Path: utzoo!attcan!uunet!cs.utexas.edu!tut.cis.ohio-state.edu!purdue!haven!mimsy!chris
From: chris@mimsy.umd.edu (Chris Torek)
Newsgroups: comp.arch
Subject: Re: fork, spawn, vfork: an overview
Message-ID: <25910@mimsy.umd.edu>
Date: 6 Aug 90 13:57:08 GMT
References: <920@dgis.dtic.dla.mil> <5830@titcce.cc.titech.ac.jp> <1990Aug6.010044.8638@watdragon.waterloo.edu>
Organization: U of Maryland, Dept. of Computer Science, Coll. Pk., MD 20742
Lines: 28

In article <1990Aug6.010044.8638@watdragon.waterloo.edu>
tbray@watsol.waterloo.edu (Tim Bray) asks the very reasonable question:
>How much does fork() cost? ... Quantitatively, how much does vfork()
>save?  Is the picture significantly different on a workstation which spends
>much of its time running a window system and a few technical applications?
>
>Somebody must have measured this at some time.  Or did somebody in the bsd
>gang just say "That looks inefficient. Let's fix it" and create vfork()?

I have only two data points on this myself (the vfork thing was done
before my time---I only got into this Unix business in late 1981...).
I know that the Franz Lisp group at Berkeley drove some of the changes
to the VM system between various 4BSD releases (including, I suspect,
vadvise); and I know that the VM group at Sun who did the SunOS 4.x
virtual memory took vfork out (made it a simple copy-on-write fork) and
then put it back.  Whether the latter was only temporary, I cannot say,
but surely someone from Sun should remember:  Was this for performance
or just to accomodate broken programs like csh?

Anyway, old Berkeley papers should describe the vfork savings (which
for 6+ MB Lisp jobs on a VAx-11/780 without a copy-on-write fork will
no doubt be quite impressive); but keep in mind that these will compare
`copy everything' against `copy nothing' without mentioning a middle
ground (`copy PTEs and written pages').
-- 
In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 7163)
Domain:	chris@cs.umd.edu	Path:	uunet!mimsy!chris
	(New campus phone system, active sometime soon: +1 301 405 2750)