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

Groups > comp.lang.forth > #26337 > unrolled thread

A clear victory for MMP

Back to article view | Back to comp.lang.forth

Contents

  A clear victory for MMP Chris <xrissmith@me.com> - 2013-10-06 18:24 -0700
    Re: A clear victory for MMP Paul Rubin <no.email@nospam.invalid> - 2013-10-06 18:48 -0700
      Re: A clear victory for MMP jkn_gg@nicorp.f9.co.uk - 2013-10-08 10:51 -0700
    Re: A clear victory for MMP Howerd <howerdo@yahoo.co.uk> - 2013-10-07 04:44 -0700
      Re: A clear victory for MMP Andrew Haley <andrew29@littlepinkcloud.invalid> - 2013-10-07 07:13 -0500
        Re: A clear victory for MMP Howerd <howerdo@yahoo.co.uk> - 2013-10-07 05:43 -0700
      Re: A clear victory for MMP visualforth@rocketmail.com - 2013-10-08 10:42 -0700
      Re: A clear victory for MMP rickman <gnuarm@gmail.com> - 2013-10-11 02:18 -0400
        Re: A clear victory for MMP "Elizabeth D. Rather" <erather@forth.com> - 2013-10-10 20:35 -1000
      Re: A clear victory for MMP "Rod Pemberton" <dont_use_email@nohavenotit.com> - 2013-10-25 18:05 -0400
    Re: A clear victory for MMP Brad Eckert <hwfwguy@gmail.com> - 2013-10-10 11:28 -0700
      Re: A clear victory for MMP albert@spenarnc.xs4all.nl (Albert van der Horst) - 2013-10-10 23:01 +0000
        Re: A clear victory for MMP Rafael Deliano <rafael_deliano@arcor.de> - 2013-10-11 20:35 +0200
          Re: A clear victory for MMP forther <forther@gmail.com> - 2013-10-12 12:47 -0700
            Re: A clear victory for MMP rickman <gnuarm@gmail.com> - 2013-10-12 22:36 -0400
              Re: A clear victory for MMP forther <forther@gmail.com> - 2013-10-13 23:12 -0700
                Re: A clear victory for MMP rickman <gnuarm@gmail.com> - 2013-10-25 13:24 -0400
                  Re: A clear victory for MMP Paul Rubin <no.email@nospam.invalid> - 2013-10-25 12:06 -0700
                    Re: A clear victory for MMP rickman <gnuarm@gmail.com> - 2013-10-26 02:40 -0400
                      Re: A clear victory for MMP Paul Rubin <no.email@nospam.invalid> - 2013-10-26 00:14 -0700
          Re: A clear victory for MMP rickman <gnuarm@gmail.com> - 2013-10-12 22:32 -0400
            Re: A clear victory for MMP Rafael Deliano <rafael_deliano@arcor.de> - 2013-10-13 10:36 +0200
              Re: A clear victory for MMP Paul Rubin <no.email@nospam.invalid> - 2013-10-13 20:46 -0700
              Re: A clear victory for MMP rickman <gnuarm@gmail.com> - 2013-10-25 13:55 -0400
                Re: A clear victory for MMP Rafael Deliano <rafael_deliano@arcor.de> - 2013-10-27 09:25 +0100
                  Re: A clear victory for MMP Paul Rubin <no.email@nospam.invalid> - 2013-10-27 03:00 -0700
                    Re: A clear victory for MMP Rafael Deliano <rafael_deliano@arcor.de> - 2013-10-27 11:50 +0100
                  Re: A clear victory for MMP albert@spenarnc.xs4all.nl (Albert van der Horst) - 2013-10-27 10:02 +0000
                    Re: A clear victory for MMP Rafael Deliano <rafael_deliano@arcor.de> - 2013-10-27 13:37 +0100
                      Re: A clear victory for MMP rickman <gnuarm@gmail.com> - 2013-10-27 12:28 -0400
                  Re: A clear victory for MMP rickman <gnuarm@gmail.com> - 2013-10-27 06:07 -0400
                    Re: A clear victory for MMP Rafael Deliano <rafael_deliano@arcor.de> - 2013-10-27 13:03 +0100
          Re: A clear victory for MMP the_gavino_himself <visphatesjava@gmail.com> - 2013-11-06 16:57 -0800
    Re: A clear victory for MMP Chris <xrissmith@me.com> - 2013-10-16 20:11 -0700
      Re: A clear victory for MMP albert@spenarnc.xs4all.nl (Albert van der Horst) - 2013-10-17 12:14 +0000
        Re: A clear victory for MMP John Rible <google@sandpipers.com> - 2013-10-17 10:51 -0700
        Re: A clear victory for MMP "Rod Pemberton" <dont_use_email@nohavenotit.com> - 2013-10-18 03:30 -0400
    Re: A clear victory for MMP Chris <xrissmith@me.com> - 2013-11-01 20:13 -0700
    Re: A clear victory for MMP the_gavino_himself <visphatesjava@gmail.com> - 2013-11-06 16:59 -0800

