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

Re: F21D CHIP


GARY B. LAWRENCE wrote:
> 
> I have been fairly busy and didn't see the status report that Jeff Fox
>  has tested the boot up rom and dynamic ram operation of the F21D chip.
>  I am glad to see that the chip is working and hope to hear that the
> video and I/O portions check out ok too. I have a few questions for Jeff
>  about the new chip. I read the new article on Chuck's color forth and
>  was worried at the time that the F21D would have the same problem with
>  transistors overheating. It was sad to hear from Jeff that the F21D has
>  the same problem and that the chip needs to use nops to keep the main
>  transistors cool enough to run well.
>   Questions for Jeff
>   1. Will you spend time figuring out which combination of opcodes will
>      work together?
> 
>   2. If the rest of the chip tests good, do you think Chuck can
>      solve      the temperature problem just by increasing the size of a few
>      of the transistors in the next mask?
> 
>   3. What affect would larger transistors have on the speed of the chip?
> 
>   4. Will you try to fix up an assembler for this F21D to insert three
>      nops between each consecutive opcode?
> 
>   5. If Chuck thinks he can solve the temperature problem with larger
>      transistors, will you try to get a new mask done fairly quickly?
>      I am assuming that he will know the answer about using
>      larger transistors from  the status of a new ITV mask run. I
>      realize that you can not comment on ITV chips but I am hoping that
>      Chuck will be able to prove that he now understands how to model the process
>      correctly as he mentioned as possible in the color Forth article.
> 
>     6. Will you compile a new p21 style forth for this chip? I could
>        see it working much faster than the one on the Mup21 even if
>        three out of every four opcodes are nops because of the larger
>        stacks and faster speed.
> 
>     7. Will the F21D chip be able to do a non jitter display even
>        when the main processor is working in a high level Forth?
> 
>     8. What voltage ranges does the F21D work at?
> 
>     9. Does it slow down in a linear fashion if the voltage is reduced?
> 
>     10. How accurate is the real time clock circuit?
> 
>    11. In your article " Low Fat Computing " you mentioned that a GUI
>        could be done in about 1K. Could you describe what features you
>        have in such a GUI and how you got it so small. In Windows
>        95 each shortcut icon has a 90 byte file so that 12 icons
>        alone take up 1K, but nobody ever said Windows was meant to be small.
> 
>    12. You also mentioned in the low fat article that Chuck
>        thinks about how he is going code alot before he writes any
>        code. I would like to know if there are any books that help
>        in doing code
>        in this manner? I have read " Thinking Forth" and then went back
>        to redesign part of a database program. I thought about the data
>        structure quite a bit and then redesigned it. I produced code        
         that was a partial win, the stuctures were easier to follow
when
>        looking for a particular field, but I incorporated the
>        actions used upon the fields into a complex execution array
>        which had to        be set to a particular "state" in many places and it
>        was not        apparent which " state" was set when looking at the field
>        code.
>        I did remember a tip in Thinking Forth that Chuck had made about
>        sometimes using more words and letting the Forth dictionary 
         determine the "condition" by compiling the right word.
>        When I redo my field code, I will return to a separate field word
>        for each action and remove the "State" words. The point that I am
>        trying to make is that it seems that having alot of good
>        tips from a book like Thinking Forth is not enough, that considerable
>        experience is needed to know which particular tool to use in
>        the situation at hand. One book mentioned to me by friends who
>        use C was Code Complete. I got this book and found that so much of
>        the book dealt with ideas on how to restructure scource code to
>        make it easier to follow the basic logic. I was disappointed because
>        a Forth program which is moderately well factored did not have  
         many of the problems mentioned. Most of what I gained from
the   
         book was an understanding of why even moderately small compiler
>        programs can be harder to understand and debug than fairly well
>        written Forth programs. It seems to be a case where the greater
>        amount of so called natural language wording makes it harder not
>        easier to understand the structure and logic of the program.
> 
>     13. Is there any set of listings of code Chuck produced say 10,20 or
>         30 years ago I could read to try to understand how he evolved to                   this style of programming?
> 
> 
>      I would like to close with an idea that I think the F21D would be
>      able to do better than any current design. On the internet now
>      people are setting up linux or unix servers to handle auctions
>      using pearl script and one or more linked linux boxes. A similar
>      internet product could be done with interlinked F21 chips.
>      Since Pearl is working mostly in string arrays and not floating point
>      math Forth string code could be used on a F21 at better speed and
>      at much lower interconnect cost. Very large number of input           
       requests could be handled with much less processoroverhead since
>      unix style processors are doing many background processes. I also
>      think that a possible large scale game database could be set up on 
       the internet using a F21 network to hold the database and
interact
>      within the database between each player and they are moved around
>      the database "space". The best way I think to design such a game
>      would be to have a graphic client running on pcs and to have only
>      small amounts of position, status and graphic information sent over
       the internet. The main advantages of the F21 design involve the
use
>      of the simple parallel processing, the much lower cost of each node
>      and the fact that a well written Forth code could be maintained and
>      expanded with greater ease.
> 
       Since the last time I tryed to send this on 1/6/99 I have learned
that you
       have had problems in the business and I hope you can get your
chips out and 
       sold to help with funding. I will try to send some money at least
for the  
       video on the Chuck's color forth in february. 
                With best wishes for a prosperous new year 
						Gary Lawrence