Release OpenJDK

[axenzoku]

OpenJDK

Ok, this is a fully open-source Java. Sun has even said it works just as good as regular java, it's just open source, and from now on OpenJDK will be where their releases come from... So it seems there is yet a chance that the Pandora can have an official Open-Source Java package endorsed by Sun themselves. (If anyone can port it)

Download links right on the official sun website.

 

[dflemstr]

Dude,

1. OpenJDK is the standard JDK on Linux. It already exists everywhere
2. The official Sun JDK is going to be dropped soon, and only the OpenJDK branch is going to be developed
3. These are old news that have existed since Sun became open source, in ~2003
4. Ångström already has OpenJDK and it has always been the plan to include it on the Pandora
5. OpenJDK != a VM, it's only the libraries. We'll still need a good JVM.

 

[Klaus]

Dude,

1. OpenJDK is the standard JDK on Linux. It already exists everywhere
2. The official Sun JDK is going to be dropped soon, and only the OpenJDK branch is going to be developed
3. These are old news that have existed since Sun became open source, in ~2003
4. Ångström already has OpenJDK and it has always been the plan to include it on the Pandora
5. OpenJDK != a VM, it's only the libraries. We'll still need a good JVM.

HotSpot & Cacao seems torun on the BeagleBoad, we should be fine. http://elinux.org/BeagleBoard#Java

 

[dflemstr]

HotSpot & Cacao seems torun on the BeagleBoad, we should be fine. http://elinux.org/BeagleBoard#Java

Cacao? Wow, that's great news! I couldn't have asked for anything better! I thought that it didn't work on ARM for some reason.

 

[Kramy]

It says Hotspot runs in interpreted mode, which means it'll probably be a good 10x slower than running on x86 Windows. Less memory usage, though.

What about JamVM? I recall seeing an article where JamVM took the lead in every test. Despite being fully interpreted, it's pretty lean and mean.

 

[dflemstr]

It says Hotspot runs in interpreted mode, which means it'll probably be a good 10x slower than running on x86 Windows. Less memory usage, though.

What about JamVM? I recall seeing an article where JamVM took the lead in every test. Despite being fully interpreted, it's pretty lean and mean.

Well, it has very nice startup times, that's for sure, but it has reached it's limits, so to speak: to improve its performance, they have to have JIT.
And Java benchmarks can be very misleading; if they only do, say, in-place maths, or only blocking I/O calls to test performance, then of course JamVM gets good results because you wouldn't gain much performance by JITing anyways. I'd like to see a test that tests object creation times (say, spawn 1000000 objects a second and see how it reacts) to be able to make a verdict.

 

[samuraicrow]

Maybe the Shark JIT will work from the Zero project of IcedTea. They are processor independent. http://llvm.org/ProjectsWithLLVM/#icedtea

 

[Kramy]

I'd like to see a test that tests object creation times (say, spawn 1000000 objects a second and see how it reacts) to be able to make a verdict.

I'd like to see that too.

Anyway, relevent links:

http://www.advogato.org/person/rlougher/diary/9.html (check out the old buglabs benchies - even the follow up is promising.)
http://www.advogato.org/person/rlougher/diary/10.html (Not quite sure what this is)
http://www.advogato.org/person/rlougher/diary/12.html (Performance looks good on a Beagle. )

Originally I had links to other sites, but then I discovered all were already posted on his blog. :lol:

 

[Guy Montag]

What about using GCJ to compile java code to native machine code? There would be no need for a jvm.

But I don´t know if it works on ARM.

 

[Klaus]

What about using GCJ to compile java code to native machine code? There would be no need for a jvm.

But I don´t know if it works on ARM.

While it would work for free Software, packaged for the Pandora, it wouldn't work with WebStart, Applets or downloaded Java-programs from somewhere on the web. A JVM would be nicer.

 

[dflemstr]

What about using GCJ to compile java code to native machine code? There would be no need for a jvm.

But I don´t know if it works on ARM.

While it would work for free Software, packaged for the Pandora, it wouldn't work with WebStart, Applets or downloaded Java-programs from somewhere on the web. A JVM would be nicer.

Is this entirely accurate?
Yes, it wouldn't work for WebStart or Applets, but it would work for any other kind of Java program, even if it's packaged in JAR or in the compiled byte-code format. GCJ can compile anything into machine code... (And why wouldn't it be able to; JIT-compilers in JVMs can do it after all)

 

[Klaus]

Is this entirely accurate?
Yes, it wouldn't work for WebStart or Applets, but it would work for any other kind of Java program, even if it's packaged in JAR or in the compiled byte-code format. GCJ can compile anything into machine code... (And why wouldn't it be able to; JIT-compilers in JVMs can do it after all)

Oh, I didn't know that GCJ could compile bytecode into machinecode, but it makes sense. Thanks.

 

[Guy Montag]

I tried to compile some of my Java games with gcj. I´m not happy with the results. GCJ seems to be pretty useless for code that contains Swing or AWT components. Even a simple Space Invaders clone didn´t compile.

 

[PhonicUK]

How about the Blackdown JRE? Has anyone tried compiling that for ARM?