On Thu, 2007-09-27 at 08:16 +0100, Alex Buell wrote:
> On Wed, 26 Sep 2007 14:49:10 +0000, I waved a wand and this message
> magically appears in front of Ferris McCormick:
> > > As for there being a lack of a Java VM on sparc, if anyone's
> > > working on porting Sun's JVM, I can help with testing the builds.
> > Supposedly, this is very hard to hopeless (according to sun). There
> > are more promising alternatives, such as jamvm or perhaps cacao. I
> > have played a bit with jamvm, and I think it could be made to work.
> > It's mostly a matter of figuring out what order the arguments to a
> > couple macros which are used for test-and-swap are given, and making
> > a few memcopy macros work right. I've had it running; just not quite
> > correctly.
> I thknk gcj could do the job just as well. It can compile to bytecode.
jamvm is a java machine, not a compiler.
> > jamvm (or something like it) would make a very nice project. Sun java
> > would probably not (I think sun have it on their to-do list, but
> > considered long-term and difficult).
> I think it's enormously vital that Sun get off their butts and do
> something about it - porting the linuxintel version to linuxsparc
> should work. I've actually dug around in the jvm sources and they do
> have some assembly in various places. These would have to be rewritten
> for the sparc platform.
Interesting. Do you think it would be easier to port from linux intel
than from solaris/sparc? That might make a lot of sense.
(Depending on how much assembly code there is, and whether anyone can
read it. How much is needed outside, perhaps, the atomic operations
like atomic_add or test_and_set?)
> > If you want to be an AT, please set it up with jmbsvicetto or with me;
> > it's mostly a matter of getting you through the ebuild quiz.
> Fearsome grindings.
Thanks and regards,
Ferris McCormick (P44646, MI) <email@example.com>
Developer, Gentoo Linux (Devrel, Sparc)