Android view preallocation drains only when a frame callback starts, so every new view is created inside a frame
维护者通常 2 天内回复
还没有人认领这个 Issue。
评估
- 难度
- 4/5
- 预计耗时
- 3-5 天
- 新手友好度
- 48/100
调研方向
Start with FabricMountingManager.cpp queueing and batch creation, FabricUIManager.java's frame callback drain, and MountItemDispatcher.kt's pre-mount budget. Trace how queue draining can run between frames when work is queued, then check how destroyed families are handled. Compare the 48-row reproducer's frame creation counts and timings; done means views are drained between frames and frame-time creation falls in the reported scenario.
由索引模型根据 Issue 内容生成。
描述
Description
FabricMountingManager::maybePreallocateShadowNode queues each view JavaScript creates (FabricMountingManager.cpp), and the queue is drained only when FabricUIManager's frame callback starts (FabricUIManager.java). So every new view is created inside a Choreographer#doFrame:
- When a commit's
executeMountruns before that frame, none of its views is registered yet, and its batch creates each one with aCREATEin the frame that mounts it (FabricMountingManager.cpp). - The views the drain does register are created by the same frame's pre-mount budget, which runs until half of a 60 Hz frame after the vsync (MountItemDispatcher.kt), or all at once before the next batch (MountItemDispatcher.kt).
The UI thread is often idle between frames while JavaScript renders, and none of this work can use that time.
Release RNTester from main at 928fd149cb, with the reproducer's two View examples: a band of 12 or 48 rows, each a View, a circle and a Text, replaced every 250 ms beside a native-driven rotation. Three 24 s traces of each on a Galaxy S23, Android 14, 120 Hz:
| 12 rows every 250 ms | 48 rows every 250 ms | |
|---|---|---|
| views created inside a frame | all 3348–3384 | all 13392–13536 |
| by the frame's pre-mount budget | 2411–2548 | 10112–10373 |
| by a batch, from the pre-mount queue | 747–852 | 1219–1507 |
by a CREATE in the batch |
16–125 | 1713–2061 |
| frames past 8.33 ms, of 2741–2780 | 93–94 | 123–125 |
In the longest 48-row frame, 29.89 ms, the batch created 72 views in a 12.37 ms CREATE section and the pre-mount budget created 25 more.
In a 0.86.3 app with #59057, I also drained the queue in #59057's between-frames slices from the moment its first view was queued, and dropped a destroyed family's view from the queue. On a Galaxy S23, a list whose rows are keyed by item then created 0–3 of its views per run inside a frame, against 1515–1829 before, and no frame passed 8.33 ms where one did in every run (six runs each). Would a drain that starts when the queue goes non-empty and runs between frames fit the direction you want for preallocation?
Steps to reproduce
- Apply the reproducer's two patches to
mainat 928fd149cb: the scenes in RNTester's View examples, and #59244's trace section around the drain call. - Build RNTester release for
arm64-v8aand open View → "Pre-allocation band: 48 rows every 250 ms". - Record a Perfetto trace of 24 s with the app's
viewandgfxatrace categories. - Every
SurfaceMountingManager::createViewUnsafeslice on the main thread lies inside aChoreographer#doFrameslice, underMountItemDispatcher::premountViews,MountItemDispatcher::mountViews preMountItemsor anIntBufferBatchMountItem::mountInstructions::CREATEsection, and everyFabricUIManager::drainPreallocateViewsQueueslice lies inside aChoreographer#doFrametoo.
Expected: views queued for pre-allocation are created between frames, so a new row's views are not created inside the frame that mounts it.
Actual: every view is created inside a frame (step 4), and the longest 48-row frame took 29.89 ms.
