From owner-freebsd-hardware  Wed Mar 26 04:54:56 1997
Return-Path: <owner-hardware>
Received: (from root@localhost)
          by freefall.freebsd.org (8.8.5/8.8.5) id EAA07179
          for hardware-outgoing; Wed, 26 Mar 1997 04:54:56 -0800 (PST)
Received: from dfw-ix6.ix.netcom.com (dfw-ix6.ix.netcom.com [206.214.98.6])
          by freefall.freebsd.org (8.8.5/8.8.5) with ESMTP id EAA07174
          for <freebsd-hardware@FreeBSD.ORG>; Wed, 26 Mar 1997 04:54:52 -0800 (PST)
Received: (from smap@localhost)
          by dfw-ix6.ix.netcom.com (8.8.4/8.8.4)
	  id GAA25077; Wed, 26 Mar 1997 06:54:20 -0600 (CST)
Received: from wck-ca9-25.ix.netcom.com(204.31.231.121) by dfw-ix6.ix.netcom.com via smap (V1.3)
	id sma025069; Wed Mar 26 06:54:01 1997
Received: (from asami@localhost) by silvia.HIP.Berkeley.EDU (8.8.5/8.6.9) id EAA10369; Wed, 26 Mar 1997 04:53:58 -0800 (PST)
Date: Wed, 26 Mar 1997 04:53:58 -0800 (PST)
Message-Id: <199703261253.EAA10369@silvia.HIP.Berkeley.EDU>
To: robsch@robkaos.ruhr.de
CC: freebsd-hardware@FreeBSD.ORG
In-reply-to: <m0w9dg4-00060xC@robkaos.ruhr.de> (robsch@robkaos.ruhr.de)
Subject: Re: Slow 'worldstone' performance on a P6
From: asami@vader.cs.berkeley.edu (Satoshi Asami)
Sender: owner-hardware@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

 * 'make world'  takes 2 hours and 50 minutes on my P6 system.

What clock speed? :)

 * (P6NP5 mb with 64 MB EDO, SC-200 HA, 2 Conner 2105S hard drives,

It takes 2 hours and 10 minutes on my P6-200 (256K) system with 96MB
non-EDO RAM, Quantam Atlas I + II and 53c825.  Oh, and the system is
2.2.

 *  /usr/src and /usr/obj are on different drives, TMP on a Conner
 *  850 MB IDE drive (is this the bottleneck?))
 * The system is FreeBSD-current (last Saturday).
 * I find this quite slow. I think it should take about 100 minutes?

Not really.  It depends on your compile options too.  Mine is "-O2
-pipe", which means the compiler is taking more time optimizing (I
remember it was like 20 minutes faster with "-O") and is not writing
anything to /tmp.

Also, I don't think having /usr/src and /usr/obj on different drives
help unless they are both on small partitions and physically located
near the other places on the disk that get accessed during the build
(swap, root FS, etc.).

If your /usr/src and /usr/obj are both on big partitions that span the
entire disk (like mine), your average disk seek latency won't improve
by having two disks (unless you run parallel makes or something).  I
guess I should try putting them both on the same disk and see what
happens.

Satoshi