Page 2 of 2 — ← Prev page 1 [2]

#26442

On 10/11/2013 2:35 PM, Rafael Deliano wrote:
>> I pray for 10 million royalties, 5 million investment in application
>> notes and the enlightment to produce part of the GA chips in DIL
>> packages.
>> (but I never believed in praying anyway...)
>
> I have put the GA board in a box in the cellar by now: the cpu is
> too limited for anything usefull, decent documentation wouldn't help.

I won't argue that the GA144 is a universally useful device.  The first 
DSP chips were real dogs and highly specialized.  The first FPGAs lacked 
so many of the features FPGAs now have.  One of the first MCUs was the 
8051, the device everyone loves to hate.

The GA144 is the first device of it's kind.  Yes, it is crude and has a 
lot of limitations, but it has a lot of potential.


> GA144 is Chuck Moore's hobby horse masquerading as a commercial company.
> And his customer orientation is after all legendary. Novix
> multiplication bug: "It never affected any of my applications...".

I also won't argue with this.  I wasn't aware of the Novix mult bug. 
But I don't see GA as a great company in many respects, but then they 
ware working on a shoestring budget.  It's actually amazing they can get 
product out the door really.


> Whats needed is someone like Jeff Fox who can sell.
> And a more decent product. Not a bunch of crippled transputers,
> at 1,8 Volt, in a unbreadboardable package.

Can't say I follow here.  I agree that a programmable device with 1.8 
volt I/Os is a bit pointless... you need to do level conversion for 
nearly any I/O commonly used.  But what is wrong with the package?  If 
you are expecting DIPs you are so out of step with the real world, the 
world that actually buys product and keeps factory doors open.

My complaint with most of the FPGA makers is that they don't use a 
package like this.  Instead they use packages that require very fine 
pitch board layouts and very, very small drills.  0.4 mm spaced uBGAs 
are common.  Try using one of those, the GA144 10 mm sq package will 
seem a breath of fresh air.


> My odd advice:
> * if possible team up with Chip Gracey of Parallax.
> The propeller is much more usable and reasonably successfull.
> But obviously not enough for a second generation chip.
> The propeller multitasking is nice, the CPU is so-and-so,
> the Spin-language ugly.
> Something like a RTX2000 CPU and a FORTH in ROM combined
> with the propeller multitasker would be a decent combination
> that would appeal to two established small user communities.

If you are comparing the parallax with the GA144 you don't understand 
the GA144.  It is *not* just a collection of many processors.  Viewing 
it that way gets you nothing.  Think of a 22V10 compared to an FPGA, 
that would be the Parallax compared to the GA144.


> * test the market and finance via kickstarter / indiegogo.
> If a well prepared campaign doesn't succeed with the hobbyist
> and small volume user then its no go. There are no big volume
> industrial buyers readily available.

That is an interesting idea actually.

-- 

Rick

[toc] | [prev] | [next] | [standalone]

#26460

> The GA144 is the first device of it's kind.

Not really.
   There has been a long tradition of arrays down to
ILLIAC IV. They had big CPUs ( ILLIAC, Transputer ) and
small "CPUs"/PEs like CLIP4 and DAP. The later were
actully only SIMD-arrays, no independent CPUs. But
they are somewhat comparable to the crippled GA144.
The DAP was arguably commercially viable, as it was
still around in the 80ies. It used a fat SRAM early on
and  "a variant of Fortan because thats what the
customers wants" to paraphrase them. So no
Color-Forth, no Spin-language, no Occam.
The CLIP group and everyone else soon too came to the
conclusion that they needed more RAM per PE too.
   The Martin Marietta GAPP had even a odd Forth
connection: Ting used a NC4000 as I/O-processor.
   There is a vast amount of literature out there
about these machines, how they evolved, what applications
were viable. Guess computer cowboy Moore isn't a reader.

   Key problem: these machines were and are almost
exclusively for the "government projects" market.
   The real computer cowboy is Chip Gracey because
his chip you can buy at digikey and ebay. Usable
for the hobbyist today, can slowly penetrate to industry.

