Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!cs.utexas.edu!uunet!mcvax!ukc!dcl-cs!aber-cs!pcg
From: pcg@aber-cs.UUCP (Piercarlo Grandi)
Newsgroups: comp.arch
Subject: Re: 80486 vs. 68040 code size
Summary: Compare .o files...
Message-ID: <902@aber-cs.UUCP>
Date: 2 May 89 21:45:11 GMT
Reply-To: pcg@cs.aber.ac.uk (Piercarlo Grandi)
Distribution: eunet,world
Organization: Dept of CS, UCW Aberystwyth
	(Disclaimer: my statements are purely personal)
Lines: 33

In article <13718@steinmetz.ge.com> davidsen@crdos1.UUCP (bill davidsen) writes:
    
      This totally mystifies me. Did someone bring in some MS-DOS figures?
    My figures were clearly for Xenix, and I clearly stated the versions and
    stuff (I think the info was quoted and appears above).

But the comparisons you made may still be apple vs. oranges... I have the
strong suspicion that you compared executable sizes, which tipically include
libraries that are very different. For example, stdio with/without printf is
very different between SunOS and System5/386. Also, you have shared
libraries in some systems (SunOS 4, Unix 5.3), and not in others. There are
many other factors that influence executable size; e.g. SunOS uses 8k
buffers for stdin/stdout, etc...

The comparison ought to be made between the .o files of fairly large
modules.  As to this, my impression is that 386 objects are a bit smaller
than 68020 objects, thanks to many common 386 instructions being short (as
somebody correctly remarked about unorthogonality). I would venture to say
(base on a very small sample) that 386 code is about 5-10% smaller than
68020 code. Both compilers used are highly optimizing.

Byt the way, both the 386 and the 68020 (which are supposedly very similar
to the 486 and 68040) use linear 32 bit addresses (virtually nobody uses the
48 bit pointer large model of the 386), so this is not a factor.

The 68020 has more registers than the 386, but this ought to count for
little, as according to literature the knee of the curve for C is before or
at 4 registers (important note: for typical Unix utilities, e.g. excluding
floating point) because of the average simplicity of C expressions.
-- 
Piercarlo "Peter" Grandi           | ARPA: pcg%cs.aber.ac.uk@nsfnet-relay.ac.uk
Dept of CS, UCW Aberystwyth        | UUCP: ...!mcvax!ukc!aber-cs!pcg
Penglais, Aberystwyth SY23 3BZ, UK | INET: pcg@cs.aber.ac.uk