Hi,
I’m new to Art Of Illusion and I experimented thed ay before yesterday at home with BEA JRockit 1.4.2 on Windows/XP , ia32 : very nice !
Then I tried to use BEA JRockit 1.5.0_04 and I got a crash of the JVM at start of Art Of Illusion.
Then I experimented on Linux FC2, ia32, with BEA JRockit 1.4.2 and 1.5.0_02 : OK. With 1.5.0_04 : again a crash.
I’ve warned BEA guys: http://forums.bea.com/bea/thread.jspa?threadID=600007913&tstart=0, but they say it seems to be a problem in JOGL.
Can someone have a look to this ?
Regards,
Tony
Do you have any information like a crash dump or a stack trace from the crash?
If things are working well with Sun’s JDK then I would recommend you use that, and the onus would be on BEA to prove that it’s a problem in JOGL.
Yes, I have several dumps.
I’ve attached one I got on my FC2 ia32 machine.
BEA thinks the problem is in JOGL …
I’m discussing with them.
Regards,
Tony
I know BEA JRockit should go faster …
At least JRockit 1.4.2 and 1.5.0_02 work well with AoI and JOGL.
So I guess the problem is in their new JVM.
Tony
Here is a dump from Windows/XP.
Tony
I would dispute claims that BEA’s JVM should be faster than HotSpot. Have you tried with the latest Mustang builds and -server?
I don’t see JOGL or the app doing anything obviously illegal so my guess would be that BEA has somehow broken native method calls or JNI support like GetDirectBufferAddress in their latest release.
Not yet. I’ll do that with AoI later. Have you played with JRockit previously ?
[quote]I don’t see JOGL or the app doing anything obviously illegal so my guess would be that BEA has somehow broken native method calls or JNI support like GetDirectBufferAddress in their latest release.
[/quote]
I’m in a private discussion with BEA guys and I’m waiting to know where is the problem.
Tony
Not personally. However I work in the Java HotSpot VM group at Sun Microsystems and some of our recent internal benchmarking results contradict BEA’s claims of JRockit being faster than HotSpot.
The problem was indeed in JRockit and has been tracked down. See http://forums.bea.com/bea/message.jspa?messageID=600016675&tstart=0.
Regards,
/Staffan
Of course which JVM is faster always depends on what you are measuring on - and “our internal benchmarking results” is hardly very specific. JRockit isn’t written to be the fastest JVM for client applications - but watch it on the server!
Regards,
/Staffan - yupp, I work on JRockit
[quote=“Ken Russell,post:8,topic:25568”]
Interesting ! When do you plan to publish results ? Which benchmarks are you using ? SPEC benchmarks ? SPECjbb2005 is now much more close to real applications, I’ve been said. And what about SPECjAppServer2004 ?
BTW, I no more see a Sun JVM for Itanium2 (5.0 update 6). Have you decided to stop delivering such a JVM for WIndows and Linux ?
Tony
I can’t make any further comment regarding the comparative performance of the two JVMs. I can confirm that Sun is not shipping Java SE 5.0 for Itanium.