> But what is wrong with the package?  If
> you are expecting DIPs you are so out of step with the real world, the
> world that actually buys product and keeps factory doors open.

   That world doesn't buy from a company like GA. If they had any
credibility there they would get venture capital.
Hobbyists accept GA, small industry too. If you run small
projects using exotic chips isnt much risk. But a lot of projects
die at the breadboard stage. Not everyone has the equipment
to get BGAs and similar stuff in and out of boards.
   Most users wouldn't mind if the "DIL40" is a FR4 carrier board
with a small LCN/BGA and capacitors on it that plugs in a DIL40
socket. Yes its more expensive, but it are boutique chips anyway.

> If you are comparing the parallax with the GA144 you don't understand
> the GA144.  It is *not* just a collection of many processors.  Viewing
> it that way gets you nothing.  Think of a 22V10 compared to an FPGA,
> that would be the Parallax compared to the GA144.

   As far as i remember there was an evolution from smallish single
CPUs MuP21 ( Ting ), F21 ( Jeff Fox ) that worked on external memory.
And were in no way competitive with controllers with integrated
memory from industry. So that was a dead end.
   The F18 CPU was scaled down further to exclusively use SRAM on chip.
And a very rudimentary 4-way interface was patched on so these
can form an array. And thats it. There were never any thoughts
who should use it and why and how. Otherwise simulations how that
thing could do a FIR-filter or a fourier transform ( via Chirp-Z )
would have been done and a application notes about it would be
ready now.
   On the other hand the machine of Chip Gracey has vast amounts
of SRAM ( 32kByte ), is easier to understand by the average user
and a better fit to his needs.

 >> * test the market and finance via kickstarter / indiegogo.
 > That is an interesting idea actually.

Beeing low volume ( <500 ) patching a-ready-to-run
hardware on a FR4 carrier board its doable by stackprocessor
on FPGA. Even a controller with FORTH-in-ROM/FLASH would
be applicable. Kickstarter / indiegogo is not about fancy
technology.

MfG  JRD

[toc] | [prev] | [next] | [standalone]

#26488

Rafael Deliano <rafael_deliano@arcor.de> writes:
>   Most users wouldn't mind if the "DIL40" is a FR4 carrier board

That's sort of what a Schmartboard is, though not a DIL:

  http://schmartboard.com/index.asp?page=products_csp&id=532

[toc] | [prev] | [next] | [standalone]

#26679

On 10/13/2013 4:36 AM, Rafael Deliano wrote:
>> The GA144 is the first device of it's kind.
>
> Not really.
> There has been a long tradition of arrays down to
> ILLIAC IV. They had big CPUs ( ILLIAC, Transputer ) and
> small "CPUs"/PEs like CLIP4 and DAP. The later were
> actully only SIMD-arrays, no independent CPUs. But
> they are somewhat comparable to the crippled GA144.
> The DAP was arguably commercially viable, as it was
> still around in the 80ies. It used a fat SRAM early on
> and "a variant of Fortan because thats what the
> customers wants" to paraphrase them. So no
> Color-Forth, no Spin-language, no Occam.
> The CLIP group and everyone else soon too came to the
> conclusion that they needed more RAM per PE too.
> The Martin Marietta GAPP had even a odd Forth
> connection: Ting used a NC4000 as I/O-processor.
> There is a vast amount of literature out there
> about these machines, how they evolved, what applications
> were viable. Guess computer cowboy Moore isn't a reader.

What is similar to the GA144 other than the fact these were multiprocessor?


> Key problem: these machines were and are almost
> exclusively for the "government projects" market.
> The real computer cowboy is Chip Gracey because
> his chip you can buy at digikey and ebay. Usable
> for the hobbyist today, can slowly penetrate to industry.

Again, nothing like the GA144 except multiple processors.  If you think 
the GA144 is just a multiprocessor chip you just don't get the GA144.


>> But what is wrong with the package? If
>> you are expecting DIPs you are so out of step with the real world, the
>> world that actually buys product and keeps factory doors open.
>
> That world doesn't buy from a company like GA. If they had any
> credibility there they would get venture capital.
> Hobbyists accept GA, small industry too. If you run small
> projects using exotic chips isnt much risk. But a lot of projects
> die at the breadboard stage. Not everyone has the equipment
> to get BGAs and similar stuff in and out of boards.
> Most users wouldn't mind if the "DIL40" is a FR4 carrier board
> with a small LCN/BGA and capacitors on it that plugs in a DIL40
> socket. Yes its more expensive, but it are boutique chips anyway.

