home .. forth .. misc mail list archive ..

Re: questions (about x21)


This is a composite reply to some of the recent dialog on the MISC list.

***
Greg:

There are indeed many documents on Jeff's website that flesh out the
technical picture for both P21 and X21. You will find some of the P21 info
you are looking for at:

http://www.ultratechnology.com/p21intro.html

In "Thoughtful Programming", Jeff's remarks on the state of the admittedly
incomplete documentation stresses the experimental nature of the venture
at this point. It's only natural to skip the effort involved in producing
definitive documentation under the circumstances. We are used to seeing
datasheets from the semiconductor manufacturers that at least on the
surface offer instant gratification, but actually refer the reader to other
documents that aren't even necessarily available on-line. One thing we
never see from the establishment is datasheets with the status of
EXPERIMENTAL. What we see instead is ADVANCED or PRELIMINARY
or FINAL, and we still need to be vigilant about chip errata, which are
usually published in separate documents.

One thing I have always admired is the honesty in reporting the problems
encountered in the evolution of x21 chip versions. It takes guts to admit
to less-than-perfect results, and the readership has the benefit of being
informed of the failure analysis and proposed solutions.

I think the best way to research x21 is to accept the fact that one needs
to "camp out at the library" over as many browsing sessions as necessary,
and not expect everything to pop out all at once.

To add another ironical perspective to the documentation issue, I have had
the experience of several ASIC and CPLD designs in a corporate
environment that followed the pattern of:

1) definitive target spec and design roadmap
2) target spec review and signoff
3) detailed design and simulation
4) engineering sample verification
5) production

I found much to my surprise that the up-front documentation I was charged
with producing was with few exceptions given only a cursory review by
"concerned" colleagues and supervisors, and that it wasn't until stage 5
that I got questions like "How come you didn't include a blah blah blah?"
and also constant questions about stuff that was fully documented already!
Although I still value good documentation, it is surprising how far short
of the ideal the audience's attention span, comprehension, and retention
can be.

***
Jeff:

Thanks for pointing out the existence of CORDIC sin/cos routines in the P21
boot ROM release. Although I have no doubt about the quality of the
compiled code and the instructive value of the source code, I still want to
explore the issues back to the fundamental algorithms themselves, which is
what I always thought Forth is all about. Despite the extra effort, I would
rather reinvent the wheel for the joy of learning. I see Forth in general
and Machine Forth in particular as a game that is played where part of the
appeal is to determine how simple one can possibly set the "rules" to be.
It is the various insights into the true and secret nature of digital
computing in general and the MISC stack machines in particular that are
inspiring me right now. Language has a great effect on the mental
symbology that we all use to reason and dream with, and I feel that there
are still veils to penetrate in my own personal perception of things. At
the risk of boring everyone, I would like to provide an example.

F21 has a COM instruction, but not a 1+ or NEGATE. Subtracting a
constant is easy: + a negative value, but subtraction of variables would at
first blush require COM + LIT #1 + for each term which is slightly
objectionable from a bus bandwidth standpoint. It is not so obvious to
someone *who thinks they know something* that you can streamline things
down to COM + for each term and wrap things up with LIT #N +, where N is
the number of negative terms involved. Maybe I'm a dummy and it should
have been obvious from the start, but at least I know it and will persevere
in my efforts to retrain part of my mind into thinking like an optimizing
compiler for the MISC architecture and instruction sets. I have obvious
gaps in my
education, and experiments with the MISC concept are exposing them and
stimulating me to rethink things at a fundamental level.

One thing that has bothered me since joining the MISC list is the "all
talk, no action" pattern that has developed re: fund-raising activities to
finance the next fab run of F21. T-shirts were suggested but have not
materialized (at least to my knowledge, but I might have missed that while
in transit from California back to Canada). My recent notion of "The 2's
Complement Cookbook" (hardcopy @US$100 ea. proposed, procedes to
the F21 program) has resulted in some good feedback, but I can see that it
will be a long haul in getting the readership to actually cough up the
dough.
For that matter, I have no idea what the real demographics of the MISC
readership is *or whether or not such activities are welcome in your's and
Chuck's view*. I believe it should be an outright gift with no strings
attached in the understanding that F21(e?) *might not be the last stepping
required*. On the other hand, I hope that you and Chuck are willing to put
a capper on high-risk experiments and consolidate your hard-won
experience into producing a definitive version. Personally, I am more
interested in an F21 version that simply provides the basic CPU core, boot
ROM, (S?)DRAM, and demuxed bus interface than F21d's multimedia
peripheral suite, but I also understand that the settop box market is what
you have been going for. It still might be worth considering as a stepping
stone and cash cow to release a lower-integration version if an unqualified
success in the resulting silicon would result. I am not a well-heeled
philanthropist and despite my genuine wish to see the F21 emerge as a
winner, I just can't bring myself to spend US$750 on a sample of F21d with
it's known (and unknown) problems. As an experienced chip designer
myself, I know damn well that it is possible to "get it right the 1st time"
at
least in a pure digital design by respecting silicon process limitations
and
doing things "by the book". At any rate, I hope that you and Chuck are
either in the home stretch with the full concept OR are tempted to start
making some money in the short-term with a more "conservative"
implementation.

The F21/Machine Forth concept is an extremely valuable alternative to
current industry trends. Money is needed to bring it to fruition. This
would need to be done in such a way that Jeff and Chuck retained full
control of their baby, i.e. donations, not buying shares.

This isn't exactly "The American Way", but as an aviator I am aware of the
history of Count Ferdinand Zeppelin's efforts to develop the airship into a
practical form of air transport, his personal financial stake, and the
grassroots support from the German people that saved the venture from
financial death. I know that this is a politically risky parallel to draw
due to millions of Led Zeppelin covers in our record collections, WW1
bombing raids etc., but the fact of the matter is that Count Zeppelin did
succede, and the Hindenberg disaster never would have occured if it wasn't
for the US government's denial of the sale of helium to loft it with.
Anyone can verify this by researching the many books on the subject. My
point here is that the grassroots coughed up the dough as a very concrete
demonstration of support for a new and magnificent endeavour and I hope
that the MISC readership is prepared to do the same now or that some other
solution materializes.

My question to Jeff and Chuck is: Do you want guys like me embarking on
fund-raising activities and other donations of effort in the first place?
This is really the crucial question. There is no point in any of us
performing in the capacity of a "ladies auxiliary" if the efforts are not
welcome. As a rule, I suspect the motives of busybodies and don't get
involved in commitees and such, but I am willing to do so if it contributes
to being able to buy fully-operational silicon from UltraTechnology.

My question to the MISC list is: How many of you would actually kick in
some cash in support of fund-raising, how much, and for what deliverables?

Let's be honest and see what (if anything) can be done. If the outcome is
negative, I'll stay out of it and mind my own business (of which I have
plenty) from now on and simply stay tuned for further developments. One
way  or the other, I want to and ultimately will build a stack computer. I
would rather it were F21-based for various reasons.

Myron Plichota