With KSP2, I’m finding that the <CatchUp> project ...

# ksp

[Zac Sweers]

With KSP2, I’m finding that the CatchUp project requires nearly 4g of memory compared to2.5gb on K1/KSP 1. I know KSP2 requires more memory, but this seems… extreme. Is this intentional?

[Ting-Yuan Huang]

The upcoming release will reduce a good amount of memory when the compile classpath is large. KSP 1.0.20 and before used a workaround that loads the entire compile classpath into memory. The workaround is no longer needed after an upstream improvement. Context: https://youtrack.jetbrains.com/issue/KT-66689

[Ting-Yuan Huang]

Other than that, we'll try to use CatchUp to profile CPU and memory usage and see what can be done.

[Ting-Yuan Huang]

btw, the removal of aforementioned workaround will also improve startup time considerably, again, when the classpath is large.

[Zac Sweers]

Awesome!

[Zac Sweers]

Oof, that’s missing 2.0 release it seems

[Ting-Yuan Huang]

There will be a corresponding release for Kotlin 2.0, with the fix.

[Zac Sweers]

ah no I mean that the issue you linked is targeted at 2.0.20-Beta1

[Zac Sweers]

not 2.0

[Zac Sweers]

are they going to cherrypick it?

[Ting-Yuan Huang]

KSP2 bundles some near tip-of-tree Kotlin compiler builds so no cherry-pick is needed.

[Zac Sweers]

👍

[Zac Sweers]

any chance of a release sooner than 2.0.0? Currently KSP2 is far from usable in our nontrivial projects, would love if there was another release we could test against before the final release

[Jiaxiang]

can you list the blocking issues for your use cases so I can prioritize?

[Zac Sweers]

• https://github.com/google/ksp/issues/1861
• https://github.com/google/ksp/issues/1842
• https://github.com/google/ksp/issues/1845
• https://github.com/google/ksp/issues/1846
• https://github.com/google/ksp/issues/1839
• https://github.com/google/ksp/issues/1843

[Jiaxiang]

#1841 and #1843 should be fixed in a patch coming out soon.