Dual-stack (Forth) processors

"Martin Euredjian" <0_0_0_0_@pacbell.net> wrote in message
news:5vkZb.27977$_76.2895@newssvr25.news.prodigy.com...
> rickman wrote:
> > You had to go and say that, didn't you!  This is being posted to the
> > Forth newsgroup and you will hear a few comments about this...  ;)
>
> I realize that.  No flames please.  I have two decades of Forth
experience.
> I'm from the days when you built your own computer from chips, wrote a
> monitor, got Forth in there, wrote your own editor and then developed your
> apps.  I enjoy and love Forth.  I truly do.  I also love a language called
> APL.  I think it should rule the World.

Heh.  Heheh.

I learned APL when I was a kid.  I love it too.  It taught me respect for
languages that
don't have an implicit order of operations.

I, too, have been programming in Forth for 20+ years.  I wrote a multi-user
virtual
reality system that used byte-coded Forth to move virtual objects between
hosts
on the Internet.  One of the things that struck me is that's it's less
awkward to write
long floating-point expressions multiplying sine and cosine terms in Forth
than it is in C.  That's
why Forth came to mind as language for designing filters, etc.  Also for
physical modeling
of musical instruments.  It's easy to build up waveguides, etc into function
blocks that can
be strung together on the command line.  *Everything* is better with an
interactive command
prompt.

The scripting language for the virtual reality objects was Forth, but the
Forth interpreter itself was written in
plain-vanilla C for portability.  It ran fast enough (on Pentium 90s with
software-only 3D rendering!)
that the extra work of an assembler implementation wasn't worth it.  I used
Forth exactly where
it made the most sense -- as a platform-independent VM-based object code -- 
and used C in a
way that reflected _its_ strengths.  I chose the languages based on their
technical merits for a particular
application, something I'd like to see more programmers do.

The majority of the coding that I have done, both for pay and for
self-directed projects, has been
in C.  I have been living in the same world of wanting portability,
maintainability and ease of finding programmers.

My goals for asking about DSP and Forth processors:

I'm teaching myself digital design and DSP.

My first substantial FPGA project was a minimal instruction set 16-bit Forth
processor.  It simulates at 90+ MIPS
in a small Altera Cyclone part.  It uses about 800 LEs.

Looking at the Stratix family, I see they're chock-full of optimized
multiply-and-accumulate blocks.  Very nice!
I did a stint at Korg R&D years ago, working on a legendary unreleased synth
(the OASys).  It was my intro
to state-of-the-art synthesis, using physical modeling of instruments.  At
that time, I would have very much
liked to build/buy some hardware to experiment with at home, but the cost
was prohibitive.  Now, I have
a real chance to build some music hardware.

And hardware for separating signals on the ham radio bands.  And hardware to
develop multi-processing
ideas I've been toying with for years.

I'm going to experiment and read all the papers I can get my hands on and
pursue what's interesting, and if
something I do looks like it has better cost/performance than something on
the market, I'll look into commercializing
it.

After I have a project or two that I have physically realized, maybe someone
will give me a job doing FPGA design.

-Davka

Davka wrote:

> I learned APL when I was a kid.  I love it too.  It taught me respect for
> languages that don't have an implicit order of operations.

Small World, isn't it!
The other thing that's nice about APL is the ability to abstract away from
the mechanics of programming and live within the problem domain.


> My goals for asking about DSP and Forth processors:
>
> I'm teaching myself digital design and DSP.
<SNIP>
> My first substantial FPGA project was a minimal instruction set 16-bit
Forth
> processor.  It simulates at 90+ MIPS in a small Altera Cyclone part.  It
uses about 800 LEs.
<SNIP>
> state-of-the-art synthesis, using physical modeling of instruments.
<SNIP>
> liked to build/buy some hardware to experiment with at home, but the cost
> was prohibitive.  Now, I have a real chance to build some music hardware.

