Path: utzoo!dciem!chk
From: chk@dciem.dciem.dnd.ca (C. Harald Koch)
Newsgroups: comp.sys.amiga
Subject: Re: AmigaDOS scheduling
Message-ID: <1751@dciem.dciem.dnd.ca>
Date: 23 May 89 17:39:07 GMT
References: <17148@usc.edu> <24279@agate.BERKELEY.EDU> <18268@cup.portal.com> <17183@usc.edu> <21814@srcsip.UUCP> <18037@cisunx.UUCP> <1861@internal.Apple.COM> <761@boing.UUCP> <216@doctor.Tymnet.COM> <6560@ardent.UUCP>
Reply-To: chk@dretor.dciem.dnd.ca (C. Harald Koch)
Organization: NTT Systems, Inc., Toronto, Canada
Lines: 22

In article <6560@ardent.UUCP> rap@rap.ardent.com (Rob Peck) writes:
[ description of a process that changes priorities of other processes in
  the system deleted. ]

The Amiga Exec calls itself through its own jump table in a lot of places.
In particlular it indirects via -$24(ExecBase) whenever an interrupt occurs,
to check if task rescheduling is necessary.

It seems to me that it would be possible (and relatively easy!) to replace
this jump vector (and maybe a couple of other entry points that massage the
task queues) to install any arbitrary scheduling algorithm.

The only problem I can see with this is that there are tasks out there that
assume the current scheduling algorithm. Some systems that I have come
across assume that a really high priority task will get all the CPU whenever
it wants it. This task then avoids locking certain structures because it
knows that nobody else has access while it runs. Of course, this is a bad
practice, and is rare, so it shouldn't be a major problem.

-- 
C. Harald Koch		NTT Systems, Inc., Toronto, Ontario
chk@gpu.utcs.utoronto.ca, chk@dretor.dciem.dnd.ca, chk@chkent.UUCP