Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 exptools 1/6/84; site ihnss.UUCP
Path: utzoo!linus!decvax!harpo!ihnp4!ihnss!knudsen
From: knudsen@ihnss.UUCP
Newsgroups: net.micro
Subject: The BEST measure of micro word-size
Message-ID: <1926@ihnss.UUCP>
Date: Wed, 15-Feb-84 20:22:59 EST
Article-I.D.: ihnss.1926
Posted: Wed Feb 15 20:22:59 1984
Date-Received: Thu, 16-Feb-84 05:46:55 EST
Organization: AT&T Bell Labs, Naperville, IL
Lines: 20

Been a lot of discussion here about honest & correct ways to determine
a given micro-proc's "true" wordsize.  I submit that the MOST IMPORTANT
measure is: (the envelope please):
	The largest wordsize that the Programmer can THINK in while using
this micro.
	This is directly related to the registers and the ALU, but *not*
to the external pinout.  For example, the 8086 lets me think in 16-bit terms;
I can load, add, and store (even multiply) 16-bit integers.  The 8088
appears ientical to the programmer, despite its half-sized data bus,
so still deserves to be called a 16-bit machine.
	Note that allowing a programmer to think in N-bit words means much
faster and shorter object code than on an N/2-bit machine.  Even tho the
8088 uses extra micro- (and bus) cycles to add a 16-bit word, it beats
hell out of a Z80 executing several instructions to do double-precision.
So it isn't just programmer convenience I'm thinking of.
	By this criterion, the 68000 is a 32-bit micro, until you go to
multiply.  The 6809 is a 16-bit machine for simple arithmetic, but only
8-bit for logical and multiply work.  With their floating-point attachments,
the 8086/8 and some PDP-11s are 64-bit, since you can treat floats as
single entities, and let microcode (ie, hardware) do the work.  mike k