Have you consider just doing it the FPGA way?  I haven't stopped to think
about what's required but you can certainly create building blocks in
hardware (say, filters).  Also, with the Virtex 2 Pro family you have the
ability to use the PowerPC processor/s as, well, processors and the rest of
the FPGA is available for custom peripherals (if you want to think of it
that way).

Of course, an optimized Forth machine would/could make it very interactive
and fun to play with.  Is there an implementation of Forth for PowerPC?

Sounds like you are going to have some fun.



-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Martin Euredjian

To send private email:
0_0_0_0_@pacbell.net
where
"0_0_0_0_"  =  "martineu"

"Martin Euredjian" <0_0_0_0_@pacbell.net> wrote in message
news:KX7%b.2619$6k.0@newssvr27.news.prodigy.com...
> Davka wrote:
>
> > I learned APL when I was a kid.  I love it too.  It taught me respect
for
> > languages that don't have an implicit order of operations.
>
> Small World, isn't it!
> The other thing that's nice about APL is the ability to abstract away from
> the mechanics of programming and live within the problem domain.

You're not the first fellow nerd that I've met who likes both Forth
and APL.  :-)

> > was prohibitive.  Now, I have a real chance to build some music
hardware.
>
> Have you consider just doing it the FPGA way?  I haven't stopped to think
> about what's required but you can certainly create building blocks in
> hardware (say, filters).  Also, with the Virtex 2 Pro family you have the
> ability to use the PowerPC processor/s as, well, processors and the rest
of
> the FPGA is available for custom peripherals (if you want to think of it
> that way).

That's the plan.  I've already put dozens of hours into learning the Altera
tool
and their chip families, so that's the way I will go initially.  They have
their
own tiny RISC implementation (called NIOS), so if I don't do my own
processor,
I'll use that one.

The idea for the system architecture is to have a bunch of DSP modules of
various flavors
in the hardware which can be dynamically allocated by the software system
that manages them.

> Of course, an optimized Forth machine would/could make it very interactive
> and fun to play with.  Is there an implementation of Forth for PowerPC?

:-)  I think it's called Open Firmware.

-Davka

Davka wrote:
> 
> "Martin Euredjian" <0_0_0_0_@pacbell.net> wrote in message
> news:KX7%b.2619$6k.0@newssvr27.news.prodigy.com...
> > Davka wrote:
> >
> > > I learned APL when I was a kid.  I love it too.  It taught me respect
> for
> > > languages that don't have an implicit order of operations.
> >
> > Small World, isn't it!
> > The other thing that's nice about APL is the ability to abstract away from
> > the mechanics of programming and live within the problem domain.
> 
> You're not the first fellow nerd that I've met who likes both Forth
> and APL.  :-)
> 
> > > was prohibitive.  Now, I have a real chance to build some music
> hardware.
> >
> > Have you consider just doing it the FPGA way?  I haven't stopped to think
> > about what's required but you can certainly create building blocks in
> > hardware (say, filters).  Also, with the Virtex 2 Pro family you have the
> > ability to use the PowerPC processor/s as, well, processors and the rest
> of
> > the FPGA is available for custom peripherals (if you want to think of it
> > that way).
> 
> That's the plan.  I've already put dozens of hours into learning the Altera
> tool
> and their chip families, so that's the way I will go initially.  They have
> their
> own tiny RISC implementation (called NIOS), so if I don't do my own
> processor,
> I'll use that one.
> 
> The idea for the system architecture is to have a bunch of DSP modules of
> various flavors
> in the hardware which can be dynamically allocated by the software system
> that manages them.

I found a mention of a Forth for DSP called FROTH.  I believe it is an
open source project, but the page does not seem to have been updated in
quite a while.  Some of the other posters here may know more about it.  

-- 

Rick "rickman" Collins

rick.collins@XYarius.com
Ignore the reply address. To email me use the above address with the XY
removed.

