Re: Simple Forth processor witten in VHDL.
- To: "Myron Plichota" <myron.plichota@xxxxxxxxxxxx>, "Don Golding" <angelus@xxxxxxxxxxxxx>, <misc>, "John Rible" <jrible@xxxxxxxxxxxxxx>
- Subject: Re: Simple Forth processor witten in VHDL.
- From: "vic plichota" <atsvap@xxxxxxxxxxx>
- Date: Sat, 12 Feb 2000 11:46:40 -0500
- References: <200002121447.JAA03028@smtp13.bellglobal.com>
From: Myron Plichota <myron.plichota@sympatico.ca>
> A 12-bit design might be worthwhile if
> you want to conserve target chip resources.
Don Golding:
> > What I would like to see is a simple 8 bit Forth engine
> > with a 64K address space that is writen in VHDL
...Or an 8-bit machine with 256 bytes address space (double that,
with a dedicated 'Harvard' program bus) -- no bloody good for a web
browser, but you could sure build a fancy washing machine, or house
thermostat!
> > option of creating special purpose co-processors
Not sure what Don is driving at, here (floating point? A/D? serial
link?), but it seems an inescapable fact of life that above a certain
size / sophistication, the CPU complexity experiences a sudden
quantum-growth:
? - does the target machine really need it's own native interpreter /
compiler? Maybe not; it's surprising what can be accomplished (most
conveniently) for cross-development with even a lowly PC/XT.
? - At what point does the hassle of encoding, aligning, and issuing
small opcodes (within a relatively wide bus) become more trouble than
it's worth? Given a plethora of memory, it's simpler to waste the
extra bits, and to hell with any alignment / compaction mechanisms.
No architecture is perfect, but I've seen some that come awfully
close -- it all depends on exactly what sort of jobs you intend to
throw at it.
My favorite methodology (these days) is to take radically different
CPU's (each tailor-fitted for a specific regime) and get them to
communicate. Admittedly, this usually involves off-the-shelf
merchant silicon from "the big boys", but hey, I have bills to pay
(and big-boy merchant silicon has some awesome capabilities)... I am
NOT averse to using a design from the MISC list if it is truly cheap
+ easy -- and suited for my particular job.
The big flaw with big-boy silicon, is that it is tending more and
more towards a big investment in time and intellectual dedication --
and you are stuck dead with the development tools that are offered --
whereas a homebrew solution (for a similar investment) provides the
opportunity for complete control; if it's broke, you can fix it
yourself!
...So I guess I am trying to say that it's not just the design, but
the whole politics + business target that makes concocting a fresh
CPU such a challenge. Why bother?
WHY NOT??? :-)
cheers - vic