Path: utzoo!mnetor!uunet!husc6!hao!gatech!udel!burdvax!sdcrdcf!trwrb!felix!ccicpg!harald
From: harald@ccicpg.UUCP ( Harald Milne)
Newsgroups: comp.sys.amiga
Subject: Re: The REAL problem is the nature of personal computers.
Message-ID: <9011@ccicpg.UUCP>
Date: 12 Jan 88 11:37:58 GMT
References: <7967@g.ms.uky.edu> <1363@sugar.UUCP> <8692@ccicpg.UUCP> <38467@sun.uucp>
Organization: CCI CPD, Irvine CA
Lines: 64

In article <38467@sun.uucp>, cmcmanis%pepper@Sun.COM (Chuck McManis) writes:
> >	In summary, I don't feel UNIX is reasonable for the home computing
> >market in general. It's just too expensive to implement. And its overkill
> >for a machine that's supposed to be a personal computer, not an 80 user
> >supermicro UNIX system.
> 

	I knew I should have qualified that statement with price. In terms
of "reasonable for the home computing market" I was thinking around $1000 to
$2000, which I think is about the upper limit for the computer buying public
at large for now. When I told my friends about the A2000 prices, they
scrambled for the A500 en masse. Well some. But this is simply a marketing/
affordability issue.

> Well, all I can say is don't limit yourself arbitrarily. It's coming no 
> matter what you or I say about it. :-)

	Believe me, I'm not! I'm very encouraged by the things I see happening.
For instance, the MMU on the 68020 board CBM showed at COMDEX. I doubt
seriously it was intended for some future version of AmigaDos, and if it 
appears this way in a consumer version, it certainly opens the door to UNIX.

	Add to this, CSA's "Over 030" board. It also sports an MMU. The ad
in Amiga/World even went as far to say "capability for running UNIX (when
available)". When, not if! Perhaps Im getting a bit too enthusiastic, but
thats the way I am. Damm the Amiga! 8^)

	AmiExpo is 3 days away..tick..tick.. CSA is gonna see a lot of me!

	The 68030 has a "cache extensible architecture". If done properly,
the 68030 can run at full speed internally. It could SCREAM like HELL. I
have done numerous cache size studies for our upcomming RISC computer, and
it is amazing what even a minimal cache (say 4-8k) can do with REAL slow
memory (relatively speaking of course). The thing that has me bummed out, is
cache integrity has to be maintened somehow. But after giving this more
thought, I came to the conclusion that the Amiga "knows" at the OS level
when it is doing DMA. The Amiga could appropriately purge/maintain cache
integrity, BUT, is has to be built into the OS. (Hint, hint 8^)). Why not?
I know Im an 030, and I know Im bad! (Ack, bad Jackson pun.)

	If this was done, you wouldn't need expensive 32 bit memory which
does not exist in the entire Amiga address space. You would need REAL
expensive cache memory, BUT you would not need as much, (since it's cached)
and you would be compatible with 16 bit memory. The performance gained this
way is phenomanel, and with a large enough cache (say 32-64k) memory access
goes down to almost nil. (Well almost) Of course, there is a down side to
this, which is how long does it take to purge a cache this size (yanking a
hardware line would be wonderful!) takes in code and how often you have to
do it (like every blit, DMA, etc). But this scheme still has enough benifits
to warrent investigation.

	Somewhere along the line, this strategy becomes viable/unworkable.
And I would like to put more effort into this, but then I don't work for
CBM.  And believe it or not, I don't know enough to talk about this, without
some kind of feedback that Im totally flashed off the deep end. Oh well.

	Jeez, I'm rambling again and it's 4:00 in the morning.

	See you all at AmiExpo! I'll be there Sunday for sure.

-- 
Work: Computer Consoles Inc. (CCI), Advanced Development Group (ADG)
      Irvine, CA (RISCy business! Home of the CCI POWER 6/32)
UUCP: uunet!ccicpg!harald