Real memory usage (RSS) is lower and fluctuates. The toolbox max heap is set to 160m.

@vergauwensimon.bsky.social @kotlinlang.org I think the Ktor custom plugin doc need some refactorings - ktor.io/docs/server-.... We can now supply the config from file easily using the new config deserialization support and can make the config a lot simpler.

I'm glad this is somewhat fixed in the new Kotlin Amper build tool; it seems they try to parallelize things by default. I sometimes struggle with Gradle, even after working on it for a while. 😬

#Kotlin
Compose Desktop apps can also benefit from this feature. A quick test showed improved startup times. In combination with compact object headers (openjdk.org/jeps/483) and GenZGC, desktop UI applications on the JVM can become much smaller and more efficient in the near future.

The new AOT class linking feature in Project #Leyden, which is now available in the latest #OpenJDK 24 Early Access build, is quite promising. Despite the code cache not being implemented yet, I've already seen a significant 2.5x startup time improvement even on a basic #Kotlin #Ktor application.

Beware of updating to the latest Copilot IntelliJ plugin. This is what the resource usage of their newest copilot-language-server looks like, ~8GB memory and 1000 platform threads 🙄

Ah, yes, the classic 'why not just' question 🙄

openjdk.org/jeps/450 just landed on JDK 24. That's going to help with in the short term until we get Valhalla, that will remove the whole object header altogether.

Here is the new object layout with JEP 450 (Lilliput), shared by Aleksey Shipilëv

Nice... modified with sealed class for exhaustiveness