Epiphany OpenJDK support

[over9000]

Michal: I agree best bet is on OpenJDK.

I contacted Ross McIlroy and he agreed to share his code on github: https://github.com/rmcilroy/HeraJVM
Thanks Ross.
But the code was made on the old Jikes RVM( version 2.9.3) and the second problem is that Jikes RVM does not run on ARM processors. But there might be some precious gems to pick out.

Wow. That's really interesting. A big thanks to everyone involved for bringing this code onto github. I'm guessing that for the moment, it's only of immediate interest to people who want to try it on the PS3. Since I don't have my Parallella yet, I'll be very interested in doing that.
Thanks again!

[michal.warecki]

Here is the article I wrote few days ago:
http://michalwarecki.blogspot.com/2013/ ... lella.html

Now I have some free time for coding the compiler

Best,
Michal

[8l]

if jvm available on parallella, then scala also available on it.
yay! thanks for the info.

[Gerald]

Hi any news on this? Can you already run java on the paralella board?

[michal.warecki]

Hi Gerald,

It is a long way to support Java on Parallella. For example the whole team from AMD on full-time job worked on HSAIL compiler for a year or even more. They are still working on a dispatching mechanism.
You have to be very patient (I'm programming the compiler in my free time)

Best,
Michał

[9600]

Hi any news on this? Can you already run java on the paralella board?

Just to be clear: you can run Java on ARM processor, of course.

Cheers,

Andrew

[notzed]

Here is the article I wrote few days ago:
http://michalwarecki.blogspot.com/2013/ ... lella.html

Missed this - looks interesting. And rather challenging ...

If it would help in any way I wouldn't necessarily stick to the C abi (register usage) since you will probably need your own on-core runtime anyway - apart from LR anything else can be reassigned (some global context registers might be useful). But I guess it depends on if the runtime will need to call C compiled code often.

Does sumatra (or even lambdas) have a more relaxed/different memory model for actively running code? I don't see how it could work very well with a gpu if it didn't. e.g. with opencl there's no guarantee on how many workgroups are even executing concurrently so there's no way you can write inter-workgroup communications (amongst other things).

[Gerald]

hi,

are people working on this? any github?

Thanks,
Gerald