Return to BSD News archive
Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!news.rmit.EDU.AU!news.unimelb.EDU.AU!munnari.OZ.AU!news.ecn.uoknor.edu!news.ysu.edu!usenet.ins.cwru.edu!agate!howland.reston.ans.net!newsfeed.internetmci.com!in1.uu.net!news.artisoft.com!usenet
From: Terry Lambert <terry@lambert.org>
Newsgroups: comp.unix.bsd.freebsd.misc
Subject: Re: Symmetric Multi-Processing
Date: Mon, 29 Apr 1996 13:41:02 -0700
Organization: Me
Lines: 51
Message-ID: <3185295E.41F1F0E0@lambert.org>
References: <3180D16D.41C6@wcom.com> <31827FD0.FF6D5DF@FreeBSD.org> <31829E21.46253B04@lambert.org> <4m01n3$t88@nntpd.lkg.dec.com>
NNTP-Posting-Host: hecate.artisoft.com
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
X-Mailer: Mozilla 2.01 (X11; I; Linux 1.1.76 i486)
Michael C. Cambria wrote:
] From a user point of view, how does FreeBSD provide SMP from
] _an_ application (or does it)? I'm interested in pthreads.
] I'll settle for Draft 4, but obvioulsy like the actual
] standard interface.
In general, any SMP in an OS is visible to applications only if:
1) The applications are multithreaded
2) The OS supports kernel multithreading
3) The threads implementation is kernel-threads-aware
So FreeBSD (and Linux) do not currently support SMP scalability
at the application level, only on a per process basis. That is,
multiple processes can run on seperate CPU's.
Note that there are some *significant* disadvantages in terms
of quantum utilization in the Solaris/SVR4 thread model; the
single overriding advantage is SMP scalability. Depending on
how the application makes its calls, this break-point over
LWP (or pthreads) may not come until the 4 or 8 processor
level for any given multithreaded applications.
The best implementation model is a hybrid user/kernel cooperative
scheduler utilizing async I/O to fully consume the process
quantum on an N:M, N>M mapping of user to kernel threads. I
have only seen this model implemented in the lab.
] If pthreads is available, and makes use of SMP, in my case,
] I'll start my work RSN and wait for the SMP to meet
] everyones "production quality". I don't even have an SMP
] machine at my complete disposal yet. Will the SMP code (and
] of course, pthreads) run on a single CPU FreeBSD system?
The pthreads implementation does *not* make use of SMP because
it is a user-space threading system.
The SMP code will only detect one processor on a single
processor system when it does the APIC probe, so it will
run. It would be more efficient to replace the mutexes
with semaphores at compile time for a strictly UP system
(even though the multithreading, when complete, will more
than make up for the cache fluch overhead of the mutex
memory that semaphores don't have).
Terry Lambert
terry@lambert.org
---
Any opinions in this posting are my own and not those of my present
or previous employers.