[MachineForth] 25x Forth engine
- Subject: [MachineForth] 25x Forth engine
- From: Jeff Fox <fox@xxxxxxxxxxxxxxxxxxx>
- Date: Sun, 25 Nov 2001 11:50:29 -0800
- Organization: UltraTechnology
David K Walker wrote:
> Then, I am going to replace
> the GUI with any standard browser anywhere in the network,
> and the replicated database with a mixture of my own code
> and cheap, open database software or optionally any data
> base software that PHP will talk to. Leaving the Forth
> middleware as the interesting service for me to develop
> on ever-cheaper hardware. And new middleware.
Interesting vision, not unlike some other forward looking
strategies from major players.
> The issue as I see it for bringing Chuck's designs to
> bear on this development, is new applications that are
> very compute-intensive and very parallel.
Yes, certainly. Compression/decompression, encryption,
decryption, protocol translation, routing, simulations,
hardware emulation, all seem to fit.
Also, think of 25x as a sort of new type of programmable
part. It has a huge amount of processing power for
computation tasks or for doing things very fast on I/O
pins. Exactly what all those pins do is mostly software
so the part could do the job that a lot of other
specialized systems do now but cheaper or faster or
whatever. Software defined hardware emulation as it
were.
Many Forth programmers are familiar with the idea of
trading software for hardware. Most programmers work on
a layer of software far far removed from the
hardware/software layers. Only a few programmers write
drivers let alone design CPU or compilers or OS etc.
or defined new hardware parts using software...
It is hard to tell someone "exactly what" an FPGA
part can do. It is hard to tell someone "exactly
what" any given hardware can do by varying the software.
It is hard to tell someone "exactly what" DNA
can do either. People want to know, but no one knows.
The 25x is not easy to understand, perhaps much
harder to really understand than F21, because the
"coprocessors" are almost more software than hardware.
The inside is both hard and soft and the outside is
both hard and soft.
It is obviously designed to exploit parallelism in
many ways. One way is as a scalable cluster. Use
all c18 core more or less in the same way...
But it could be a stand-chip programmed as a very
fast very cheap hardware part. It could be a stand-
alone computer system with a RAM chip and some
software, but still 256K 18-bit words of SRAM
(as an example) is not enough to run software
that requires megabytes. That is enough for
cheap network appliances with iTV style software
and software modems and network interfaces and
maybe even software generated video interfaces.
There is enough memory for some types of routers
for instance that today cost more than $1.
It appears that the external 18-bit SRAM interface
is just one way to program the ROMs on 25x,
and is more defined by the size of the chip and
the number of pins than anything else since
software is so variable. Even the 5x5 matrix
on 25x is just a software definition at one
level in OKAD.
Jeff Fox
------------------------
To Unsubscribe from this list, send mail to Mdaemon@xxxxxxxxxxxxxxxxxx with:
unsubscribe MachineForth
as the first and only line within the message body
Problems - List-Admin@xxxxxxxxxxxxxxxxxx
Main Machine Forth site - http://www.