Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.forth > #21149

Re: There you go boys and girls, a challenge ..

Show all headers | View raw

On 3/24/2013 6:54 PM, AKE wrote:
> On Sunday, March 24, 2013 9:40:36 PM UTC, rickman wrote:
>>
>> Interesting.  Note that the 32 bit processor array will have "an
>> entirely new instruction set".  So what is left of the old architecture?
>>    I guess a bunch of processors on a die...
>>
>
> Any opinions on what holy grail is being searched for with 'entirely new instruction set'?

I would expect it will be similar to the existing ISA.  Instead of three 
and a half, five bit instructions per word, I guess they might give it 
six and a quarter instructions... lol  Or they might go with six bit 
instructions only having five and a quarter per word.  The multiple 
instruction per memory word seems like a very odd way to do things, but 
it can be very powerful.  Having more instructions per word gives a bit 
more flexibility but also can waste a bit more space, if that is 
important...

I found it interesting that even at the chip level, because Chuck's CPU 
is so optimized for speed, there is still the imbalance between CPU and 
memory cycle time.  That has always been a problem because of going off 
chip.  He sees it even on chip and this is how he deals with it.

With six bits, there can be twice as many instructions allowing the 
encoding of a SWAP instruction finally.  Although I found that not using 
a SWAP is often the better choice oddly enough.  Not really much more 
effort than the usual stack gymnastics.


> Seems odd to me that another 'radical' processor would be attempted before any of the preceeding designs reached acceptance in the market?

"Acceptance"?  How would you know?  It's not like GA has a press release 
every other week with announcements.  They claim to have done a second 
fab run of thousands of chips, pushing the total number of CPUs produced 
into the millions.  Somebody must be buying them.


> I can understand the original goal of wanting to design a processor closely matched to the language, i.e. having the instruction set in hardware matched to the Forth stack machine.  And if this led to the the Novix / RTX, one could say they reasonably successful -- at least they were designs that seemed to make it into production, and had some big customers.

I don't think at all that Chuck's CPUs are about being "closely matched 
to the language".  I think they are just the CPU Chuck designed and of 
course he uses Forth to develop his code.  The stack design just turns 
out to be pretty simple and still powerful if you optimize the right stuff.


> In which case, I can understand the next goal of wanting to combine a number of these simple Forth processors onto the same silicon to get cheap / simple parallel computing.

I also don't think the GA144 type chips are about "parallel computing" 
in any traditional sense.  It was when I started working with it that I 
realized it was more of an FPGA sort of approach to embedded digital 
design but with processors as the LUT-FF cell.  Yes, this is a massively 
parallel processor, but it is not about that.  It is about the CPUs 
being small and inexpensive so that you can afford to have one processor 
do a very tiny job and not try to optimize the resource.


> But looking at the long chain of special processors attempts leading up to the GA144 after the Novix / RTX 2000 ...
>          SC-32, Sh-boom, PSC1000, MuP21, i21, F21, P8/P16/P24, SEAforth24, S40, C18, GA4, GA32, GA144
>
> ... and the long list of startups that threw money and staff at the problems...
>          Computer Cowboys, Patriot Scientific Corp, UltraTechnology, iTV, Offete, IntellaSys, GreenArray
>
> ... and seemed to run out of funding before reaching stable production runs,
>
> ... starts to seem as though these are essentially 'experiments in silicion' being pursued almost as though they were software, (lots of refactoring, radical changes, etc.), except more time consuming and surely more costly.
>
> [I recall coming across an article putting the development cost of the SEAforth at $40M.]
>
> Somehow, the apparent lack of uptake of these chips commercially doesn't seem to dissuade the proposal / development / funding of the next one.
>
> Which brings back to: any thoughts on what holy grail is being searched for?

I think that Chuck might have had some issues with finding reliable 
customers/development partners over the years.  I expect GA to be a 
workable business approach if they find even just a couple of high 
volume customers to get them going enough that they can become solvent. 
  But they have a unique marketing style and it will be some time before 
they are really mainstream.

-- 

Rick

Back to comp.lang.forth | Previous | Next — Previous in thread | Next in thread | Find similar

Thread

There you go boys and girls, a challenge  .. organicmonkeymotion@gmail.com - 2013-03-24 01:30 -0700
  Re: There you go boys and girls, a challenge  .. Paul Rubin <no.email@nospam.invalid> - 2013-03-24 11:02 -0700
    Re: There you go boys and girls, a challenge  .. mhx@iae.nl (Marcel Hendrix) - 2013-03-24 20:21 +0200
      Re: There you go boys and girls, a challenge  .. albert@spenarnc.xs4all.nl (Albert van der Horst) - 2013-03-24 19:35 +0000
      Re: There you go boys and girls, a challenge  .. Paul Rubin <no.email@nospam.invalid> - 2013-03-24 12:40 -0700
        Re: There you go boys and girls, a challenge  .. mhx@iae.nl (Marcel Hendrix) - 2013-03-24 23:26 +0200
          Re: There you go boys and girls, a challenge  .. Paul Rubin <no.email@nospam.invalid> - 2013-03-24 15:57 -0700
          Re: There you go boys and girls, a challenge  .. Bernd Paysan <bernd.paysan@gmx.de> - 2013-03-25 03:35 +0100
            Re: There you go boys and girls, a challenge  .. Brad Eckert <hwfwguy@gmail.com> - 2013-03-25 11:03 -0700
              Re: There you go boys and girls, a challenge  .. Bernd Paysan <bernd.paysan@gmx.de> - 2013-03-25 19:40 +0100
          Re: There you go boys and girls, a challenge  .. rickman <gnuarm@gmail.com> - 2013-03-25 19:34 -0400
    Re: There you go boys and girls, a challenge  .. rickman <gnuarm@gmail.com> - 2013-03-24 17:40 -0400
      Re: There you go boys and girls, a challenge  .. AKE <assadebrahim2000@gmail.com> - 2013-03-24 15:54 -0700
        Re: There you go boys and girls, a challenge  .. Brad Eckert <hwfwguy@gmail.com> - 2013-03-25 11:39 -0700
        Re: There you go boys and girls, a challenge  .. rickman <gnuarm@gmail.com> - 2013-03-25 20:00 -0400
          Re: There you go boys and girls, a challenge  .. AKE <assadebrahim2000@gmail.com> - 2013-03-25 17:33 -0700
            Re: There you go boys and girls, a challenge  .. rickman <gnuarm@gmail.com> - 2013-03-25 21:56 -0400
  Re: There you go boys and girls, a challenge  .. visualforth@rocketmail.com - 2013-03-24 17:03 -0700
    Re: There you go boys and girls, a challenge  .. rickman <gnuarm@gmail.com> - 2013-03-25 19:43 -0400

csiph-web