flutter-performance

原文涵盖的主题

• Flutter performance
• Measure first
• Rebuilds
• CPU work → isolates
• Lists and grids
• Images (most common OOM cause)
• Painting and animation
• App size

Verified 2026-09-25 against docs.flutter.dev/perf (Flutter 3.47: Impeller is the default renderer on iOS, Android API 29+, macOS, Windows and Linux; not web). A frame has 16 ms at 60 Hz (8 ms at 120 Hz). Almost every real Flutter performance problem is self-inflicted: too-wide rebuilds, work in build(), eager lists, full-resolution images, or CPU work on the UI isolate. The engine (Impeller) is rarely the bottleneck. Profile in profile mode on a real (ideally low-end) device. Debug-mode numbers are meaningless. Use DevTools Performance and the Timeline (frame times, jank), the Performance Overlay, and Memory (leaks, image cache). Optimize when you see dropped frames, stuttering animations, or laggy scrolling. Otherwise write clear code first. No premature optimization. Call setState as low in the tree as possible. State high up rebuilds everything below it. Split big widgets into small widget classes (not helper methods, which rebuild with the parent). Use const constructors wherever possible; const widgets are skipped on rebuild. Nothing expensive in build() (parsing, sorting, filtering, formatting big data). Do it in initState/didUpdateWidget (recompute only when input changed) o

这里只展示节选，请以完整原文和当前要求为准。