Path: utzoo!mnetor!uunet!husc6!bloom-beacon!tut.cis.ohio-state.edu!osu-cis!apr!dss
From: dss@apr.UUCP (Daniel Sabo)
Newsgroups: comp.sys.atari.st
Subject: Re: (a) The Right to Flame & (b) WHAT ATARI NEEDS TO DO...
Message-ID: <429@apr.UUCP>
Date: 29 Apr 88 13:58:18 GMT
References: <203@eutrc3.UUCP> <1615@alliant.Alliant.COM> <1043@atari.UUCP> <303@bdt.UUCP> <27969@yale-celray.yale.UUCP> <585@galaxy>
Reply-To: dss@apr.UUCP (Daniel Sabo)
Organization: APR, Columbus, OH
Lines: 77

In article <585@galaxy> arcarese@andromeda.UUCP (John C Arcarese) writes:
>>Atari     - The Abaq w/o MMU and UNIX ? Har Har. Thats gonna be
>>            the Atari videogame for the 90's. The hardware
>>            upto the MMU is the right thing, but then... Prediction:
>>            Development discontinued... a workstation company buys
>>            INMOS. The Transputer will become a Megahit in the work-
>>            station business.  The 68030 ? Quite possibly. Not that I 
>>            am particularly fond of yet another featureless Shivi-creation,
>>            but UNIX on top of it and I might be buying.
>
>	I think the ABAQ might be a joke if you're looking for market
>	acceptance, but as far as hardware goes, it looks quite nice.
>	Knowing Atari, they would release it, with an incompatible
>	OS(meaning incompatible w/ UNIX), then everyone would have to wait
>        3yrs. for software to meet the users expectations.  I would like
>        to see Atari (and C=) use the 88000 in future designs.  That way they 
>        could get transputer speed, with compatibility.  If the transputer is
>        used for anything, I think it'll be used as a graphics controller.
>					John A.

Hi folks, I've been reading this group for a while now and can't hold
back my humble reflections any longer.  I really like my ST and a lot
of other machines that don't have segmented architectures.  

I really like the prospect of Atari in the workstation market with
a Un*x box ('030).  It can only help.  The Transputer machine (any
transputer machine) is an entirely different animal.  The 88000 and
RISC and SPARC and '436 machines are all rather clever red herrings.
The point isn't reduced instruction sets, memory management or even
MIPS per processor.  The real issue is parallelism, virtually unlimited
parallelism; and the transputer has it.

The most significant challenge to the ABAQ project is to come up with a useable
operating system which can effectively take advantage of the parallel 
processors. The base hardware already exists, and works!  The OS
(Helios), will really be a challenge.  No matter who does it (especially
someone like Atari) I'm certain it won't be 'quite right' on the first
try.  Even then I'll give it my enthusiastic support.  I believe
the present state of parallel processors is analogous to the state of
Random Access Memory in the mid 1950's!!  And I expect to see advances
in the computer industry of similar magnitude BECAUSE of parallel
processing.

I have to defend the decision to make Helios 'look-like' Unix. It will
inspire the same sort of tools, and programmers like it.  After listening
to Mr. Beckmeyer you have to agree how important that is :-).
It's just too bad that this makes people want it to BE unix.  Unix just
can't (in it's present, SVID, form) support parallelism at all levels.

I feel bad about chastising the people who site the lack of an MMU as a
liability.  They seem the closest to understanding the things that are
really important!  'tis true, without an MMU 'virtual memory' is not
going to be feasible. The ABAQ is suppose to implement a greater goal,
_Virtual_Processors_.  Each task doesn't get just its own memory to muck
about in, without worry; but its own processor (and anything on that
processor) to muck with.

Imagine the price/performance curve of present day computers. Plotting
performance vs incremental cost. At the left end are the older 8 bits,
followed by clones, ST's and other micro's. Then moving up through
mainframes in a sharp curve to present day Crays. Each additional MIPS
costs more to squeak out. 

With the transputer's version of parallelism, the ABAQ would jump onto
the curve at around the same place as a MAC II or very fast '386 clone,
and then continue in a line to the right... flat... continuing flat out
BEYOND the Crays. Need just 4 more MIPS's, add one more $800.00
processor.  Need 400 more MIPS, add 100 more $800.00 processors.  For a
total of $80,000.00, flat...

Of course this assumes that Helios is up to the task, the transputer has
been proven to work this way with OCCAM.

Well I've gone on much too long for a first post ...  thanks for
listening.

	dan