home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!europa.asd.contel.com!darwin.sura.net!mips!swrinde!elroy.jpl.nasa.gov!ames!riacs!pioneer.arc.nasa.gov!lamaster
- From: lamaster@pioneer.arc.nasa.gov (Hugh LaMaster)
- Newsgroups: comp.benchmarks
- Subject: Re: Geometric Mean or Median
- Message-ID: <1992Aug14.155857.6561@riacs.edu>
- Date: 14 Aug 92 15:58:57 GMT
- References: <PRENER.92Aug9220648@prener.watson.ibm.com> <1992Aug12.012620.3441@nosc.mil> <1992Aug12.172209.3108@nas.nasa.gov> <Aug14.142126.38458@yuma.ACNS.ColoState.EDU>
- Sender: news@riacs.edu
- Distribution: comp.benchmarks
- Organization: RIACS, NASA Ames Research Center
- Lines: 38
-
- In article <Aug14.142126.38458@yuma.ACNS.ColoState.EDU>, shafer@CS.ColoState.EDU (spencer shafer) writes:
- |>
- |>
- |> Having joined this in midstream, I risk repeating some information. If
- |> so, my apologies in advance. A discussion of this, and an offered proof
- |> of the geometric mean as preferred method is in the March 1986 issue of
- |> Communications of the ACM, "How Not to Lie With Statistics: The Correct
- |> Way to Summarize Benchmark Results," by Fleming and Wallace.
-
-
-
- Yes, and there was a rebuttal to this "proof" in CACM by, I believe,
- J.E. Smith, in October of 1988. {If I have the reference correct,}
- it is proved that the harmonic mean is the correct measure of rates,
- if you want to examine a fixed workload and characterize the performance
- on that workload. Fleming and Wallace missed the point, IMHO, of a lot
- of things, including Amdahl's law. Always use the harmonic mean of rates
- unless you have a specific reason not to.
-
-
-
- I was glad to see that Gordon Bell provided the "LFK(hm)" column in the
- performance summary of his recent article in CACM, since this is the
- harmonic mean of an untuned workload of Livermore Loops. This is the one
- of the many possible measures, which I find most meaningful when comparing
- raw floating point performance. It tends to be pessimistic with respect
- to tuned rates on vector machines by perhaps a factor of 3 or 4, but it
- is rather realistic when looking at untuned rates.
-
- ("raw" -- that is, a first cut, before you do some "real" benchmarking :-)
-
-
-
- --
- Hugh LaMaster, M/S 233-9, UUCP: ames!lamaster
- NASA Ames Research Center Internet: lamaster@ames.arc.nasa.gov
- Moffett Field, CA 94035-1000 Or: lamaster@george.arc.nasa.gov
- Phone: 415/604-1056 #include <usenet/std_disclaimer.h>
-