Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!apple!agate!darkstar!ce.ucsc.edu
From: karplus@ce.ucsc.edu (Kevin Karplus)
Newsgroups: comp.benchmarks
Subject: trivial benchmark (not again!)
Message-ID: <9724@darkstar.ucsc.edu>
Date: 4 Dec 90 23:51:12 GMT
Sender: usenet@darkstar.ucsc.edu
Reply-To: karplus@ce.ucsc.edu (Kevin Karplus)
Organization: University of California, Santa Cruz
Lines: 17

I was the one who got the ridiculously poor time for the 
	echo 2^5000/2^5000 | /bin/time bc 
on a Sparcstation SLC.  The time I got is not representative of other
SLC's, but it was accurate for this one.  I even repeated the
benchmark on different days, and without suntools running.  I still
don't know the reason for the bad performance.  We tried booting
another SLC in place of this one, to check out all the software
possibilities---it ran the trivial benchmark at a perfectly reasonable
speed.  Now, mysteriously, my SLC runs at the right speed also:
       15.7 real        15.1 user         0.1 sys  
The systems people are all away at a conference, and so I don't know if
they fixed anything, or if simply rebooting caused the problem to go away.

Moral: Never use a trivial benchmark as an indicator of system speed.
Moral2: Even trivial benchmarks may be useful for pointing out problems.

Kevin Karplus


Brought to you by Super Global Mega Corp .com