Android Benchmarking and other stories
summary
Shows how to benchmark a real app using Jetpack micro and macro benchmarks and the Android Profiler, compares outputs (microbenchmarks for tight CPU hot spots, macrobenchmarks producing JSON with min/median/max and the macrobenchmark startup metric), and demos the new in‑app JankStats (released 2022‑02‑09) that reports per‑frame nanos and an isJank boolean and can aggregate before upload. Describes storing results, running on real devices or Firebase Test Lab, and detecting regressions with a sliding-window step-fitting algorithm to avoid false positives while alerting on regressions or unexpected improvements, using Android Profiler traces to guide what to benchmark.
AI-written summary · about
chapters (11)
- 0:00Introduction and speakers
- 1:10What is Android benchmarking
- 4:22Android profiling tools
- 5:45Microbenchmarking basics
- 8:16Macrobenchmarking overview
- 10:45Introducing JankStats
- 13:54Detecting regressions in CI
- 21:10Regression detection approaches
- 25:02Statistical regression detection
- 29:13Key takeaways and resources
- 31:20Q&A and closing
related talks
33mAndroid Benchmarking and other stories
44mAndroid Benchmarking and other stories
40mMonitor App performance using Jetpack Benchmark
43mBenchmarking your application
42mPerformance Paradox: Navigating the Complexities of Mobile App Optimisation
20mMeasure What Matters with Jetpack Benchmark (Lightning)
39mAndroid App Performance in a Nutshell
05mWorkshop: Inspect and improve app performance
