[comp.parallel] Applicative languages on transputer systems

wg@tub.UUCP (Wolfgang Grieskamp) (06/28/90)

Hello!

The OPAL group TU-Berlin plays with an experimental applicative 
language to research transformations to real efficient code for
sequential and parallel machines. The language OPAL has pure 
applicative evaluation order but a builtin STREAM concept which seems 
to be closely related to the transputer architecture. 

I'm searching for some information about the implementation of 
applicative languages on transputer systems or related ones.    
Imperative languages are also of interest, but in general reference 
transparency drops many hard problems and opens mind for others.

I would be glad to get hints according to one of the following
topics:

o       trusty papers about the transputer  from the compiler 
        constructors viewpoint.

o       availability of T800 configuration emulators which support 
        communication/synchronization monitoring (as far such is possible) 
        under sun3 or sun4.

o       availability of OCCAM for sun3 or sun4 machines.

o       experiences/opinions/papers about the compilation of applicative 
        languages for transputers or related systems in general:

        o       Dataflow, abstract interpretation or other approaches to 
                calculate the distribution of stream consumers/producers 
                throughout a static network at compile time. 

        o       The problem of mapping dynamic nets to static ones. 
                Approaches to estimate the extension and structure of 
                a dynamic subnet?
        
        o       In OPAL we express (at source level) large-grain parallelism
                via recursive network equations which connect stream 
                processors.  The user would not like to write down nets when 
                he/she actually thinks in plain functions; the compiler must 
                exploit implied parallelism of high-cost applications using 
                source transformations to produce weak equivalent networks.
                Any researches in this direction?
                
o       Critical discussions of the stream concept at all. Alternate
        ways to handle communication and parallelism at source level? 

Many thanks for your effort (I will summarize on interest).


- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Wolfgang Grieskamp       "Die Welt ist tief! Und tiefer als der Tag gedacht!" 
wg@opal.cs.tu-berlin.de        tub!tubopal!wg         wg%opal@DB0TUI11.BITNET