Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!ncar!ames!lll-lcc!pyramid!csg
From: csg@pyramid.pyramid.com (Carl S. Gutekunst)
Newsgroups: comp.sys.tahoe
Subject: Re: what are all the tahoe machines?
Message-ID: <57196@pyramid.pyramid.com>
Date: 31 Jan 89 06:10:51 GMT
References: <23129@watmath.waterloo.edu> <127500005@hcx1>
Organization: Pyramid Technology Corp., Mountain View, CA
Lines: 23

In article <127500005@hcx1> dick@hcx1.SSD.HARRIS.COM writes:
>  The numbers I am quoting are for AIM and Neal Nelson benchmarks which
>are the property of AIM Technologies and Neal Nelson Associates
>respectively.  These results are available from AIM and NN.

Please note that the Neal Nelson benchmark is meaningless when comparing
machines with different architectures, or when comparing different com-
pilers within the same architecture. It is weighted against machines
using optimizing compilers (including strength reduction, loop unrolling,
and common subexpression elimination), register allocators, branch pre-
diction, and many other modern compiler and architectural enhancements. 

AIM, on the other hand, has proved itself to be quite meaningful, as syn-
thetic benchmarks go.

Anyone who is at all serious about benchmarking should get a copy of John
Mashey's "MIPS Performance Brief," and the MIPS benchmark suite. (Last I
knew, MIPS was giving the suite away for $200, on 9-track mag tape.) I'd
like to see every systems vendor agree to run this under John's super-
vision, and publish the results. Would settle a lot of bar bets. :-) Of
course, I suspect John has more important things to do with his time. :-)

<csg>