> I've been thinking about the possibility of running JOS on a Java processor.
> Being able to run Java on a Java processor (i.e. achieving the speed of
> native code) is kind of like ultimate goal of Java as far as I'm concerned.
As a matter of fact, as slow as the current picoJava cores are,
and as fast as the say, K7, is, you'd be better off running a good JVM
(jikes, hotspot). You'd get better performance running your Java apps on
a fast-enough 'conventional' computer -- though not as good as native code
on the same machine, most machines are so grossly over-powered for the
work they're doing it's rather disgusting. (The only way I can slow the
box I'm working on down enough (running Linux) to want more power is to
run seti@home in the background (nice -10), which chews 98.6% of the
processor :)) While it's no excuse for poor coding, the muscle available
to us (and more so when jos release 1.0 is ready :)) limits my worry about
the speed of things...
> I.e., is the design of the JOS kernel(s)/JVM(s) currently suited for
> execution on Java processors?
Well, the kernel, small as it is, shouldn't be too hard to port,
thought it's not my area of expertise. However, I don't have the
slightest idea of what kind of work the hardware does for handling
bytecodes... if things like constant pool resolution generate hardware
traps, it's more ofa kernel job than a JVM one... while some of the same
code might be useable, the port would be so far off the beaten path that
it couldn't be maintained with the rest of them, I would think.
However, I have more than enough work waiting for me to finish in
decaf for the i386 to squash anything more than idlest speculation.
-_Quinn
_______________________________________________
Kernel maillist - [EMAIL PROTECTED]
http://jos.org/mailman/listinfo/kernel
