OpenJ9 VS Graal

Source

I keep forgetting about J9 but they're not doing themselves any branding favors since there actually is J17 on J9 :-/ https://github.com/eclipse-openj9/openj9/blob/openj9-0.35.0/... (Also that 0.35 versioning ...)

As best I can tell, these are the docker images: https://hub.docker.com/_/ibm-semeru-runtimes

    $ docker run --rm ibm-semeru-runtimes:open-11-jdk java -version

On another note, I'm still not sure if there is a viable way to microbench code running on OpenJ9. It seems that there is still no official support from JMH, at least I'm getting warnings such as "This VM is not supported by JMH. The produced benchmark data can be completely wrong". Apparently it should work, however, my results for runs on OpenJ9 show (by a large margin) much higher variance compared to Hotspot which doesn't exactly inspire confidence.

-Xjit:disableGuardedStaticFinalFieldFoldingFlat out improves performance, working around a bug in -XaggressiveEnables performance optimizations and new platform exploitation that are expected to be the default in future releases of OpenJ9. -Xmns128M -Xmnx1024MSets minimum and maximum size of the nursery for the gencon (default) GC. Having a small nursery allows garbage collection to be really fast, especially with how many short lived objects Minecraft makes. These values shouldn't need to be changed.If you want to know more about the gencon GC and its nursery and tenure zones you can find something here. -XdisableexplicitgcDoesn't allow mods to force a full garbage collection. Removes some lagspikes from misbehaving mods. -Xgc:concurrentScavengeLets gencon GC collect garbage in the background, without stopping the game thread to do it. Gives a very noticeable boost to "smoothness". -Xgc:dnssExpectedTimeRatioMaximum=95 -Xgc:dnssExpectedTimeRatioMinimum=70Lets gencon GC know that it's gotta spend most of its time cleaning up the nursery, instead of the rest of the heap. Most of the garbage is in the nursery instead of the tenure zone so this works incredibly well on modded MC.

I like this part "We continue to employ dozens of developers that work directly and openly in the Eclipse OMR and Eclipse OpenJ9 projects at GitHub. IBM doesn’t produce a separate enterprise version of OpenJ9; we don’t hold back any of the innovation in our runtime."

OpenJ9 (heavily based on OpenJDK, especially later versions): https://github.com/eclipse/openj9/blob/master/jcl/src/java.base/share/classes/java/lang/Throwable.java

In 1.18.16 they "added support for compiling projects with OpenJ9". Turns out the hack to access Hotspot's sun.misc.Unsafe doesn't quite work with OpenJ9 . Oh, really? So surprising. This is exactly the reason why the OpenJDK project pushes their encapsulation agenda so hard!

Contrary to what vocal Kotlin advocates might believe, Kotlin only matters on Android, and that is thanks to Google pushing it no matter what.

https://spectrum.ieee.org/the-top-programming-languages-2023

https://snyk.io/reports/jvm-ecosystem-report-2021/

And even so, they had to conceed Android and Kotlin on their own, without the Java ecosystem aren't really much useful, thus ART is now updatable via Play Store, and currently supports OpenJDK 17 LTS on Android 12 and later devices.

As for your question regarding numbers, mostly Java 74.6%, C++ 13.7%, on the OpenJDK, other JVM implementations differ, e.g. GraalVM is mostly Java 91.8%, C 3.6%.

https://github.com/openjdk/jdk

https://github.com/oracle/graal

Two examples from many others, https://en.wikipedia.org/wiki/List_of_Java_virtual_machines

Pkl was built using the GraalVM Truffle framework. So it supports runtime compilation using Futurama Projections. We have been working with Apple on this for a while, and I am quite happy that we can finally read the sources!

https://github.com/oracle/graal/tree/master/truffle

Disclaimer: graalvm dev here.

That's pretty interesting. It's not as aggressive as Bee sounds, but the Espresso JVM is somewhat similar in concept. It's a full blown JVM written in Java with all the mod cons, which can either be compiled ahead of time down to memory-efficient native code giving something similar to a JVM written in C++, or run itself as a Java application on top of another JVM. In the latter mode it obviously doesn't achieve top-tier performance, but the advantage is you can easily hack on it using all the regular Java tools, including hotswapping using the debugger.

When run like this, the bytecode interpreter, runtime system and JIT compiler are all regular Java that can be debugged, edited, explored in the IDE, recompiled quickly and so on. Only the GC is provided by the host system. If you compile it to native code, the GC is also written in Java (with some special conventions to allow for convenient direct memory access).

What's most interesting is that Espresso isn't a direct translation of what a classical C++ VM would look like. It's built on the Truffle framework, so the code is extremely high level compared to traditional VM code. Details like how exactly transitions between the interpreter/compiled code happen, how you communicate pointer maps to the GC and so on are all abstracted away. You don't even have to invoke the JIT compiler manually, that's done for you too. The only code Espresso really needs is that which defines the semantics of the Java bytecode language and associated tools like the JDWP debugger protocol.

https://github.com/oracle/graal/tree/master/espresso

This design makes it easy to experiment with new VM features that would be too difficult or expensive to implement otherwise. For example it implements full hotswap capability that lets you arbitrarily redefine code and data on the fly. Espresso can also fully self-host recursively without limit, meaning you can achieve something like what's described in the paper by running Espresso on top of Espresso.

I'm also using GraalVM if that's of any help.

Quarkus is one of Java frameworks for microservices development and cloud-native deployment. It is developed as container-first stack and working with GraalVM and HotSpot virtual machines (VM).

Apologies, I didn't mean to imply DCEVM went poof, just that I was sad it didn't make it into OpenJDK so one need not do JDK silliness between the production one and the "debugging one" since my experience is that's an absolutely stellar way to produce Heisenbugs

And I'll be straight: Graal scares me 'cause Oracle but I just checked and it looks to the casual observer that it's straight-up GPLv2 now so maybe my fears need revisiting: https://github.com/oracle/graal/blob/vm-23.1.0/LICENSE

> to be compiled to a single executable is a strength that Java does not have

I think this is very outdated claim: https://www.graalvm.org/

https://github.com/oracle/graal/issues/7182