Performance Zone is brought to you in partnership with:

Fabrizio Giudici is a Senior Java Architect with a long Java experience in the industrial field. He runs Tidalwave, his own consultancy company, and has contributed to Java success stories in a number of fields, including Formula One. Fabrizio often appears as a speaker at international Java conferences such as JavaOne and Devoxx and is member of JUG Milano and the NetBeans Dream Team. Fabrizio is a DZone MVB and is not an employee of DZone and has posted 67 posts at DZone. You can read more from them at their website. View Full User Profile

Java Performance at Jazoon Keynote

  • submit to reddit

Just followed the Keynote at Jazoon: William O'Mullane talked about Gaia, the European Space Agency's astrometric platform, and why they mostly use Java for it. I mean, these guys are concerned about performance, right? Well, look at this slide:

Next time you're challenged about the topic, and can't give your own data because of e.g.customer confidentiality, you can refer to this talk (slides will be soon available at the Jazoon site).

Thanks to Mirko Stocke for the shot, I wasn't fast enough to take it.



Published at DZone with permission of Fabrizio Giudici, author and DZone MVB.

(Note: Opinions expressed in this article and its replies are the opinions of their respective authors and not those of DZone, Inc.)


Casper Bang replied on Fri, 2010/06/04 - 10:52am

While it's true that (thanks to HotSpot) Java has't been "slow" since the early decade, that usually assumes only one isolated usage and environment: raw calculation throughput on a powerful server with lots of RAM, where cold-start performance and UI responsiveness is a non-issue. I get the feeling however that many people factor in upstart time, resource demand and responsiveness into this definition as well - let's call it perceived or experienced performance.

It is a fact that much Java UI stuff isn't as snappy as native widgets (that's why SWT was born after-all) and while the difference may be minute today after many years of optimization (and being somewhat OS-specific btw.), I still feel this in practice when using say NetBeans vs. Eclipse.

Drag/lag during start-up is also often a factor - I suspect the very large and bloated runtime (Java 1.7's rt.jar went up to ~57MB/~18K classes) along with a heavy method-based JIT has something to do with it. But to stay pragmatic: compare a JavaFX applet (if you can find one) with Flash to understand why people may resort to using the word "slow" about Java here as well.

Google's Android team would appear to agree with this wider definition, considering they went along to design the Dalvik interpreter to be x2 faster than Sun's and chose a trace-based JIT strategy for faster pay-of. It begs the question whether Sun's client JVM is too close to being a server JVM/JIT, since a -client mode isn't even supported on the platform I am using right now (64bit Linux).

William O'mullane replied on Mon, 2011/02/07 - 8:47am

Yes its all true about GUIs although really many of the C++/NET GUIs are just as slowe on desktops. Still I dont like people taking single slides out of context. I pointed out that we do not have any GUI here and it is all math,   the code is perfect for JIT as it runs over and over again (millions of times on differnt data).

We have also tried differnt JVMs including Jrockit (I think that exampe was Jrockit actually)

It was/is  an amusing anecdote though that out of the box the thing is faster in Java!



Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.