Anyone here tried to benchmark build times after migrating to Hilt? I tried migrating a project, but...

# dagger

[allan.conda]

Anyone here tried to benchmark build times after migrating to Hilt?
I tried migrating a project, but the benchmarks show Hilt is significantly slower.
I’m going to investigate if there’s just something wrong with my setup; and I’d love to hear your experiences if anyone can share 🙏

[Colton Idle]

What exactly is slower?

[ritesh]

I dont think there is any performance difference with dagger-android vs hilt.
It's same object graph, same dagger stuff. Hilt is just an abstraction over dagger-android which does things for you. Less boilerplate.

[allan.conda]

We don’t use dagger-android, just dagger.
Just saw this page and it’s probably due to using subcomponents
https://dagger.dev/hilt/subcomponents-vs-deps

One of the main advantages of component dependencies is building Dagger code separately and in parallel.

…

Since Hilt has to aggregate modules, all components would be generated at the same time anyway so we wouldn’t be able to take advantage of building in parallel.

[wasyl]

You mean it’s slower to compile, not in runtime, right?

[allan.conda]

Ah, yes. sorry the for confusion

[allan.conda]

Triggering an incremental ABI change (add a public method) on a feature module triggers the dexBuilder task and kapt on the app module which weirdly take a long time and take most of the time of the total build duration after migrating to Hilt. I verified incremental kapt is enabled for both modules though.

[Karthikeyan1241997]

Use incremental annotation processing

[allan.conda]

I verified incremental kapt is enabled for both modules though.

I made sure it is enabled, I verified from kapt.verbose logs