Date: Thu, 20 May 1999 07:17:52 -0400 (EDT)
From: JAmes Atwill <james AT tainted DOT org>
To: pgcc AT delorie DOT com
Subject: Re: Benchmark PGCC vs EGCS on a K6-2
In-Reply-To: <3743ADE8.C938ADBB@informatik.hu-berlin.de>
Message-ID: <Pine.NEB.3.96.990520070839.20661O-100000@slavery.tainted.org>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Content-Transfer-Encoding: 8bit
X-MIME-Autoconverted: from QUOTED-PRINTABLE to 8bit by delorie.com id IAA08302
Reply-To: pgcc AT delorie DOT com
X-Mailing-List: pgcc AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

On Thu, 20 May 1999, Jens-Uwe Rumstich wrote:

> I still donīt know, why my results are that wrong :-(

Running benchmarks on a living breathing system generally isn't advised.
If you're running this on your own personal computer at home, some tips
(from experience).

o Quit X.  If you can't, kill off, or suspend as many of the X processes
you've got as you can (other xterms, xpdf, xemacs, etc..)
o Stop crond.   /etc/rc.d/init.d/crond stop
o Stop atd.     /etc/rc.d/init.d/atd stop
o Stop inetd.   /etc/rc.d/init.d/inet stop
o Kill off any other servers which make your system accessable from the
  outside (apache, postfix/qmail/sendmail, innd, bind, dhcpd, etc..)
o sync your disks (type "sync")
o Don't play mp3's (hey, i only did it once!)
o Check with "top" and make sure that nothing that could suddenly pipe up
  is running.  
o Try to lower the amount of I/O that a test does (unless testing I/O
  speeds).
o Run your test 3-4 times ("% time ./mytest") to get a feel for how long
  it takes and to negate (if applicable) the 2.2.x caching system. 
o Run the actual test a couple times and average the result.

Those tips have generally shown good results for me; of course YMMV. 

	JAmes