Arius - A Signal Processing Solutions Company
Specializing in DSP and FPGA design      URL http://www.arius.com
4 King Ave                               301-682-7772 Voice
Frederick, MD 21701-3110                 301-682-7666 FAX

"Davka" <mygarbagepail@hotmail.com> wrote in message news:<W01%b.2$mA3.6454@news.uswest.net>...

> Heh.  Heheh.
> 
> I learned APL when I was a kid.  I love it too.  It taught me respect for
> languages that
> don't have an implicit order of operations.
> 
> I, too, have been programming in Forth for 20+ years.  I wrote a multi-user
> virtual
> reality system that used byte-coded Forth to move virtual objects between
> hosts
> on the Internet.  One of the things that struck me is that's it's less
> awkward to write
> long floating-point expressions multiplying sine and cosine terms in Forth
> than it is in C.  That's
> why Forth came to mind as language for designing filters, etc.  Also for
> physical modeling
> of musical instruments.  It's easy to build up waveguides, etc into function
> blocks that can
> be strung together on the command line.  *Everything* is better with an
> interactive command
> prompt.
> 
> The scripting language for the virtual reality objects was Forth, but the
> Forth interpreter itself was written in
> plain-vanilla C for portability.  It ran fast enough (on Pentium 90s with
> software-only 3D rendering!)

Hmmm...that reminds me quite a bit of the "Metatopia" project.

http://metatopia.sourceforge.net/
http://www.immersive.com/

Do you still have the code?  Are you doing anything with it?

> My goals for asking about DSP and Forth processors:
> 
> I'm teaching myself digital design and DSP.
> 
> My first substantial FPGA project was a minimal instruction set 16-bit Forth
> processor.  It simulates at 90+ MIPS
> in a small Altera Cyclone part.  It uses about 800 LEs.

Cool.  You should get this added to Jeff Fox's "Forth chips" page.

> And hardware for separating signals on the ham radio bands.  And hardware to
> develop multi-processing
> ideas I've been toying with for years.

Interesting.  I do think that wireless processing are the "new frontier" for
hardware hackers.  I have scene a packat modem implemented in FPGA.  I'll
send you the link if you're interested.  Jeff should be able to give you
some info of multi-processing MISC since that's something he's worked on
in the past and seems to be the focus of the latest Chuck work (25x).

Regards,

John M. Drake

"Martin Euredjian" <0_0_0_0_@pacbell.net> wrote in message news:<KX7%b.2619$6k.0@newssvr27.news.prodigy.com>...

> Have you consider just doing it the FPGA way?  I haven't stopped to think
> about what's required but you can certainly create building blocks in
> hardware (say, filters). 

I wasn't aware that the "FPGA way" and a Forth processor were mutually
exclusive.  Certainly Forth CPUs have had DSPs built into them before.

> Of course, an optimized Forth machine would/could make it very interactive
> and fun to play with.  Is there an implementation of Forth for PowerPC?

Mops can generate PowerPC code.  As for embedded PowerPC I was surprised
that I didn't see a SwiftX or VFX implementation on their websites
although both supported ColdFire.

Regards,

John M. Drake

jmdrake wrote:
> 
> "Martin Euredjian" <0_0_0_0_@pacbell.net> wrote in message news:<KX7%b.2619$6k.0@newssvr27.news.prodigy.com>...
> 
> > Have you consider just doing it the FPGA way?  I haven't stopped to think
> > about what's required but you can certainly create building blocks in
> > hardware (say, filters).
> 
> I wasn't aware that the "FPGA way" and a Forth processor were mutually
> exclusive.  Certainly Forth CPUs have had DSPs built into them before.
> 
> > Of course, an optimized Forth machine would/could make it very interactive
> > and fun to play with.  Is there an implementation of Forth for PowerPC?
> 
> Mops can generate PowerPC code.  As for embedded PowerPC I was surprised
> that I didn't see a SwiftX or VFX implementation on their websites
> although both supported ColdFire.

