Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site fritz.UUCP
Path: utzoo!watmath!clyde!cbosgd!ukma!psuvm.bitnet!psuvax1!burdvax!sdcrdcf!trwrb!scgvaxd!felix!fritz!jim
From: jim@fritz.UUCP (Jim Gilbert)
Newsgroups: net.micro
Subject: Segmented Architectures
Message-ID: <4338@fritz.UUCP>
Date: Thu, 12-Dec-85 13:35:19 EST
Article-I.D.: fritz.4338
Posted: Thu Dec 12 13:35:19 1985
Date-Received: Sat, 14-Dec-85 23:32:52 EST
Reply-To: jim@fritz.UUCP (Jim Gilbert)
Organization: FileNet Corp., Costa Mesa, CA
Lines: 45


Burroughs "large" mainframes are a segmented architecture.  Compilers on these
systems map COBOL SECTIONs, and designated collections of Algol
procedures to segments.  Their Algol compilers automatically
segment arrays out via dope vectors.   Programs using single
data objects through multiple applications views, e.g.
a FORTRAN program which EQUIVALENCEs a 5000x1 array atop a 20x250 one
begin to need to worry about the segmentation.  Most others
do not.  I am not aware of a compiler for C on these systems.
They do support COBOL, FORTRAN, Pascal, Algol, PL/I, and BASIC.

There is, unfortunately, no popular desktop implementation of this
architecture.

For what it is worth I believe that the architectural limit on the
number of bytes in a single procedure in the Burroughs Large Systems
architecture is 8192 bytes.

I am not aware of an 64K limit on individual segments however.

These magnificent machines have been much overlooked.  Elliot
Organick wrote a pleasantly instructive book about them.
They run Algol like the wind.  It would not be much of an exaggeration
to say that they ARE Algol in hardware.

In 1961 the ancestor of these machines supported symmetric multiprocessing,
segmented virtual memory, and an all high level language operating system.
A shame it took so many of these ideas so long to catch on.

I now wait to see what interesting flames this draws.  How many of
you folks have an "I Touched a B-5000" button?  How many of you even
know what the book named "The Descriptor" is, let alone have read it?
Whomever answers, "Who cares?" chooses to ignore a fair amount of
interesting architectural pioneering.

Technology advances have changed many of the economic tradeoffs around
which these systems were designed.  I, for one, believe that anyone
who wishes to speak on the pro's and con's of segmented architectures
should do their homework on computing history.  Microprocessors were
not invented at the beginning of time.

                                Jim Gilbert
                                FileNet Corporation

                                ihnp4!trwrb!felix!jim