Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!iuvax!cica!gatech!hubcap!gmt
From: gmt@arizona.edu (Gregg Townsend)
Newsgroups: comp.parallel
Subject: Re: Announcing SR version 1.1
Summary: answers to questions
Keywords: SR distributed parallel language release
Message-ID: <6164@hubcap.clemson.edu>
Date: 4 Aug 89 15:02:28 GMT
Sender: fpst@hubcap.clemson.edu
Lines: 63
Approved: parallel@hubcap.clemson.edu

In article <6119@hubcap.clemson.edu>, James.Price.Salsman@cat.cmu.edu
asks some questions about SR:

    - Could you elaborate on "An SR program runs on one or more networked
	machines of the same architecture."?

All SR programs execute using lightweight processes.  These may run in
multiple Unix processes, but there are always many more lightweight
processes than Unix processes.

Depending on how a particular program is structured, the lightweight
processes are distributed
    (1) within a single Unix process
or  (2) among multiple Unix processes on a single host
or  (3) among multiple processes on multiple hosts (using rsh(1))

The details are handled within SR;  the programmer just compiles the program
and runs "a.out".  However, the multiple machines must have a close enough
hardware and software environment to let them run the same executable file.

    - How well does it compile on to various types of parallel hardware?
    - Why is the Multimax the only MIMD system that runs SR?

Right now SR doesn't take any advantage of the parallelism except where
the multiple Unix processes of cases (2) or (3) run simultaneously.
We would like to address this in the future.  For now, the Multimax
is in the list simply because one happened to be handy.

    - Any plans for a port to a multiprocessor chip like the transputer
	or the i860?

No plans right now.

    - Does SR have any of Icon's string scanning and backtracking?
	[For those who don't know, Icon is from Arizona, too.]

Nope.  SR is in most respects fairly traditional, and has more of a
Pascal flavor than anything else.

[Thanks to Ron Olsson of UC Davis for answering the rest of these:]

    - Could you compare & contrast SR with Ada's task model, Occam,
	and the BBN Butterfly's MultiScheme?

SR's process and interprocess communication model is similar in some
ways to Ada and Occam, but is considerably more flexible.  The Jan 88
TOPLAS paper gives the details.

    - What is meant by "synchronizing?"  To be more specific, is the term
	used in the asynch/sync sense or the synch/out-of-synch sense?

The former.

    - Has the underlying paradigm been formalized as in Hoare's CSP?

No.

    - Can the denotational semantics be described in terms of temporal logic?

Not sure exactly what you mean here.  Some work on using continuations
to express concurrency in denotational semantics has appeared recently
(maybe in POPL?).  However, its relationship, if any, to temporal
logic is not at all clear.