Missed the boat on this one?
Oracle and AMD propose GPU-harnessing Java project
OpenJDK project, proposed by Oracle and AMD, aims to establish
whether it is possible to implement GPU support into Java with a
Essentially, this Oracle-led initiative seeks to find out whether Java application performance could be improved by utilising the GPU for computational power, and not just for graphics. This follows in the footsteps of competitors of Microsoft, Mozilla and Google – all of whom have turned to hardware to eke out some GPU browser power
Hotspot’s Group Lead John Coomes, alongside AMD’s Gary Frost, made the proposal on a OpenJDK mailing list, detailing why he believes this investigation is critical for Oracle’s JVM, the most commonly used virtual machine by Java developers:
This project will demonstrate the performance advantages of offloading Java compute to a GPU. We propose to use the Hotspot JVM, and will concentrate on code generation, garbage collection, and runtimes. Performance will be improved, while preserving compile time, memory consumption and code generation quality.
The proposal also states the intentions to bring in Java 8 Lambda library features along the way, but it is also quick to recognise the Java API challenges that could emerge such as extensions. This would mean new standards would have to be created through the Java Community Process, and this has also been taken into account.
AMD are putting their weight behind the project, with Frost pledging committers from his development team. The GPU project would maintain one or more code repositories within the larger OpenJDK HotSpot repository and maintain a developers mailing list.
With GPU compiler projects like Rootbeer cropping up recently, there's a clear demand for greater graphics control in Java. Now seems the ideal time to pursue the option of segmenting Java processing into GPU and CPU, to lighten the load. Oracle recognises the need to accelerate performance, to accommodate for more demanding graphics within apps on smartphones and tablets. But have they moved too late? The rest of the competition have already done this.
But the biggest stumbling block could well be the length of time it would take to see this standardisation. OpenJDK certainly has noble intentions but as recent events have shown us, it takes a fair bit of time for anything to get close to getting into Java.
Pessimism aside though, it’s a bold and needed proposal that should be applauded for its intentions. Whether we see it come to fruition is another matter.