I believe that is because both companies are reactive rather than
proactive.  They port to a new platform when they are paid to do it. 
They don't speculate on their own dime.  

-- 

Rick "rickman" Collins

rick.collins@XYarius.com
Ignore the reply address. To email me use the above address with the XY
removed.

Arius - A Signal Processing Solutions Company
Specializing in DSP and FPGA design      URL http://www.arius.com
4 King Ave                               301-682-7772 Voice
Frederick, MD 21701-3110                 301-682-7666 FAX

rickman <spamgoeshere4@yahoo.com> wrote in message news:<403E4817.2E1B127A@yahoo.com>...
> jmdrake wrote:

> > Mops can generate PowerPC code.  As for embedded PowerPC I was surprised
> > that I didn't see a SwiftX or VFX implementation on their websites
> > although both supported ColdFire.
> 
> I believe that is because both companies are reactive rather than
> proactive.  They port to a new platform when they are paid to do it. 
> They don't speculate on their own dime.  
> 
> -- 
> 
> Rick "rickman" Collins

Right.  I guess I'm surprised that there (apparantly) hasn't been a 
request for PowerPC.  Although I suppose it makes sense.  Perhaps
ColdFire is a better fit for embedded systems price/performance
wise then PowerPC?

Anyway Forth Inc does have a PSC1000 port and Patriot Scientific
does have a PSC1000 core for use in FPGAs.

Regards,

John M. Drake

On Thu, 26 Feb 2004 14:25:11 -0500, rickman <spamgoeshere4@yahoo.com>
wrote:

>> Mops can generate PowerPC code.  As for embedded PowerPC I was surprised
>> that I didn't see a SwiftX or VFX implementation on their websites
>> although both supported ColdFire.
>
>I believe that is because both companies are reactive rather than
>proactive.  They port to a new platform when they are paid to do it.
Sorry Rick, but it isn't always true. We certainly port when paid
to do so, but we also do it when we like the CPU. The ARM and
Coldfire VFX ports were done on our own dime. There are several
CPUs we've done cross compilers for that I wish we hadn't wasted
time on. We just don't have the budget to do every CPU on the
market. In retrospect I consider the decision to support ARM
rather than PPC a good one.

>They don't speculate on their own dime.  
Oh yes we do. The VFX code generator was a speculative venture.
Yes, we do it with commercial (and sometimes cautious) interest,
but MPE at least has a long track record in speculative R&D
ventures, including what eventually became the basis of the
Europay OTA system.

Stephen
--
Stephen Pelc, stephenXXX@INVALID.mpeltd.demon.co.uk
MicroProcessor Engineering Ltd - More Real, Less Time
133 Hill Lane, Southampton SO15 5AF, England
tel: +44 (0)23 8063 1441, fax: +44 (0)23 8033 9691
web: http://www.mpeltd.demon.co.uk - free VFX Forth downloads

John M. Drake wrote:

> > Have you consider just doing it the FPGA way?  I haven't stopped to
think
> > about what's required but you can certainly create building blocks in
> > hardware (say, filters).
>
> I wasn't aware that the "FPGA way" and a Forth processor were mutually
> exclusive.

Neither am I.  :-)

But, seriously, there are a lot of things that you cannot do with a
sequential processor, Forth or otherwise.  Example:  32 tap FIR filter
running at 150MHz, one result per clock.  What you can do is have a little
Forth (or whatever) processor within the FPGA that loads coefficients into
that filter.  That's what I implied by the "FPGA way":  Use the high-speed
capabilities of FPGA's to develop custom processing sub-blocks that a Forth
(or something else) processor can then pull strings on.  This in contrast to
insisting that everything be done with a Forth machine just out of
attachment to the language (note that the OP is not suggesting that at all).


-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Martin Euredjian

To send private email:
0_0_0_0_@pacbell.net
where
"0_0_0_0_"  =  "martineu"