Hobbiests are unimportant to the profit of any company over $1 
million/yr revenues, unless you are Heathkit or the like.  Zero point in 
even discussing the hobby market regarding GA.


>> If you are comparing the parallax with the GA144 you don't understand
>> the GA144. It is *not* just a collection of many processors. Viewing
>> it that way gets you nothing. Think of a 22V10 compared to an FPGA,
>> that would be the Parallax compared to the GA144.
>
> As far as i remember there was an evolution from smallish single
> CPUs MuP21 ( Ting ), F21 ( Jeff Fox ) that worked on external memory.
> And were in no way competitive with controllers with integrated
> memory from industry. So that was a dead end.

You forget the RTX2000 which is the closest thing ever to an actual 
Forth CPU.  Its life has ended, but it had a useful run.


> The F18 CPU was scaled down further to exclusively use SRAM on chip.
> And a very rudimentary 4-way interface was patched on so these
> can form an array. And thats it. There were never any thoughts
> who should use it and why and how. Otherwise simulations how that
> thing could do a FIR-filter or a fourier transform ( via Chirp-Z )
> would have been done and a application notes about it would be
> ready now.

I agree that it appears the applications for the GA144 were unknown at 
the time it was designed.  Rather it seemed to be an interesting design 
optimized for certain features while totally ignoring requirements in 
many apps.


> On the other hand the machine of Chip Gracey has vast amounts
> of SRAM ( 32kByte ), is easier to understand by the average user
> and a better fit to his needs.

To Chip Gracey's needs?  Oh, the user's needs... really?  All of them? 
that's a rather large statement and ignores the advantages of the GA144. 
  What if your needs included ultra low power consumption?


>  >> * test the market and finance via kickstarter / indiegogo.
>  > That is an interesting idea actually.
>
> Beeing low volume ( <500 ) patching a-ready-to-run
> hardware on a FR4 carrier board its doable by stackprocessor
> on FPGA. Even a controller with FORTH-in-ROM/FLASH would
> be applicable. Kickstarter / indiegogo is not about fancy
> technology.

I have no idea what you are trying to say here.

-- 

Rick

[toc] | [prev] | [next] | [standalone]

#26704

>> ILLIAC IV. ... CLIP4 ... DAP ... GAPP

> What is similar to the GA144 other than the fact these were multiprocessor?

There has obviously been a long tradition of
building computer arrays chessboard like with
4 interconnections per PE.
Either with big CPUs like ILLIAC and the Transputer
and then with a reduced number of PEs.
Or with a big number of minimalist PEs like
CLIP4, DAP ... These were limited by beeing SIMD,
but that makes not much difference to the rather
crippled GA144.

There is the famous Elon Musk quote on SpaceX
in their early years:
"Well, I have tried to learn as much as possible
from prior attempts. If nothing else, we are
committed to failing in a new way."
Skipping the first sentence and just
beeing different is obviously not enough.

>  you just don't get the GA144.

Neither does any else it seems.
   I at least have the evaluation board and
had a serious look at the gadget.
And my view is that the PE is too small.
RAM obviously. And i don't like CPUs without
XOR opcode.
Minimum would be something like a von Neumann
PIC16 as PE.

> Hobbiests are unimportant to the profit of any company over $1
> million/yr revenues,

I do not see "$1 million/yr revenues" by GA selling GA144s.
Chip Gracey did prove that selling to the hobbyist is viable.
   But it has never been the typical option.
Bill Mensch got his 65816 into production
via Korea. No chip manufacturer in the US wanted the 65816.
And Koreas semiconductor industry was then so insignificant
that they wouldn't get a second source offer for a established CPU.
Apple could use the chip because it was available as sample
and in volume.

> You forget the RTX2000 which is the closest thing ever to an actual
> Forth CPU.  Its life has ended, but it had a useful run.

I think Charles Moore is still associated by the general public
with that chip. It would be an easy sell.
Nobody expects and wants a parallel processor
from him. A rather conventional 16/32 bit stackprocessor with
limited multitasking capability ( similar to Parallax )
and enough microcode so one can claim "runs high level
language in hardware" would get customers in industrial control.

> To Chip Gracey's needs?  Oh, the user's needs... really?  All of them?
> that's a rather large statement and ignores the advantages of the GA144.
>   What if your needs included ultra low power consumption?

Chip Gracey's hobby horse has proven to be commercially viable.
When Digikey stocks GA144s one can discuss the technical merits.

