Path: utzoo!mnetor!uunet!lll-winken!lll-lcc!ames!pasteur!agate!eris!doug
From: doug@eris (Doug Merritt)
Newsgroups: comp.sys.amiga
Subject: Re: Amiga UNIX
Message-ID: <9031@agate.BERKELEY.EDU>
Date: 21 Apr 88 08:26:20 GMT
References: <211@laic.UUCP> <3663@cbmvax.UUCP>
Sender: usenet@agate.BERKELEY.EDU
Reply-To: doug@eris.UUCP (Doug Merritt)
Organization: University of California, Berkeley
Lines: 37
Summary: Unfair comparison of 68K to 286

In article <3663@cbmvax.UUCP> daveh@cbmvax.UUCP (Dave Haynie) writes:
>While the 80286 machines that Xenix run on are pretty brain damaged from
                                                       ^^^^^^^^^^^^^
(definitely agree!)

>an architectural point of view, their segmentation scheme works as well
>as paging to make each process appear to run at the same location in 
>memory.  So you can just drop Xenix into an AT[Clone] and go.  The plain
>old 68000 doesn't support anything like that.

That's not quite accurate, and is in fact quite unfair to the 68000
family, which has address modes that are more general than the
80x86 family. The 68000 can be programmed to use a pure segmentation
scheme, using "register indirect with offset" and "indexed register
indirect with offset". For that matter, you can create 68K code that
is purely relocatable.

The difference that you're thinking of is that the 286 has a simple
MMU built into it to support their crumby segmentation scheme, while
the 8086 had none. The result *will* run Xenix, but due to the
addition of MMU hardware, not from some advantage to segmentation.

Quite the contrary! Having worked *extensively* with Xenix on the 286,
and with Unix on a 68000 (with a custom MMU), I can tell you definitively
that the 68000 is a *much* more pleasant architecture. As an example,
try porting any Unix program that has arrays larger than 64K. It's
trivial on a 68K, but requires a rewrite from the ground up, in general,
for the 286. And yes, this did come up often...it seems that once people
start writing for nonsegmented architechtures (e.g. VAX), they start
assuming that large arrays are ok. How silly. Everyone knows that 64K
is a reasonable absolute limit on the size of data structures! :-)

The 286 is newer than the 68000; a fair comparison is against the
68010 with the 68851 or something. In which case Unix works better
on the 68010 than Xenix on the 286. By far.
	Doug Merritt		doug@mica.berkeley.edu (ucbvax!mica!doug)
			or	ucbvax!unisoft!certes!doug
			or	sun.com!cup.portal.com!doug-merritt