From owner-freebsd-smp  Wed Sep  3 04:49:58 1997
Return-Path: <owner-freebsd-smp>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.7/8.8.7) id EAA07381
          for smp-outgoing; Wed, 3 Sep 1997 04:49:58 -0700 (PDT)
Received: from fang.cs.sunyit.edu (perlsta@fang.cs.sunyit.edu [192.52.220.66])
          by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id EAA07375
          for <freebsd-smp@FreeBSD.ORG>; Wed, 3 Sep 1997 04:49:53 -0700 (PDT)
Received: from localhost (perlsta@localhost) by fang.cs.sunyit.edu (8.8.5/8.7.3) with SMTP id LAA17715; Wed, 3 Sep 1997 11:49:52 GMT
Date: Wed, 3 Sep 1997 11:49:52 +0000 (GMT)
From: Alfred Perlstein <perlsta@fang.cs.sunyit.edu>
To: Antti-Pekka Liedes <apl@mail.cs.hut.fi>
cc: Steve Passe <smp@csn.net>, freebsd-smp@FreeBSD.ORG
Subject: Re: SMP performance on P5 systems?
In-Reply-To: <19970903102141.63694@hutcs.cs.hut.fi>
Message-ID: <Pine.BSF.3.95.970903114840.17682A-100000@fang.cs.sunyit.edu>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: owner-freebsd-smp@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk


What kind of drive technology are you using?
non-busmastering IDE will kill your perforance won't it?

> > > The current SMP performance on my Tyan Tomcat IIID dual pentium 200MMX
> > > system is not quite what I had expected, as a simple benchmark, this is
> > > what it takes for me to compile kernel, with UP kernel:
> > > make -j3  283.19s user 32.56s system 94% cpu 5:32.83 total
> > > and with SMP:
> > > make -j3  341.23s user 134.72s system 183% cpu 4:18.80 total
> > 
> > try -j16, using just 3 processes are leaving you diskbound.
> > 
> 
> -j12 or -j16 didn't give any practical differences:
> text    data    bss     dec     hex
> 1142784 98304   347076  1588164 183bc4
> make -j16  349.22s user 129.98s system 189% cpu 4:12.77 total
> 
> I guess it's just so heavily disk bound operation that SMP doesn't give
> much edge compared to UP.