Path: utzoo!attcan!uunet!wuarchive!emory!gatech!prism!dali!ken
From: ken@dali.gatech.edu (Ken Seefried iii)
Newsgroups: comp.sys.amiga
Subject: Re: UNIX sys V4.0
Message-ID: <16004@hydra.gatech.EDU>
Date: 29 Oct 90 17:55:45 GMT
References: <9010025@hpfcso.HP.COM> <15404@cbmvax.commodore.com> <2401@sauron.Columbia.NCR.COM> <1990Oct26.230141.22552@csun.edu>
Sender: news@prism.gatech.EDU
Reply-To: ken@dali.gatech.edu (Ken Seefried iii)
Distribution: na
Organization: The House Of Fun
Lines: 31

In article <1990Oct26.230141.22552@csun.edu> bcphyagi@Twg-S5.uucp (Stephen Walton) writes:
>In article <2401@sauron.Columbia.NCR.COM> stevem@sauron.UUCP (Steve McClure) writes:
>>In article <15404@cbmvax.commodore.com> ag@cbmvax.commodore.com (Keith Gabryelski) writes:
>>>
>>>[Amiga UNIX uses as swap space] Standard 10Mb.
>>>
>>
>>I guess the 68k and 386 versions are different.  386 uses 2x memory size for
>>swap.  Won't you be limiting yourself to 10M of application space this way?
>
>I don't know the answer to the question; I do know that 386 binaries tend
>to be a whole lot bigger than 68k ones, for somewhat unclear reasons.  The
>IRAF system (Image Reduction and Analysis Facility) has a recommended minimum
>of 8 MB RAM for a Sun-3 and 16 MB for a 386i;  same source code, different
>Fortran compilers.  

First...as a general rule, swap space should be 2 and 3 times physical
memory.  For a one or two user machine, closer to 2 time; more users,
more swap.  

Second, 386 code is generally less dense because of its lousy,
non-orthoginal, special-cased-to-death instruction set.  This makes
the job of automatic code generation (i.e. compilers) exceedingly
difficult.  Interested parties are directed to several papers by
Niklaus Wirth on the influence of architecture on code generation
(sorry, I don't have the references in front of me).

--
	ken seefried iii	"A snear, a snarl, a whip that
	ken@dali.gatech.edu	 stings...these are a few of
				 my favorite things..."