Christian Thalinger

Running Twitter services on Graal has been very successful and saved Twitter a lot of money on datacenter cost. But we would like to run more efficient to reduce cost even more. I mean, who doesn’t? In order to do this we are using our Machine Learning framework called Autotune to tune Graal inlining parameters. This talk will show how much performance improvement we got by autotuning Graal.

With JEP 317: Experimental Java-Based JIT Compiler it is very likely that Graal will be part of JDK 10. In fact, Graal is already available in JDK 9 due to JEP 243: Java-Level JVM Compiler Interface. Graal is itself written in Java and that brings some new properties and behavior to the table which we haven’t seen with existing HotSpot JIT compilers. This talk will show how to use Graal with JDK 9, how to compile an upstream Graal version and what to look out for when using it for benchmarking or even in production.

HotSpot's optimizing Just-In-Time compiler C2 is reaching its end-of-life and it's time to look for alternatives. One very promising replacement candidate is Graal. Graal is a Java JIT compiler written in Java. It is being developed by Oracle Labs for a couple years now and has reached a state where it's viable to be that replacement. Twitter's huge distributed system tickles every corner of the JVM and is the perfect testing ground for a new JIT compiler technology like Graal. This presentation will talk about Twitter's experiences with Graal, good and bad, the bugs we found and also the wins.

Late last year we have integrated JEP 243 into the JDK 9 repositories. This is something completely new to the JVM: a Java interface which is part of the JVM. In this case it is a Compiler Interface (CI) which can be used to plug in JIT compilers written in Java.
I will talk a bit about how it works, what it can be used for and what Oracle is using it for.