workers.
AI-written summary · about
chapters (6)
- 0:00Start
- 5:14Frame Rendering Pipeline
- 8:02Challenges & Proposals
- 14:22A+ threading use-cases
- 19:06Requirements of an ideal model
- 28:01Base cost of a worker
