RE: Gabriel Benchmarks?

[Qobi@ZERMATT.LCS.MIT.EDU (Jeffrey Mark Siskind)]

    Date: Fri, 16 Feb 90 10:40 PST
    From: snicoud@atc.boeing.com (Stephen Nicoud)

        Date: Fri, 16 Feb 90 10:55 EST
        From: davel@whutt.att.com.ARPANET

        I know this must have been asked 10,000 times, but I need a set of
        common-lisp benchmarks.  I'd appreciate pointers to actual source code
        (or the code itself), and any results you have.

        If there is interest, I'll post the results of running the tests on
        several Symbolics, TI Ex II, and HP/Apollo DN10000.

        David Loewenstern
        AT&T Bell Laboratories
        14B-253 
        Whippany, NJ 07981
        email: davel@whutt.att.com || whutt!davel
        at&t: 201-386-6516

    You might want to check out the /pub/89-02/ directory on
    sumex-aim.stanford.edu [36.44.0.6].  Here's the README file from that
    directory:

        This directory contains the source code for the benchmark programs
        used to gather data for "Performance of Two Common Lisp Programs on
        Several Systems", Stanford University Knowledge Systems Lab Report
        number 89-02.  The file KSL-TR-89-02.Hqx is a version of the MicroSoft
        word source for the document encoded with BinHex4.0.  The paper
        contains information on running the tests.

        Send comments to acuff@SUMEX-AIM.Stanford.EDU

I also have a medium sized (approx 35 page) CommonLisp program which I
have used as a benchmark. I prefer it to the Gabriel benchmarks since
it is medium sized rather than tiny and thus somewhat more reflective
of reality. Also, it is a real program (a snapshop from six months ago
of a research system that I am constructing) so rather than testing
particular nuiances of a compiler, it accurately reflects my particular
programming style which mixes functional programming with object-oriented
programming. The program is a language acquisition simulator which
figures out the meanings of words given simulated linguistic and visual
input. I have done the necessary modifications to port this program to
Lucid, Franz, AKCL, Symbolics, and Harlequin. The program has no
#+ or #- conditionals, is pure CommonLisp (+LOOP macro) and has no
declarations. I took my Symbolics program and did the minimal modification
needed to get it to run jointly on all 5 systems. I have documented those
modifications as comments at the beginning of the program. I have run
the program on the above compilers on numerous platforms including
3630, XL400, NeXT, SparcStation330, SparcStation1, HP????? with various
combinations of proclaimations of OPTIMIZE.

I will give the program to anyone who wants it. I will also give the
results of my benchmark to anyone who wants it. Be warned though that
I have *NO DOCUMENTATION* nor do I have time to create any. I have two
files: benchmark.lisp containing the program and benchmark.text containing
the results. I someone can tell me of a public place where I can post
these two files I will do so.
        Jeff