MfG  JRD

[toc] | [prev] | [next] | [standalone]

#26705

Rafael Deliano <rafael_deliano@arcor.de> writes:
> There has obviously been a long tradition of building computer arrays
> chessboard like with 4 interconnections per PE.  Either with big CPUs
> like ILLIAC and the Transputer and then with a reduced number of PEs.

ILLIAC was a hypercube, I thought.  Not sure about Transputer.
Chessboard would be 8 interconnections per PE, I'd hope (kingwise
moves).  It seems unfortunate that the GA didn't also have 8, since it
would still be very local, yet make routing a lot easier.  Some
further-reaching interconnects would be good too.  FPGA's have a mix of
these, from what I understand.

> And my view is that the PE is too small.  RAM obviously. And i don't
> like CPUs without XOR opcode.

The GA144 has XOR, their assembler just use the mnemonic OR for it, due
to idiosyncracy.

> Minimum would be something like a von Neumann PIC16 as PE.

They're committed to stack machines, which have had some success, but
yeah, more memory would help it a lot.  Not necessarily ram: with some
program flash at each node, the 64 words of ram would then be sufficient
for lots of things.

> Chip Gracey did prove that selling to the hobbyist is viable.

Somebody is buying all those raspberry pi's...

> A rather conventional 16/32 bit stackprocessor with limited
> multitasking capability ( similar to Parallax ) and enough microcode
> so one can claim "runs high level language in hardware" would get
> customers in industrial control.

Maybe in the 1980's.  That wouldn't compete today with the ubiquitous
ARM cores and compilers everywhere.

[toc] | [prev] | [next] | [standalone]

#26708

> ILLIAC was a hypercube, I thought.

No details on that in http://en.wikipedia.org/wiki/ILLIAC_IV
So had a look at Slotnicks description in Scientific American 1971:
its a flat square. There was an alternative view of that
with all 64 PEs in a line but with an additional link to another
PEs 8 places away.
Cubes were an idea of the 80ies with everything "massively
parallel". Even the original 256 PE ILLIAC was not that.
But it had big 64bit float PEs.

 > Not sure about Transputer.

Always 4 links as far as i remember.

> Chessboard would be 8 interconnections per PE, I'd hope (kingwise
> moves).

CLIP4 ( early 70ies ) chip had originally 8 because its known that
has technical advantages in image processing. But after that everyone
settled on 4 because easier/cheaper in hardware.

>> Minimum would be something like a von Neumann PIC16 as PE.
> They're committed to stack machines, which have had some success,

Having seen the MARC4 instruction set i have no doubt that a
similar compact stack processor is possible. But that would
have to be similar in performance to a PIC16. Not only in
available memory, but also a instruction set that is acceptable
to someone used to programming 8 bit CPUs.

>> A rather conventional 16/32 bit stackprocessor with limited
>> multitasking capability ( similar to Parallax ) and enough microcode
>> so one can claim "runs high level language in hardware" would get
>> customers in industrial control.

> Maybe in the 1980's.  That wouldn't compete today with the ubiquitous
> ARM cores and compilers everywhere.

Some time ago i tried toggling a portpin with a ARM from Energy Micro: 
slower then the 8 bit 8MHz MC68H908 i am normaly using.
In that ARM the ports were called by C routines provided by the 
manufacturer because not much documentation about the rather complex
internals of the I/O were available.
I am in no doubt that ARMs are good for running Linux. But i am very
much in doubt that the main use of industrial controllers is running
a PC/minicomputer operating system.

MfG  JRD

[toc] | [prev] | [next] | [standalone]

#26706

In article <526ccde1$0$9515$9b4e6d93@newsspool1.arcor-online.net>,
Rafael Deliano  <rafael_deliano@arcor.de> wrote:
>>> ILLIAC IV. ... CLIP4 ... DAP ... GAPP
>
>> What is similar to the GA144 other than the fact these were multiprocessor?
>
>There has obviously been a long tradition of
>building computer arrays chessboard like with
>4 interconnections per PE.
>Either with big CPUs like ILLIAC and the Transputer
>and then with a reduced number of PEs.
>Or with a big number of minimalist PEs like
>CLIP4, DAP ... These were limited by beeing SIMD,
>but that makes not much difference to the rather
>crippled GA144.

There is a tremendous difference with the transputer links ...

1. The transputer links are not setup in a fixed layout
2. The transputer has just one order of magnitude between the off chip
transfer of a word and its processing.
3. The transputer has a word size that is sensible, plus floating point.

