🚀 go-pugleaf

In article I wrote:
|> 
|> Hi,
|> 
|> assumed
|> 1. I have a SUN but no Transputer-Board and
|> 2. I want to compile OCCAM programs for Transputer-Networks and
|> 3. I want to test if the programs are correct
|> 
|> What can I do?


The main hint was given by Iain Phillips (Iain.Phillips@cs.man.ac.uk)
who suggested to use the Toolset "isim" simulator.

This program simulates a T425-Transputer on a SUN and it IS
possible to simulate great transputer networks if you set
the UNIX environment variables IBOARDSIZE to something large.

Additionally you have to declare the memsize in the .pgm-file
sufficiently large (e.g. SET Root(type, memsize := "T425", 104*mega) ).

I tried it with a 26 Transputer network without any problems!


Many thanks to Ian,

                          Alf.


-- 
*     Alf Wachsmann                | Uni/GH Paderborn                  *
*                                  | FB 17 Informatik                  *
*     e-mail: alf@uni-paderborn.de | Warburger Str. 100                *
*     Phone:  +49 5251 60 3312     | D-W-4790 Paderborn                *
*                                  | Germany                           *

In Article 61 of comp.sys.transputer:
alf@uni-paderborn.de (Alf Wachsmann) writes

> The main hint was given by Iain Phillips (Iain.Phillips@cs.man.ac.uk)
> who suggested to use the Toolset "isim" simulator.
> 
> This program simulates a T425-Transputer on a SUN and it IS
> possible to simulate great transputer networks if you set
> the UNIX environment variables IBOARDSIZE to something large.
> 
> Additionally you have to declare the memsize in the .pgm-file
> sufficiently large (e.g. SET Root(type, memsize := "T425", 104*mega) ).
> 
> I tried it with a 26 Transputer network without any problems!

The isim program is only a single processor simulation, what you must have
been doing is to configure multiple configuration level processes on a single
processor, isim doesn't mind this as it is trying to simulate the execution
of this as would happen on a real transputer. You might have 26 configuration
level processes in your system but you don't have 26 PROCESSORS defined to
the configurer. Of course with having identical communications instructions on
the transputer whether using internal channels or hard links you should be 
able to assume that a 26 configuration level process system 
will easily spread over a multiprocessor system.

also in Article 65 of comp.sys.transputer:
luigi@pical2.uucp (Luigi Rizzo) writes

> Recently there has been some discussion on isim. As pointed out by some
> users, the Unix (well, SUN) version has some advantages over the real
> thing because it can emulate a system with a lot of memory (if I'm not
> wrong, limited only by swap space ?); I don't think the
> Transputer-based isim (the one shipped with MSDOS systems) has this
> ability.

isim is a functionally similar simulation of the transputer, there is no
guarantee that the results of running on the transputer will agree with the
results of running on isim (although we have tried to make isim as good as
possible). isim is a tool for debugging low level software and in this
case, and this case alone isim has an advantage over the transputer, also
it is considerably slower than any real transputer (you might get a few 10s of
K transputer instructions/sec). For anyone who wants to run a transputer
application they should run on real transputer hardware, this gets the customer
accuracy of execution model and execution speed.
In the IMSD7214 and IMSD7205 products isim runs on a transputer board and
is limited to the physical memory available - this overcomes the standard
hard limit of MS-DOS 640k memory limits. If you had a 104MB TRAM isim would
quite happily let you use this memory.

> but we'd still like to get/buy a version of isim running on a
> DECstation. Is such a thing available ...

NO

> As an aside, it would be VERY NICE of Inmos if the source for isim were
> made publicly available, same as they did with their servers and other
> stuff (e.g. the 'check' program). I don't think this would stop people
> from buying Transputers!

If anyone wants to develop a public domain transputer simulator they are 
quite at liberty to do so. INMOS do not have a policy of giving away 
valuable intellectual property (especially that containing proprietary 
information). You have already demonstrated that the availability of isim 
has had some effect in stopping people from buying transputers.


regards,

Steve

Steve Doyle, Software Marketing, INMOS Ltd   | Tel +44 454 616616
1000 Aztec West                              | 
Almondsbury                                  | UK: steved@inmos.co.uk
Bristol BS12 4SQ, UK                         | US: steved@inmos.com

In article <1992Oct2.090607.334@titan.inmos.co.uk> 
steved@lion.inmos.co.uk (Stephen Doyle) writes:
[lots of good comments deleted]

   > but we'd still like to get/buy a version of isim running on a
   > DECstation. Is such a thing available ...

   NO

Pray, why not? (see below)

   > As an aside, it would be VERY NICE of Inmos if the source for isim were
   > made publicly available, same as they did with their servers and other
   > stuff (e.g. the 'check' program). I don't think this would stop people
   > from buying Transputers!

