hey folks, I've finally gotten the slack android r...

# k2-adopters

[Zac Sweers]

hey folks, I've finally gotten the slack android repo to a benchmarkable state with K2 and unfortunately it does not look good for performance. The TL;DR is it's ~17% slower for us. Going to continue working through a few scenarios (this one is a non-abi kotlin change to a low-level subproject that should be pretty IC-friendly). But definitely feeling surprised right now

[tapchicoma]

With which Kotlin version are you testing? Is it KMP project?

[Zac Sweers]

2.0.0-RC1, not a KMP project. Mostly android-heavy project with ~850 subprojects

[tapchicoma]

would be helpful if you could share Kotlin build metrics report in json format + Gradle build scan for RC1

[tapchicoma]

and does project uses KAPT?

[Zac Sweers]

Will try to collect them. It does use kapt but in only 3-4 subprojects

[gildor]

I think Slack uses compile plugins a lot, at least Anvil.
Maybe it also affects compilation time

Also a case for us, so curious about this

[Zac Sweers]

Anvil runs almost entirely via KSP now. The only compiler plugin we run from it is IR merging, in the same projects that run kapt

[gildor]

But KSP is also a compiler plugin, isn't it?

[Zac Sweers]

KSP2 is not

[Zac Sweers]

But fair KSP 1 still is

[Zac Sweers]

But it’s running in its own task too

[Edoardo Luppi]

@Zac Sweers hey! Just wondering if you've managed to understand what was/is going wrong on the K2 side

[Zac Sweers]

No. Results seem to vary widely across repos