From: Randal Whittle (rwhittle_at_usa.net)
Date: Sun Jan 04 1998 - 23:09:38 EST
At 06:42 AM 1/4/98 EST, Robert Dewar wrote:
>The Norton SI benchnmark is a notorious piece of junk. It relies heavily
>on the timing for integer multiply and divide, which are not at all
>indicative of general performance. It was written way back when Peter
>really did not know much about architecture. It served to measure
>relative rates for the same chip, but across chips it is seriously
>misleading.
>
>There were also alignment problems in at least one version of SI (Peter
>did not understand that alignment of code is critical in benchnmarks),
>but I think that has been corrected.
Without commenting on the quality of the Norton SI benchmark (I only
included it because it was quick & simple, in addition to the much more
involved and detailed benchmarks I ran), it should be noted that Peter
Norton hasn't been involved in any of the Symantec products bearing his
name for *many* years.
I'm sure that most flaws from his earlier work have been dealt with and
the above criticism is worthy only as a historical note.
-------
Randal J. Whittle whittle_at_usc.edu (213) 740-7775
Director, Electronic Commerce Program
Marshall School of Business, University of Southern California
This archive was generated by hypermail 2.1.3 : Thu Jan 23 2003 - 09:53:52 EST