React Native Version
nightly (main at 928fd149cb)
Affected Platforms
Runtime - Android
Areas
Fabric - The New Renderer
Output of npx @react-native-community/cli info
System:
OS: Windows 11 10.0.26200
CPU: "(32) x64 AMD Ryzen 9 9950X3D 16-Core Processor "
Memory: 142.62 GB / 255.67 GB
Binaries:
Node:
version: 26.8.1
path: C:\Program Files\nodejs\node.EXE
Yarn:
version: 1.22.22
path: ~\AppData\Roaming\npm\yarn.CMD
npm:
version: 12.0.1
path: C:\Program Files\nodejs\npm.CMD
Watchman: Not Found
SDKs:
Android SDK:
API Levels:
- "33"
- "35"
- "35"
- "35"
- "36"
- "37"
Build Tools:
- 35.0.0
- 36.0.0
- 36.1.0
- 37.0.0
System Images:
- android-26 | Google APIs Intel x86_64 Atom
- android-28 | Google APIs Intel x86_64 Atom
- android-29 | Google APIs Intel x86_64 Atom
- android-30 | Google APIs Intel x86_64 Atom
- android-31 | Google APIs Intel x86_64 Atom
- android-32 | Google APIs Intel x86_64 Atom
- android-33 | Google APIs Intel x86_64 Atom
- android-34 | Google APIs Intel x86_64 Atom
- android-35 | Google APIs Intel x86_64 Atom
- android-36.1 | Google Play Intel x86_64 Atom
- android-36.1 | Pre-Release 16 KB Page Size Google Play Intel x86_64 Atom
- android-37.0 | Google APIs Intel x86_64 Atom
Android NDK: Not Found
Windows SDK: Not Found
IDEs:
Android Studio: AI-252.27397.103.2522.14617522
Visual Studio: Not Found
Languages:
Java:
version: 17.0.20.1
path: /c/Program Files/Microsoft/jdk-17.0.20.101-hotspot/bin/javac
Ruby: Not Found
npmPackages:
"@react-native-community/cli": Not Found
react:
installed: 19.3.0
wanted: 19.3.0
react-native: Not Found
react-native-windows: Not Found
npmGlobalPackages:
"*react-native*": Not Found
Android:
hermesEnabled: Not found
newArchEnabled: Not found
iOS:
hermesEnabled: Not found
newArchEnabled: Not found
Run in the React Native checkout I built RNTester from, main at 928fd149cb.
Stacktrace or Logs
No crash. Galaxy S23, Android 14, 120 Hz; release RNTester from main at 928fd149cb,
View → "Pre-allocation band: 48 rows every 250 ms", the longest frame of one trace:
main Choreographer#doFrame 29.89 ms
MountItemDispatcher::premountViews 2.75 ms, 25 views created
IntBufferBatchMountItem::mountViews 19.92 ms
mountInstructions::CREATE numInstructions=72 12.37 ms, 72 views created
mqt_v_js FabricMountingManager::executeMount, before that frame 0.60 ms
MANDATORY Reproducer
https://github.com/YevheniiKotyrlo/react-native-evidence/tree/main/preallocation-band
Screenshots and Videos
The scenes, the build, how they were traced and every reading are on this page.
- 主要语言
- C++
- 星标
- 127k
- 派生
- 25.4k
- 平均合并
- 2 天 9 小时
- 30 天内合并 PR
- 91
环境准备
- 没有 Dockerfile 或 Docker Compose 文件
- 有 Pull Request 模板
- 阅读贡献指南
从这里开始
- 先读完整个 Issue,再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库,在一个分支上完成修改。
- 提交 Pull Request,并在描述里引用这个 Issue 编号。
react/react-native 的其他 Issue
-
难度 2/5 1-3 小时 新手友好度 70/100
react/react-native#59722 ·
维护者通常 2 天内回复
-
Android: new WebSocket() crashes the app for a URL that OkHttp or java.net.URI rejects可能已有人在做 @sorinc03 于 1 天前认领。 未关闭🌐Networking Needs: Author Feedback Needs: Repro Platform: Android
难度 2/5 1-3 小时 新手友好度 72/100
react/react-native#59704 · 1 条评论 ·
维护者通常 2 天内回复
-
难度 1/5 1 小时以内 新手友好度 88/100
react/react-native#59694 ·
维护者通常 2 天内回复
-
Android autolinking keeps a stale autolinking.json when only bun.lock changes, since bun.lock isn't in the default lockFiles可能已有人在做 @lazerg 于 3 天前认领。 未关闭
难度 1/5 1 小时以内 新手友好度 85/100
react/react-native#59693 ·
维护者通常 2 天内回复
-
Image.resolveAssetSource's TypeScript declaration drops the null it returns for an unknown asset可能已有人在做 @Hassan1305 于 3 天前认领。 未关闭
难度 2/5 1-3 小时 新手友好度 78/100
react/react-native#59648 · 1 条评论 ·
维护者通常 2 天内回复
查看 react/react-native 的全部 Issue
相似的 Issue
-
难度 2/5 1-3 小时 新手友好度 80/100
-
难度 1/5 1 小时以内 新手友好度 82/100
-
难度 2/5 1-3 小时 新手友好度 72/100
eclipse-score/lifecycle#929 ·
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 62/100
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 68/100
colbymchenry/codegraph#3223 ·
维护者通常 1 天内回复