I was thinking about doing a fast gcd (which is fundamental to factoring
large numbers and so indirectly related to cryptography.).
It needs two pipelines that can contain a large number, where each member of the
pipeline can talk to the corresponding other member. 128 processors for a
1Kbit number.
This kind of pipeline cannot be accommodated.

>
>There is the famous Elon Musk quote on SpaceX
>in their early years:
>"Well, I have tried to learn as much as possible
>from prior attempts. If nothing else, we are
>committed to failing in a new way."
>Skipping the first sentence and just
>beeing different is obviously not enough.
>
>>  you just don't get the GA144.
>
>Neither does any else it seems.
>   I at least have the evaluation board and
>had a serious look at the gadget.
>And my view is that the PE is too small.
>RAM obviously. And i don't like CPUs without
>XOR opcode.

With 32 instructions to study this remark undermines your authority.

Obviously XOR is the most fundamental opcode allowing
reversable computing on quantum computers.
So you've studied the chip? Apparently not enough. It has an
instruction that is called OR, bit it really is an XOR.

>Minimum would be something like a von Neumann
>PIC16 as PE.
>
>> Hobbiests are unimportant to the profit of any company over $1
>> million/yr revenues,
>
>I do not see "$1 million/yr revenues" by GA selling GA144s.
>Chip Gracey did prove that selling to the hobbyist is viable.
>   But it has never been the typical option.
>Bill Mensch got his 65816 into production
>via Korea. No chip manufacturer in the US wanted the 65816.
>And Koreas semiconductor industry was then so insignificant
>that they wouldn't get a second source offer for a established CPU.
>Apple could use the chip because it was available as sample
>and in volume.
>
>> You forget the RTX2000 which is the closest thing ever to an actual
>> Forth CPU.  Its life has ended, but it had a useful run.
>
>I think Charles Moore is still associated by the general public
>with that chip. It would be an easy sell.
>Nobody expects and wants a parallel processor
>from him. A rather conventional 16/32 bit stackprocessor with
>limited multitasking capability ( similar to Parallax )
>and enough microcode so one can claim "runs high level
>language in hardware" would get customers in industrial control.
>
>> To Chip Gracey's needs?  Oh, the user's needs... really?  All of them?
>> that's a rather large statement and ignores the advantages of the GA144.
>>   What if your needs included ultra low power consumption?
>
>Chip Gracey's hobby horse has proven to be commercially viable.
>When Digikey stocks GA144s one can discuss the technical merits.

Ouch! That hurts, but you're right.

>
>MfG  JRD
>
-- 
Albert van der Horst, UTRECHT,THE NETHERLANDS
Economic growth -- being exponential -- ultimately falters.
albert@spe&ar&c.xs4all.nl &=n http://home.hccnet.nl/a.w.m.van.der.horst

[toc] | [prev] | [next] | [standalone]

#26711

> There is a tremendous difference with the transputer links ...

I am not saying the PEs of ILLIAC & transputers are similar
to the GA144. But patching up PEs with 4 links as array
doesn't make it "the first device of it's kind".

> With 32 instructions to study this remark undermines your authority.

It doesn't. It simply says something about their documentation
if they call XOR anything else. There are crippled commercial CPUs 
without XOR out there like the ST62, so i wasn't too surprised.

Since i had the board then, someone asked me if i could do
something like 10000 0 DO LOOP to see how fast it would go.
Answer from eForth: DO LOOP is advanced technology and not
implemented in the basic package. That did surprise me for
a moment. How could anything be shipped that would foul up
on "Hello world" class programs ?
I think time has shown that hardly any who bought the board
did any demo application with it at all. Not to mention any
that was worth bragging about.

MfG  JRD

[toc] | [prev] | [next] | [standalone]

#26716

On 10/27/2013 8:37 AM, Rafael Deliano wrote:
>> There is a tremendous difference with the transputer links ...
>
> I am not saying the PEs of ILLIAC & transputers are similar
> to the GA144. But patching up PEs with 4 links as array
> doesn't make it "the first device of it's kind".
>
>> With 32 instructions to study this remark undermines your authority.
>
> It doesn't. It simply says something about their documentation
> if they call XOR anything else. There are crippled commercial CPUs
> without XOR out there like the ST62, so i wasn't too surprised.
>
> Since i had the board then, someone asked me if i could do
> something like 10000 0 DO LOOP to see how fast it would go.
> Answer from eForth: DO LOOP is advanced technology and not
> implemented in the basic package. That did surprise me for
> a moment. How could anything be shipped that would foul up
> on "Hello world" class programs ?
> I think time has shown that hardly any who bought the board
> did any demo application with it at all. Not to mention any
> that was worth bragging about.

