Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!uwm.edu!cs.utexas.edu!tut.cis.ohio-state.edu!ucbvax!prlvax1.prl.philips.co.uk!EDMONDS
From: EDMONDS@prlvax1.prl.philips.co.uk (Mark from Philips)
Newsgroups: comp.sys.transputer
Subject: (none)
Message-ID: <900924121517.00002C28131@prlvax1.prl.philips.co.uk>
Date: 24 Sep 90 12:15:17 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 56

A thought occurred to me the other day.  Some people can't afford transputer 
hardware, even though it is relatively cheap these days.  It would also be 
nice to see occam running on other machines - so what can one do?

What would people think if, rather than using transputers, networks of
workstations were used instead.  Thus, you would write your occam code as
usual; however, in the configuration description (PROGRAM fold), rather than
saying: 


            PLACED PAR
              PROCESSOR 0 T8
              ... placed channels and SC PROC

              PROCESSOR 1 T8
              ... placed channels and SC PROC

              PROCESSOR 2 T8
              ... placed channels and SC PROC


etc, you would write:


            PLACED PAR
              PROCESSOR 0 node_name
              ... placed channels and SC PROC

              PROCESSOR 1 node_name
              ... placed channels and SC PROC

              PROCESSOR 2 node_name
              ... placed channels and SC PROC


where node_name (yes, you've guessed it) is the name of a node (workstation) 
on your network.  Communication (if one is running under UNIX) would take 
place via sockets rather than links.  One big advantage is that you are not 
limited to a fixed number of physical channels - instead you have virtual 
channels!  This makes it somewhat compatible with the H1.

Is this a practical solution?

Would people be interested?

We have a package that would allow you to do this easily, if only we had the
source to the occam compiler/configurer, etc. 

It means that you can achieve parallel processing, without buying transputers, 
and it makes occam more attractive to run on non-transputer based machines.


Any comments?

Mark Edmonds
Philips Research Labs