About this ebook
Most Android performance advice stops at a list of tips. Mastering Android Performance teaches the discipline behind them. You'll learn how to measure a problem correctly, trace it to its actual cause, fix it, and lock the fix in with automated checks so it never quietly comes back. The book covers every layer of an app, from the first frame of a cold start to the thermal limits of a phone that has been busy for an hour.
Every chapter opens with an incident at FreshCart, a fictional grocery-delivery app built from failure patterns common in real production apps. Each one follows the problem from symptom to trace to fix, with concrete before-and-after numbers.
What you'll learn:
How Android really runs and renders your app: Zygote, ART, garbage collection, the main thread, and the frame pipeline
Measuring correctly with Android Studio profilers, Perfetto, Macrobenchmark and real-user field data
Cutting startup time with lazy initialization, App Startup, and Baseline and Startup Profiles
Finding and fixing ANRs, main-thread stalls, lock contention and thread explosions
Smooth rendering in both Views and Jetpack Compose, including stability, strong skipping and compiler reports
Memory leaks, process death, and designing for low-RAM and Android Go devices
Battery, network and storage efficiency, including WorkManager and the Android 12–16 foreground service rules
R8, app size, build speed, and CI gates that catch regressions before release
The topics most books skip: native code and the mandatory 16 KB page size, on-device AI with LiteRT and Gemini Nano, and thermal-aware adaptive performance
Built to be used, not just read. All 32 chapters end with a hands-on lab you run against your own app on a real device, plus pitfalls, a checklist, exercises and version notes. Four appendices give you a command cheat sheet, 20 ready-to-use Perfetto SQL queries, performance checklists and a full metric reference to keep open at your desk.
Who this book is for: Android developers comfortable with Kotlin and either Views or Jetpack Compose. No prior profiling experience is needed; Part I builds the mental model every later chapter depends on.
