[4880] in java-interest
The Parallel Java Computer Box
daemon@ATHENA.MIT.EDU (XTALV1@delphi.com)
Wed Jan 17 12:15:08 1996
From: XTALV1@delphi.com
Date: Wed, 17 Jan 1996 10:16:47 -0500 (EST)
To: java-interest@java.sun.com
X-VMS-To: INTERNET"java-interest@java.sun.com"
Is there reliable evidence that somwhere there are some renegade home
workshop hot engineers building a parallel Java computer (PJC) out of
banks of high-end 8051s (or 68HC11s, or whatever)? The nativecode Java
run-time system of the PJC could be written in FORTH on multiple Moore
P21s, for example. Since platform doesn't matter anymore, all future
Java code would run on the PJC. Internet/Web would be the OS. The PJC
would be a consumer-directed object, say $600 a box. It captures your
existing keyboard, mouse, monitor, and modem, and comes with a gigando
hard drive. Since the Java run-time system (array bounds checking,
null pointer checking, I/O, garbage collection, bytecode interpretation,
thread and exceptions management; anything else?) can evidently be
expected automatically to take advantage of multiple processors
(presumably sharing a blackboard memory), all correctly compiled Java
code (by the standard of a commercial release (gamma comes after beta?)),
will run 50-100 times faster than any of the latest n86 CISC chips. Maybe
that, together with the promise of automated bytecode-to-nativecode
converters, will get us off the ground. The PJC automatically melds with
other PJCs by modem (most of which will be on ISDM-type lines, full-time),
in other words, Java applets could conceivably draw on the parallel
computing power of thousands of multiprocessors. For example, a box with
12 CPU chips, times 1000, gives 12,000 threads communicating via shared
RAM. The Sun people like to say, "Have fun." I say, let's laugh ourselves
silly.
Ellis D. Cooper xtalv1@delphi.com http://www.ec3.com
-
This message was sent to the java-interest mailing list
Info: send 'help' to java-interest-request@java.sun.com
