Path: utzoo!censor!geac!torsqnt!news-server.csri.toronto.edu!cs.utexas.edu!usc!apple!agate!darkstar!cs.washington.edu
From: pardo@cs.washington.edu (David Keppel)
Newsgroups: comp.os.research
Subject: Re: Toy Experiment with Shared Memory Programming
Message-ID: <8859@darkstar.ucsc.edu>
Date: 10 Nov 90 18:34:44 GMT
Sender: usenet@darkstar.ucsc.edu
Organization: University of Washington, Computer Science, Seattle
Lines: 23
Approved: comp-os-research@jupiter.ucsc.edu


jms@central.cis.upenn.edu (Jonathan M. Smith) writes:
>[Programming with pipes vs. shared memory]

Some other lessons from the experiment:
* Programming with good, familiar primitives is always easier.
* Never confuse MODEL with IMPLEMENTATION -- there are pipes that
  are implemented using shared memory (Mach, Topaz) and shared memory
  implemented using pipes (more or less, anyway; Ivy).
* The results are always different on a multiprocessor: the
  shared-memory version has a simple implementation that (after setup)
  makes no kernel calls and does no copying -- at least on some
  platforms.
* There is also a pipe model, implemented using shared memory, that
  runs on (some) multiprocessors with no kernel calls and no copying.
* Simple, understandable examples are almost controversial :^)

Thank you for this opportunity to generalize...

    ;-D on  ( General Semaphore?  This is Private Memory.  Sir! )  Pardo
-- 
		    pardo@cs.washington.edu
    {rutgers,cornell,ucsd,ubc-cs,tektronix}!uw-beaver!june!pardo