So are you planning to toss it or what?  I'd be happy to pay the 
shipping if you wish to give it a toss my way.

-- 

Rick

[toc] | [prev] | [next] | [standalone]

#26707

On 10/27/2013 4:25 AM, Rafael Deliano wrote:
>>> ILLIAC IV. ... CLIP4 ... DAP ... GAPP
>
>> What is similar to the GA144 other than the fact these were
>> multiprocessor?
>
> There has obviously been a long tradition of
> building computer arrays chessboard like with
> 4 interconnections per PE.
> Either with big CPUs like ILLIAC and the Transputer
> and then with a reduced number of PEs.
> Or with a big number of minimalist PEs like
> CLIP4, DAP ... These were limited by beeing SIMD,
> but that makes not much difference to the rather
> crippled GA144.
>
> There is the famous Elon Musk quote on SpaceX
> in their early years:
> "Well, I have tried to learn as much as possible
> from prior attempts. If nothing else, we are
> committed to failing in a new way."
> Skipping the first sentence and just
> beeing different is obviously not enough.
>
>> you just don't get the GA144.
>
> Neither does any else it seems.
> I at least have the evaluation board and
> had a serious look at the gadget.
> And my view is that the PE is too small.
> RAM obviously. And i don't like CPUs without
> XOR opcode.
> Minimum would be something like a von Neumann
> PIC16 as PE.
>
>> Hobbiests are unimportant to the profit of any company over $1
>> million/yr revenues,
>
> I do not see "$1 million/yr revenues" by GA selling GA144s.
> Chip Gracey did prove that selling to the hobbyist is viable.
> But it has never been the typical option.
> Bill Mensch got his 65816 into production
> via Korea. No chip manufacturer in the US wanted the 65816.
> And Koreas semiconductor industry was then so insignificant
> that they wouldn't get a second source offer for a established CPU.
> Apple could use the chip because it was available as sample
> and in volume.
>
>> You forget the RTX2000 which is the closest thing ever to an actual
>> Forth CPU. Its life has ended, but it had a useful run.
>
> I think Charles Moore is still associated by the general public
> with that chip. It would be an easy sell.
> Nobody expects and wants a parallel processor
> from him. A rather conventional 16/32 bit stackprocessor with
> limited multitasking capability ( similar to Parallax )
> and enough microcode so one can claim "runs high level
> language in hardware" would get customers in industrial control.
>
>> To Chip Gracey's needs? Oh, the user's needs... really? All of them?
>> that's a rather large statement and ignores the advantages of the GA144.
>> What if your needs included ultra low power consumption?
>
> Chip Gracey's hobby horse has proven to be commercially viable.
> When Digikey stocks GA144s one can discuss the technical merits.

Chip Gracey's product is designed for and marketed to hobbyists?  I 
don't think so.  It may be bought by hobbyists, but that is not the 
intended market nor the profit source.

The rest of your post is pretty much not anything I feel the need to 
continue to respond to.  The GA144 is not a bunch of Transputers or ARM 
CPUs, it is not trying to be that.  It is something different and until 
you can accept that and learn how to use it you will fail to understand 
it.

I have a current product using an FPGA because there were no MCU chips 
that could provide all the interfaces needed.  The GA144 is the only 
device that could potentially replace the FPGA in this design.  The 
processing would map well to the GA144 elements and the low power 
capability would be useful.  The only downside is the 1.8 volt I/Os 
which would require additional chips or transistors to provide level 
shifting.

That said, the GA144 will not be used because of the uncertainty of the 
company, not the design of the chip.

I won't argue that the GA144 is an ideal product.  But when you express 
your preference for larger processors or more memory you are simply not 
considering the device properly.  If you need a screwdriver, then use a 
screwdriver, not a wrench.  But don't say the wrench is a poor tool 
because it doesn't turn screws very well.

-- 

Rick

[toc] | [prev] | [next] | [standalone]

#26710

> That said, the GA144 will not be used

Wonderfull, we are coming to
the same conclusion.
Oddly enough pretty much the same one
every one else has come to.

MfG  JRD

[toc] | [prev] | [next] | [standalone]

#26801

I agree.

