如何理解Low CPU usage and high latency？

返回 技术表达与英语 目录

标准回答

Low CPU usage does not mean low latency. The critical thread may be waiting for a lock, a Binder reply, I/O, or GPU work. It may also be runnable but not scheduled quickly enough. A device-wide average can hide a bottleneck on one core.

I would first identify the slow interval, then separate running time, runnable time, and blocked time. After that, I would follow the dependency to the thread or service that can make progress. This helps me decide whether to optimize computation, scheduling, or a resource wait.

中文要点：先给结论，再列候选，最后描述取证顺序。用may表示尚未确认的原因，避免把推断说成事实。

追问与回答

What evidence would you collect next

I would collect the relevant scheduler events and call stacks, and correlate the client and server timelines if the thread is waiting on Binder.

内容来源

• Android 与 AI 工程 Wiki · 原条目 q101