Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!tut.cis.ohio-state.edu!purdue!haven!adm!smoke!gwyn
From: gwyn@smoke.BRL.MIL (Doug Gwyn)
Newsgroups: comp.unix.wizards
Subject: Re: Large Unix, SysV.4 for Sperry
Message-ID: <12412@smoke.BRL.MIL>
Date: 23 Mar 90 08:11:42 GMT
References: <22847@adm.BRL.MIL> <0a=302Ns95IT01@amdahl.uts.amdahl.com>
Reply-To: gwyn@brl.arpa (Doug Gwyn)
Organization: Ballistic Research Lab (BRL), APG, MD.
Lines: 18

In article <0a=302Ns95IT01@amdahl.uts.amdahl.com> littauer@amdahl.uts.amdahl.com (Tom Littauer) writes:
>In article <22847@adm.BRL.MIL> archunix@stl-08sima.army.mil (Bernie J. Potter) writes:
>>Encore's Multimax 300's (4-40 mips) and 500's (17-170 mips).
>>Amdahl's 5890,5980, or 5990 machines, using UTS (5-100 mips).
>A word of caution here... as I'm sure you know, MIPS stands for
>Meaningles Indicator of Processor Speed.

That's for sure -- which is better, -153 mips or -95 mips?

The real reason for this note is to remark that most benchmarks
I've seen are also rather meaningless.  (Indeed, I suspect some
compiler vendors of tuning their compilers so that the "standard"
benchmarks used by the toy-computer magazines work well but any
reasonable application works horribly if at all.)  A really good
benchmark should simulate the anticipated application loading of
the system.  DA MINIS didn't do too badly in this area, actually.
BRL typically uses benchmarks based on ray-tracing code, since
that's one of our big application areas.