Path: utzoo!censor!geac!torsqnt!news-server.csri.toronto.edu!cs.utexas.edu!uunet!auspex!guy
From: guy@auspex.auspex.com (Guy Harris)
Newsgroups: comp.arch
Subject: ~8-job "knee" in response curves on Suns (was Re: IBM RS6000)
Message-ID: <5257@auspex.auspex.com>
Date: 15 Jan 91 18:45:57 GMT
References: <1991Jan10.214122.9506@news.arc.nasa.gov> <amos.663857722@shum>
Organization: Auspex Systems, Santa Clara
Lines: 19

>(This "ledge" at 8 processes is a feature of SUNOS, and appears on all
>Sun hardware).

I suspect it's more a feature of the Sun MMU which has, on many
machines, 8 contexts, one of which is used by the kernel and the other 7
of which are used for user-mode processes; if the job to which you're
switching contexts has a context in the MMU, you can switch address spaces
setting the context register, but otherwise, you have to find a free
context or steal one from another process and load up translations for
that process into that context.

Some Sun hardware has more than 16 contexts, and some doesn't have
contexts at all - the Sun-3/80 and Sun-3/4xx have 68030s and use the
on-chip MMU, not a Sun MMU, and the Sun386i uses the on-chip MMU as well.

Have you seen this "knee" on Suns without a Sun MMU, and on machines
with a Sun MMU but more than 8 contexts, have you seen the knee at 8 or
so jobs?  If not, I suspect it's MMU-related, not SunOS-related
("SunOS", not "SUNOS", please; in Sun's name, "Sun" isn't an acronym).