Improving App Performance with Baseline Profiles

Posted by Kateryna Semenova, DevRel Engineer; Rahul Ravikumar, Software program Engineer; Chris Craik, Software program Engineer

Why is startup time essential?

A number of apps discover correlation between app efficiency and person engagement. Individuals count on apps to be responsive and quick to load. Startup time is without doubt one of the main metrics for app efficiency and high quality.

A few of our companions have already invested loads of time and sources for app startup optimizations. For instance, take a look at the Fb story.

On this weblog submit we’ll focus on Baseline Profiles and the way they enhance app and library efficiency, together with startup time by as much as 40%. Whereas this blogpost focuses on startup, baseline profiles additionally considerably enhance jank as nicely.

Historical past

Android 9 (API stage 28) launched ART optimizing profiles in Play Cloud to enhance app startup time. On common, we’ve seen that apps’ chilly begins are a minimum of 15% sooner throughout quite a lot of units when Cloud Profiles can be found.

How do Profiles work?

When the app is first launched after set up or replace, its code runs in an interpreted mode till it’s JITted. In an APK, Java and Kotlin code is compiled as dex bytecode, however not absolutely compiled to machine code (since Android 6), on account of the price of storing and loading absolutely compiled apps. Courses and strategies which can be incessantly used within the app, in addition to these used for app startup, are recorded right into a profile file. As soon as the gadget enters idle mode, ART compiles the apps based mostly on these profiles. This accelerates subsequent app launches.

Beginning with Android 9 (API stage 28), Google Play additionally gives Cloud Profiles. When an app runs on a tool, the profiles generated by ART are uploaded by the Play Retailer app and aggregated within the cloud. As soon as there are sufficient profiles uploaded for an utility, the Play app makes use of the aggregated profile for subsequent installs.

Drawback

Whereas Cloud Profiles are nice when they’re out there, they don’t seem to be all the time prepared for use when an app is put in. Accumulating and aggregating the profiles often takes a number of days, which is an issue when many apps replace on a weekly foundation. Many customers will set up an replace earlier than the Cloud Profile is offered. The Google Android staff began in search of different methods to enhance the latency of profiles.

Answer

Baseline Profiles are a brand new mechanism to offer profiles which can be utilized on Android 7 (API stage 24) and better. A baseline profile is an ART profile generated by the Android Gradle plugin utilizing a human readable profile format that may be supplied by apps and libraries. An instance would possibly appear like this:

HSPLandroidx/compose/runtime/ComposerImpl;->updateValue(Ljava/lang/Object;)V
HSPLandroidx/compose/runtime/ComposerImpl;->updatedNodeCount(I)I
HLandroidx/compose/runtime/ComposerImpl;->validateNodeExpected()V
PLandroidx/compose/runtime/CompositionImpl;->applyChanges()V
HLandroidx/compose/runtime/ComposerKt;->findLocation(Ljava/util/Record;I)I

The binary profile is saved in a selected location within the APK property listing
(property/dexopt/baseline.prof).

Baseline Profiles are created throughout construct time, shipped as a part of the APK to Play, after which despatched from Play to customers when an app is downloaded. They fill the hole within the ART Cloud Profile pipeline, when Cloud Profiles will not be but out there, and mechanically merge with Cloud Profiles when they’re.

This diagram shows the baseline profile workflow from creation via end-user supply.

One of many largest advantages of Baseline Profiles is that they are often developed and evaluated regionally so builders can see real looking end-user efficiency enhancements. They’re additionally supported on a decrease model of Android(7 and better) than Cloud Profiles, that are solely out there beginning in Android 9.

Affect

App devs

In early 2021, Google Maps switched from a two-week to a one-week launch cycle. Extra frequent updates meant extra incessantly discarding native pre-compilation, and extra customers experiencing gradual launches with out Play Cloud Profiles. Through the use of Baseline Profiles, Google Maps improved their common startup time by 30% and noticed a corresponding improve in searches by 2.4%, an immense acquire for such a longtime app.

Library devs

Code in a library is rather like that of an app – it is not absolutely compiled by default, which generally is a downside if it does vital work on the essential path of startup.

Jetpack Compose is a UI library that isn’t part of the Android system picture and thus not absolutely compiled when put in, not like a lot of the Android View toolkit code. This was inflicting efficiency issues, particularly for the primary few chilly launches of the app.

To resolve this downside, Compose makes use of profile installer. It ships baseline profile guidelines which scale back startup time and jank in Compose apps.

Google PlayStore’s search outcomes web page has been re-written with Compose. After incorporating the Baseline Profile guidelines from Compose, time to render the preliminary search outcomes web page with photos improved by ~40%.

The Android staff has additionally added Baseline Profiles to related AndroidX libraries. This advantages all Android apps utilizing these libraries. Constraint Format has discovered transport profile guidelines reduces animation body occasions by multiple millisecond.

How one can use Baseline Profiles

Create a customized Baseline Profile

All apps and library builders can profit from together with Baseline Profiles. Ideally, builders create profiles for his or her most important person journeys to make sure that these journeys have constantly quick efficiency no matter whether or not cloud profiles can be found. Take a look at the detailed information on learn how to arrange Baseline Profiles for each app and library builders.

Replace dependencies

In case you are not able to generate Baseline Profiles on your app proper now, you may nonetheless profit from them by updating your dependencies. For those who construct with Android Gradle Plugin 7.1.0-alpha05 or newer, you may get Baseline Profiles included in your APK which can be already supplied by libraries (comparable to Jetpack). Google Play compiles your app with these profiles at set up time. You may complement these profiles as a part of constructing your utility.

Measure Enhancements

Don’t neglect to measure enhancements. Comply with the steps on learn how to measure startup with the generated profile regionally.

Present suggestions

Please share your suggestions and tell us your expertise!