1 pc running firefox with network and firefox ish browser and then microsoft and longsoon and arm are all running in fear since forth chip would probably cost 1% and use 1% power.

ret are racing to get everyoen on proprietary cell networks b4 everyone switches to openbsd desktop running icewm or to a colorforth 144 chip greenaray box.

when someoen does a aweb start up on 10 144 green array boxes and can gnerate fast webapps with 1% power and money of amd or intel server pizzaboxes then everyoen will SHIT

[toc] | [prev] | [next] | [standalone]

#26602

> Intel is paying for at least one of the patents.  I found an article 
> that says ARM was resolved of infringement of at least one of the 
> patents.  So I'm not sure what products are left that aren't using one 
> of those two sources of CPUs? 

http://patriotscientific.com/index.php?option=com_content&task=view&id=307&Itemid=29

[toc] | [prev] | [next] | [standalone]

#26613

In article <1aa7a8a7-b63a-401a-b1d7-398c4eb8e6d9@googlegroups.com>,
Chris  <xrissmith@me.com> wrote:
>> Intel is paying for at least one of the patents.  I found an article
>> that says ARM was resolved of infringement of at least one of the
>> patents.  So I'm not sure what products are left that aren't using one
>> of those two sources of CPUs?
>
>http://patriotscientific.com/index.php?option=com_content&task=view&id=307&Itemid=29

I would be interested whether Intel pays rightly, i.e. not by extorsion.
Is Intel really using idea's from the portfolio that they don't came up
with themselves, or is it a matter of Lecrone being the first to patent
an obvious idea in Chuck Moore's name that Intel overlooked to patent?

(Of course Intel pays legitimately, i.e. according to lex, the law.
It is a matter whether the law reflects right. We know in patents it
often doesn't.)

Groetjes Albert
-- 
Albert van der Horst, UTRECHT,THE NETHERLANDS
Economic growth -- being exponential -- ultimately falters.
albert@spe&ar&c.xs4all.nl &=n http://home.hccnet.nl/a.w.m.van.der.horst

[toc] | [prev] | [next] | [standalone]

#26615

On Thursday, October 17, 2013 5:14:46 AM UTC-7, Albert van der Horst wrote:
> In article <1aa7a8a7-b63a-401a-b1d7-398c4eb8e6d9@googlegroups.com>,
> [snip]
> I would be interested whether Intel pays rightly, i.e. not by extorsion.
> Is Intel really using idea's from the portfolio that they don't came up
> with themselves, or is it a matter of Lecrone being the first to patent
> an obvious idea in Chuck Moore's name that Intel overlooked to patent?
> 
> (Of course Intel pays legitimately, i.e. according to lex, the law.
> It is a matter whether the law reflects right. We know in patents it
> often doesn't.)
> [snip]

In 2004-5 an Intel lawyer using a friend's login came on clf looking for info 
about Chuck's ideas. He was outed and then revealed that Intel engineers were 
starting to use the same techniques (2004) that Chuck had 'invented' (and 
patented) ten years earlier. The result was a legally complicated mess that's 
still playing out today.

-John

[toc] | [prev] | [next] | [standalone]

#26627

On Thu, 17 Oct 2013 08:14:46 -0400, Albert van der Horst  
<albert@spenarnc.xs4all.nl> wrote:

> I would be interested whether Intel pays rightly, i.e. not by extorsion.

We all know you meant "extortion" not "extorsion". 's' -> 't'

But, it's kind of funny...  Yes, both are English words.


Rod Pemberton

[toc] | [prev] | [next] | [standalone]

#26784

I would like to have a go with a GA144 evaluation board. I must get around to buying one.

-The GA144's well thought out machine instructions look interesting.
-polyFORTH is cool, I am still using 16bit polyFORTH under MS-DOS!!!?

GreenArrays has just updated arrayForth®
User's Manual

Updated to correspond with software release 02b. IDE and framer support programming both chips on an EVB. A new booting option loads one or both chips using a minimal boot stream through serial port, considerably faster than the IDE method, without altering flash. polyFORTH F18 code now supports Ethernet daughter board.

[toc] | [prev] | [next] | [standalone]

#26802

work just bought 6 rasberry pi mini pcs for little simple jobs around the place

if g144 chip could run firefox and or be used to generate dynamic html for e commerce letting you skip amd and intel and arm then it could take off bigtime

I wonder if netbsd could be ported to g144??

if check can emualte a unix, hek most web startups are on debianm which is garbage comapred to netbsd or freebsd or archlinux

provide a platform!!

[toc] | [prev] | [standalone]

Page 2 of 2 — ← Prev page 1 [2]

Back to top | Article view | comp.lang.forth

csiph-web