OOPSLA Java VM panel notes

[Thomas Poff]

Someone said:

>Modern processors are plenty fast enough for OO. You just have
>to glue the appropriate compiler technology to it and do live variable
>analysis, basic block evalution, code motion, inlining etc.
----

This is interesting and mostly true but assumes you are in the  
workstation/PC world and _are_ using modern processors.

There are many things you can do to a processor to make it execute faster  
that are just plain old being ignored... primarily because of the huge gap  
between certain s/w engineers and ASIC folks.

A _lot_ of folks are finding their processors fitting on chips with large  
amounts of gates going completely unused these days.  Why not then start  
fitting parts of the object-based runtime on board with a processor in the  
gate array?  I'm talking about object tracking mechanisms, IORQ, messaging  
and even the frameworks themselves.

I can't go into details, but essentially you can take a very low-horsepower  
processor or current-limited system, and manage a very complex view-system  
that executes with very low overhead.  Typically it is the view-system on a  
java-device that you want to accelerate... at least if you're trying to build  
product.

I believe it's the folks who can make some useful Java gizmo at a $50-$100  
pricepoint or less that are going to begin to see some real productization.   
If you're using 32-bit processors and "lots" of RAM then that's going to be  
pretty tough.  So if I wanted to build a Java-enabled consumer device, I'd  
prefer to look at an 8 or 16 bit processor core and glue what I could to it  
to get it to fulfill a required set of tasks.

BTW... from what I have garnered, corporations (OEMs) don't want Java  
processors, they want their own/favorite processors running Java with some  
kind of acceleration available.  And frankly, every time I look at Patriot  
Scientific and the Java cores from Sun and Rockwell I can see why.  Systems  
with their processors look big and bulky.  They still have a long way to go.