Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!sdd.hp.com!ucsd!rutgers!mcnc!rti!dg-rtp!bigben!beeblebrox!philip
From: philip@beeblebrox.dle.dg.com (Philip Gladstone)
Newsgroups: comp.arch
Subject: Re: Looking for a really odd computer
Message-ID: <1990Oct4.201721.3392@dle.dg.com>
Date: 4 Oct 90 20:17:21 GMT
References: <2721@crdos1.crd.ge.COM>
Sender: usenet@dle.dg.com (Net News)
Organization: Data General, Development Lab Europe
Lines: 37

In article <2721@crdos1.crd.ge.COM> davidsen@crdos1.crd.ge.com (bill davidsen) writes:
>
>  There has been discussion of computer word size, does the number of
>bits have to be a power of two for new systems, etc. I was looking at a
>discussion in another group and saw a really nice way to solve the
>problem, but rejected it because it wasn't portable to any system. The
>question is, has anyone ever made a general purpose computer with and
>odd word size? No one doesn't count, thank you bit slicers.

I once owned a machine with a 7-bit byte size and it tended to use two
bytes to make a 14-bit word. It was a Univac buffer processor built in the mid
1960s. I got it cheap from British European Airways who used it (and several
like it) to front end their booking system. 

I never suceeded in doing much with it as I couldn't work out 
(a) how to do subroutine calls, and 
(b) how to do input. 

Getting the code in was no problem as it had a REALLY impressive
set of neon lights and push buttons. Each light was slaved to a bit in a 
register and pressing the button would force the bit to a 1 EVEN if the machine
was running. Thus you set up an auto-incrementing store as the current 
instruction, set the address register to where you wanted the code to go,
and then just filled in the accumulator with the 7-bits you wanted
to store; pressing the microcycle step button would then execute the store
and away you went.

Still, it did have a very fast interrupt context switch (0 microsecs), but
the penalty it paid for that was a very slow everything else. The basic
instruction rate was one per 28us (four memory cycles). Anything interesting
took five or six cycles (42us). My workstation runs at 1 instruction per
40ns -- I suppose that is progess of sorts!

Philip Gladstone                        philip@dle.dg.com
Development Lab Europe                  C=gb/AD=gold 400/PR=dgc/O=dle
Data General, Cambridge                   /SN=gladstone/GN=philip
England.  +44 223-67600