Experiments for your Android builds driven by Gradle Profiler

on YouTube ↗43m

" summary

Uses Gradle Profiler to run reproducible benchmarks and saves CSV reports, running 4 warm‑up 15 measure builds per scenario; compares base vs PR with a left‑tailed Student t test and reports p‑value > 0.05 when JDK11 is actually slower so there is no strong evidence of improvement. Also shows a Gradle/tooling bump with p‑value ≈ 0 indicating a real speedup, and publishes a small Python tool (pip) to automate the t‑test.

AI-written summary · about

" chapters (15)

1. 0:00Intro
2. 4:58Measuring Gradle builds
3. 7:37Installing Gradle Profiler
4. 11:06Demo
5. 15:34Scenarios
6. 16:36Results
7. 20:18Statistical Inference
8. 23:39Statistical significance
9. 25:10Probability functions
10. 26:39Pvalue
11. 28:22Tail analysis
12. 29:55Parrot test
13. 30:44Tools
14. 33:16Example
15. 38:10Alternative Hypothesis

" related talks

• 33mAndroid Benchmarking and other storiesEnrique López Mañas, Iury Souza · droidcon New York 2022
• 33mIMPROVING ANDROID BUILD PERFORMANCEStefan Oehme · droidcon Berlin 2018
• 31mImproving Android Build performanceTony Robalik · droidcon San Francisco 2018
• 40mImproving Android Build PerformanceVolker Leck · droidCon Berlin 2019
• 33mAndroid Benchmarking and other storiesEnrique López Mañas, Iury Souza · droidcon San Francisco 2022
• 30mImproving Android Build performanceTony Robalik · droidCon NYC 2018
• 28mMaking the most of your Gradle BuildsEgor Andreevici · Droidcon Italy 2017
• 43mBenchmarking your applicationXavier F. Gouchet · Android Makers 2020