A modern programming language that makes developers happier.

kotlinlang

image.png

So I've been converting over a large-ish dagger-android project to use KSP and I am seeing some pretty bad performance regression. Wondering if anyone can provide support it getting the right diagnostics gathered for issue reporting to dagger or ksp projects.

Running the gradle profiler I am showing ksp to be ~3x slower than the same setup with KAPT.

Then in order to get profiling data I've also ran the same with `kotlin.compiler.execution.strategy=in-process` which requires me to increase the heap 3-4x to not throw OOM errors, and that in-process strategy also seems to cause the ksp/kotlinc tasks to run much much slower full builds at that point taking up hours of time.

assembleKsp-8.5-cpu-simplified-flames.svg

the huge time jump from `daemon` -&gt; `in-process` makes me wonder how trust worthy the resulting flame chart even is. but so far it is the only way I've been able to get any vision into how dagger is spending its time while running

1-2 order of magnitude slow down of kapt vs ksp is visible though even on the smaller work units

ended up writing up an issue and tweaking a fork of the santa-tracker as an example of regression to hopefully help.

<https://github.com/google/dagger/issues/4180>