   If anyone wants to develop a public domain transputer simulator they are 
   quite at liberty to do so. INMOS do not have a policy of giving away 
   valuable intellectual property (especially that containing proprietary 
   information). You have already demonstrated that the availability of isim 
   has had some effect in stopping people from buying transputers.

__FLAME ON__

It is this stupid attitude that over the years has COST Inmos a lot
of sales. Why do you think other manufacturers/OEMS "give away valuable
intellectual property"? Because they know that they have a lot of
potential customers that just don't have the money to buy a sample
of everything just to evaluate a product. Because they know that
just these cashstrapped customers (university departments etc) will act as
multipliers once they get to places where money IS available. Do you
think somebody who probably has the virtual memory on a <insert
favourite workstation> is going to buy 26 transputers with 104MB
just for a test? Sheesh, get real! Are you telling me you can't easily
strip any proprietary information from isim (most of which is probably
widely known anyway, as the PD disassembler from Inmos contains a lot
of undocumented instructions...)?

The same arguments, of course, go for the occam compiler. Thankfully,
things seem to be getting better in recent times, but I just had the
feeling of being back in the dark middle ages!

__FLAME OFF__

Have a nice weekend, guys 'n girls, I'm feeling much better now.

in Article 79 of comp.sys.transputer:
jan@hermes.neuroinformatik.ruhr-uni-bochum.de (Jan Vorbrueggen)
writes:

   > but we'd still like to get/buy a version of isim running on a
   > DECstation. Is such a thing available ...

   NO

> Pray, why not? (see below)

There is little interest from our customer base in the DECstation (at least
in context of other development hosts), for the benefit of our customers
we focus attention on their preferred host platforms.

> __FLAME ON__
> ... rest deleted

Universities do great work at the grass roots level of using and developing
newer technologies, however, proliferation of a technology is not done by 
Universities but through volume design ins at major OEMs. My own experience 
(and this comes from the systems company I worked for before joining INMOS) 
is that the basics taught at Universities are valued, however the choice of 
one technology over another often relates more to commecial issues than 
technical issues (e.g. if companies x,y,z are using it we should). Intels 
fortunes now are highly focused on the PC market and volumes came from a major 
design in at IBM. Do Intel run simulators or emulators on anything other than 
Intel processors, do Sun have a SPARC simulator that runs on HP PA or 
DECstation architectures?

It would be awfully good for Ford to increase popularity by giving cars to
University students and commercially selling all their car models for < 10$ 
but I don't expect them to commit commecial suicide in doing this.

regards,

Steve 
(living and working in the real world)

These opinions are my own and do not reflect INMOS policy.

Steve Doyle, Software Marketing, INMOS Ltd   | Tel +44 454 616616
1000 Aztec West                              | 
Almondsbury                                  | UK: steved@inmos.co.uk
Bristol BS12 4SQ, UK                         | US: steved@inmos.com

In article <1992Oct5.091626.22956@titan.inmos.co.uk> steved@lion.inmos.co.uk
(Stephen Doyle) writes:
   There is little interest from our customer base in the DECstation (at least
   in context of other development hosts), for the benefit of our customers we
   focus attention on their preferred host platforms.

Hmmm. So am I to deduce that the company that invented a formally strict
language produces software (presumably, in this case, in C) that is so non
standard that a significant porting effort to the gcc compiler for the DEC
station (and all those other MIPSbased machines from MIPS, SGI, ...) would be
required? And even if so, I'm sure you'd easily find a volunteer to do it for
you (nondisclosure and all). The thing doesn't even have to simulate
the hard part, the FPU!
And don't tell me about quality, manuals etc 
we want this as an addon, not necessarily a supported product.

   Universities do great work at the grass roots level of using and developing
   newer technologies, however, proliferation of a technology is not done by
   Universities but through volume design ins at major OEMs. 

Quite right, but how are these decisions made?

   It would be awfully good for Ford to increase popularity by giving cars to
   University students and commercially selling all their car models for < 10$
   but I don't expect them to commit commecial suicide in doing this.

Nya, get real. We're talking about software, which has low incremental cost.
Why do you think DEC had a 10%oflistprice policy for software for uni
versities, and now has the CSLG?

	Jan

    Stephen Doyle writes:
       If anyone wants to develop a public domain transputer simulator they are 
       quite at liberty to do so. INMOS do not have a policy of giving away 
       valuable intellectual property (especially that containing proprietary 
       information). You have already demonstrated that the availability of isim 
       has had some effect in stopping people from buying transputers.


this is completely bogus nonsense.

isim is a simple portable C implementation of a transputer instruction
set simulator.  any half-competent programmer could write one
relatively quickly.  it might even go faster.  the only "proprietary
information" isim contains is some trivial undocumented loader
information so that it doesnt need to simulate link booting.

i'm afraid your "real world" looks nothing like mine.  most of your
competitors do indeed give away source for instruction set simulators.