guile-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Introspective benchmarking.


From: Rob Browning
Subject: Introspective benchmarking.
Date: Tue, 19 Mar 2002 21:49:45 -0600
User-agent: Gnus/5.090006 (Oort Gnus v0.06) Emacs/21.1 (i386-debian-linux-gnu)

I feel like it would be extremely helpful to have a set of benchmarks
that we can run in order to help evaluate our changes.  While I'm well
aware of the risks involved in benchmarking, I think even a moderately
good set could provide a healthy sanity check of our efforts.

What I'm wondering is how the benchmarks should be handled.  For
example, I feel like they probably shouldn't be part of the main
(core) CVS tree, if for no other reason than not having to worry as
much about about copyright assignments.  There's also the argument
that they don't belong with any one tree since we should probably be
using them to test across guile versions.

I don't know how many people have ever looked, but stalin has a fairly
large set of benchmarks in ./benchmarks.  I've played around a bit and
found so far that many of the files will run just fine as-is, though
if we want to be able to run the tests in a reasonable amount of time
we'll need to lower a lot of the final looping constants by a couple
of orders of magnitude or so :/

Stalin's source tree can actually run the set of benchmarks across a
set of scheme implementations: bigloo, stalin, scheme->c, gambit-c,
and chez, but it takes about two days on a 450MHz Xeon...

Thoughts?

-- 
Rob Browning
rlb @defaultvalue.org, @linuxdevel.com, and @debian.org
Previously @cs.utexas.edu
GPG=1C58 8B2C FB5E 3F64 EA5C  64AE 78FE E5FE F0CB A0AD



reply via email to

[Prev in Thread] Current Thread [Next in Thread]