Path: utzoo!censor!geac!torsqnt!lethe!yunexus!ists!helios.physics.utoronto.ca!news-server.csri.toronto.edu!bonnie.concordia.ca!uunet!zaphod.mps.ohio-state.edu!usc!apple!heksterb
From: heksterb@Apple.COM (Ben Hekster)
Newsgroups: comp.sys.mac.system
Subject: Re: Preemptive scheduling
Summary: Is preemptive multitasking sufficient for intensive interaction?
Message-ID: <48172@apple.Apple.COM>
Date: 16 Jan 91 20:17:25 GMT
References: <19019@shlump.nac.dec.com> <48122@apple.Apple.COM> <1991Jan16.035715.4711@maverick.ksu.ksu.edu>
Organization: Apple Computer Inc., Cupertino, CA
Lines: 62

I wrote:

>  Have you ever
> experienced mouse operations in preemptive multitasking environments?  My
> experience with X Window on high-performance workstations (various window
> managers), for instance has been that feedback is frequently so *awful*
> (intermittent, or no reaction to mouse movements at all for periods on the
> order of many seconds) as to make complicated interactions extremely
> difficult.  (No flame on X intended, just an observation on the workstation
> environment)

jxf@altair.cis.ksu.edu (Jerry Frain) responds:

> I always find this argument quite humorous, since I do all of my work
> at work on a Sun workstation using X, while I do all of my work at home
> on a Mac SE/30.
> 
> I really have no problems with mouse/pointer movement on the Sun
> workstation I use at work.  If there are fewer than ninety existing
> processes, and four or more actually wanting to do something at that
> time, then I may start to feel it.  I just renice my X server, and
> everything is fine until the load climbs up quite a bit more.

I find it surprising that you find my observation humorous.  Being the sole
interactive user on an HP 9000/400 (I think?) running X, and hpterm to emulate
a non-graphic VT100-type terminal, response was often absolutely unbearable.
Using this very terminal program from a mainframe VAX (no graphics involved,
except on the Mac end) to write this article, I have feedback problems when
scrolling, for instance.  It may just be a question of throwing more raw
processing power at it --which is what workstations are good at-- but I don't
think the answer to the multitasking question is to make everyone buy IIfx-es.

> Shove a floppy in your disk drive on your Mac and move your mouse around
> and tell me how your Mac is *so* responsive (IIfx owners may be exceptions
> to this test).

I'm not sure I understand what point you're trying to make here.  Presumably
the assumption is that when a user inserts a disk, he wants to mount the
volume, and at that time doesn't really care for his Mandelbrot background
task to run for another 1000 iterations.  Since the disk-mounting is alredy
running at the highest priority possible (nothing is preempting it) I'm not
sure I understand what good a preemptive scheduler would do here.

	The underlying premise is that getting the disk mounted as quickly as
possible is more important than moving the mouse.

> My point is, a preemptive scheduler on a single-user system is really
> not a problem -- at least no more than cooperative, especially when an
> intelligent scheme is used to assign priorities properly.

It may or may not be a *problem* (compatibility issues have already been
raised)--but that does not signify whether it is *desirable*.

	As an aside--how many of those ninety processes were running to
support the multitasking?

_______________________________________________________________________________
Ben Hekster                           | "I've had my fun
Student intern                        |  but now it's time
AppleLink:   heksterb                 |  to serve your conscience
Internet:    heksterb@apple.com       |  overseas"
BITNET:      heksterb@henut5          | --Orange Crush, R.E.M. (Green)