home *** CD-ROM | disk | FTP | other *** search
- Path: ami-cg.GraySage.Edmonton.AB.CA!cg
- Newsgroups: comp.sys.amiga.misc,comp.sys.amiga.programmer
- From: cg@ami-cg.GraySage.Edmonton.AB.CA (Chris Gray)
- Subject: Re: toy operating systems, like AmigaOS
- References: <skllsf.984983.4.5@groomlake.mil> <4kplgo$o9r@canyon.sr.hp.com> <4ksu6q$haa@nadine.teleport.com> <31724980.226C@herts.ac.uk> <Dpy2GC.1nn@wiloyee.shnet.org>
- X-NewsSoftware: GRn 2.1 Feb 19, 1994
- MIME-Version: 1.0
- Content-Type: text/plain; charset=iso-8859-1
- Content-Transfer-Encoding: 8bit
- Message-ID: <cg.7ehs@ami-cg.GraySage.Edmonton.AB.CA>
- Date: Sat, 20 Apr 96 09:57:07 MST
- Organization: Not an Organization
-
- In article <Dpy2GC.1nn@wiloyee.shnet.org> chaos@wiloyee.shnet.org writes:
-
- > for the developer, lack of memory protection is a pain. but how many
- > Unix clones can handle realtime tasks as good as amiga can? Linux can't.
- > So your Toy Operating System beats many bigger OSes just because it is
- > smaller. If all OS kernels had the same feature list, we would need only
- > one kernel. But different applications need different feature lists, and
- > different kernels. and there is still use for a small single user multi
- > tasking OS without memory protection. You must not see the missing
- > feature, you must see the advantages of a simple, performant design.
- > add memory protection to Amiga OS and you remove its main advantage.
-
- Adding memory protection to an OS does not prevent it from being a real
- time OS. The messing around with MMU tables takes only a constant (small)
- amount of work - some work when a process is created, or when it allocates
- large chunks more of memory. On process switches, only a very small amount
- of effort is needed, on most CPUs. Virtual memory, where some memory
- gets paged out to disk, *can* prevent real-time performance, but only
- if you use more memory than you physically have.
-
- The problem with memory protection for the Amiga is that having it will
- break most software, and having it increases the cost of message passing,
- even after software is fixed to work with it. My guess is that the
- increased cost would vary from 10% to 100%, depending on the application.
- Things at the high cost end would be things like terminal programs, which
- use a message per character sent or received on the serial port (which
- is bad programming even now, but I image some do it). The cost to programs
- like ray-tracers, compilers, editors, etc. would be quite small. The
- average cost is sufficiently small that it would be completely unnoticed
- if combined with a switch to a PPC CPU.
-
- I *want* memory protection, both as a programmer and as a user.
-
- --
- Chris Gray cg@ami-cg.GraySage.Edmonton.AB.CA
-