Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!usc!wuarchive!rex!ames!uhccux!virtue!ccc_ldo
From: ccc_ldo@waikato.ac.nz (Lawrence D'Oliveiro, Waikato University)
Newsgroups: comp.sys.mac.system
Subject: Re: PPC, IAC, and True Multitasking (tm)
Message-ID: <1377.26dc1ba9@waikato.ac.nz>
Date: 29 Aug 90 07:46:49 GMT
References: <1990Aug3.040513.14844@d.cs.okstate.edu> <2760@bridge2.ESD.3Com.COM> <13888@cbmvax.commodore.com> <923@mdavcr.UUCP>
Distribution: na
Organization: University of Waikato, Hamilton, New Zealand
Lines: 48

In <923@mdavcr.UUCP>, ewm@mdavcr.UUCP (Eric W. Mitchell) says:

"preemptive multitasking slows down the system because it adds to
the number of cycles required to perform a given operation."

First of all, I'd like to say that I basically agree with Eric's
viewpoint that the Mac *is* a multitasking system, albeit a non-
preemptive one.

However, even among preemptive systems, there're two broad strategies
for divvying up the CPU time among tasks: time-slicing, and
non-time-slicing.

A time-slicing system forces a fair distribution of CPU time among
a number of tasks with the same priority. It assigns a "quantum" of
CPU time to each tasks, and does a reschedule immediately the current task
exhausts its quantum. The suspended task gets a new quantum, but
it goes to the end of the queue so it won't run again until every other
task at the same priority level has had a chance.

A non-time-slicing system only suspends the current task if a higher-
priority one becomes computable. If such an event never happens,
and the current task never goes into a voluntary wait state (such
as waiting for an I/O operation to complete), it will continue running
without interference from the scheduler.

Time-slicing schemes evolved on timeshared multi-user systems, which
need to protect themselves against compute-bound jobs hogging all the
CPU and hurting the response time of other users. I consider their use
inappropriate in a single-user environment.

Consider: a PC is dedicated to serving the needs of one user. Whether
it runs a preemptive or non-preemptive OS, all the tasks are still
assumed to be *cooperating* in their service of a single user--you're
not running a timesharing service on your desktop.

Therefore, Eric's comment about preemptive multitasking *necessarily*
involving extra overhead, I don't agree with. Provided you design
the system specifically as an interactive single-user one, and not as
a cut-down version of a multiuser one, the overhead can be kept very
low indeed.

Lawrence D'Oliveiro                       fone: +64-71-562-889
Computer Services Dept                     fax: +64-71-384-066
University of Waikato            electric mail: ldo@waikato.ac.nz
Hamilton, New Zealand    37^ 47' 26" S, 175^ 19' 7" E, GMT+12:00
To someone with a hammer and a screwdriver, every problem looks
like a nail with threads.