ART of Q

In this lightning talk. I will summarize the content of a session "Understanding Android Runtime (ART) for Faster Apps" at Google I/O '19.

[Reference]
- Understanding Android Runtime (ART) for Faster Apps (Google I/O'19) - YouTube: https://www.youtube.com/watch?v=1uLzSXWWfDg
- Cloud profiles: https://youtu.be/1uLzSXWWfDg?t=210
- Application images: https://youtu.be/1uLzSXWWfDg?t=368
- Application images: https://youtu.be/1uLzSXWWfDg?t=440
- StartupAnalyzerKt: https://android.googlesource.com/platform/tools/trebuchet/
- Activity#reportFullyDrawn(): https://developer.android.com/reference/android/app/Activity#reportFullyDrawn()
- Trash Talk (Android Dev Summit '18) - YouTube: https://www.youtube.com/watch?v=Zc4JP8kNGmQ&feature=youtu.be

Transcript

1. ART of Q 2019-05-21 (mixi & CA).aab Hiroshi Kurokawa
2. Understanding Android Runtime (ART) for Faster Apps https://www.youtube.com/watch?v=1uLzSXWWfDg
3. Improvements in ART • App Startup Time • Garbage Collection
4. Improvements in ART • App Startup Time • Garbage Collection
5. Cloud profiles https://youtu.be/1uLzSXWWfDg?t=210
6. Application images https://youtu.be/1uLzSXWWfDg?t=368
7. Application images https://youtu.be/1uLzSXWWfDg?t=440
8. Unspecialized app process pool System Server Zygote App Process Fork
9. Unspecialized app process pool System Server Unspecialized App Process
10. Analyzing startup time • StartupAnalyzerKt
 https://android.googlesource.com/platform/tools/ trebuchet/ • Use Activity#reportFullyDrawn()
 https://developer.android.com/reference/android/app/ Activity#reportFullyDrawn()
11. Improvements in ART • App Startup Time • Garbage Collection
12. Generational Concurrent Compaction
13. CC Root reachable unreachable
14. CC Root To Space Evacuated Region
15. Generational CC Younger Generation Older Generation Full-heap
 GC Young-gen GC
16. Generational CC Younger Generation Older Generation
17. 2-phase collection
18. 2-phase collection • For full-heap GC • Trace live objects and compute liveness stat • Evacuate regions based on up-to-date stats
19. Object pooling • In most cases, it doesn’t make sense • Even if you are sure it improves the performance, it would be better to measure the performance
20. Reference https://youtu.be/Zc4JP8kNGmQ