From: Ben Goertzel (ben@webmind.com)
Date: Tue Jul 17 2001 - 20:19:19 MDT
Not to get too far carried away with this thread, but there are a lot of
annoying subtleties here.
For instance, a 64 bit JVM gives you more memory, but it also causes each
object to *consume* more memory, by using 64 bit ints, floats, etc.
Now, one can get around this via cleverness, e.g. by using bit-masking to
store parameters in ints (a 64 bit int stores more parameters this way than
a 32 bit one). However, if one has to do this kind of "virtual pointer
pushing" to get Java to perform decently, one has to wonder, why not just
use C++? ;>
ben
> It was true until very recently for all Windows and Linux *and
> Solaris* JVM's. Sun claims they'll have a 64 bit JVM coming out
> sometime, but the launch date has been frequently postponed.
>
> I haven't tried IBM's new 64 bit JVM, actually, though I have
> read about it, e.g.
>
> http://service2.boulder.ibm.com/devtools/news0101/art14.htm
>
> Interesting work at IBM on scalable Java is at
>
> http://www.research.ibm.com/journal/sj39-1.html
>
> All this is quite new, and I don't know how well it works.
> Experience shows that marketing literature is a poor guide to
> actual functionality for such systems. I do know IBM has
> invested a lot of time and money in this.
>
> For us, when we decided to re-implement Webmind to change the
> *architecture* (while keeping the mathematical and conceptual
> basis the same), it was a tough choice whether to stick with Java
> (in light of all the progress IBM and others are making, albeit
> slowly) or to go with C++. But ultimately, we decided that if we
> were going to re-code the system for efficiency, we might as well
> go all the way.
> >
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:36 MDT