Universal APK performance implications

[maman]

My team had built a react native app for internal usage. We use RN 0.79.1, with both hermes and new architecture enabled.

The target platform is Android, and one of the device that we target is a Portable Data Terminal (PDT) running 32 bit android 8.1.0.

When we run the app on this device, it runs very, very slow.

I never replicate the problem on my test device and emulator since both my test device and emulator runs arm64 android (m1 macbooks didn't support arm32 emulator disk image).

Out of frustration, I tried to do a split apk build, and run the armeabi-v7a apk on the PDT.. and it runs smooth.

Until now we deploy the universal apk instead, since the device target is diverse. But in this case, doing architecture-specific build seems to resolves the issue.

Is there any perf implication when doing universal apk? or rather, how can I debug this further?

I'm curious about this since AFAIK "Universal Apk" means that we bundle libs for all supported architectures inside the apk, then Android has some heuristic to only load the architecture specific lib on runtime 🤔

[cortinico]

Is there any perf implication when doing universal apk? or rather, how can I debug this further?

I'm not aware of any perf implication of universal APKs.
Universal APKs have all the libraries available, split per architecture in separate folders.
The device should just load the needed library for a given architecture.

Are you sure you're not accidentally building a debug Universal APK? As that could explain the slow-ness.

Anyway this is more of an Android general question that you can ask here https://issuetracker.google.com/