[Shootout-list] New matrix-norm not matrixy?
Dave
davejf@frontiernet.net
Mon, 28 Mar 2005 10:36:11 -0600
----- Original Message -----
From: "Sebastien Loisel" <sloisel@gmail.com>
To: "Dave" <davejf@frontiernet.net>
Cc: <shootout-list@lists.alioth.debian.org>
Sent: Monday, March 28, 2005 3:35 AM
Subject: Re: [Shootout-list] New matrix-norm not matrixy?
>> > One of the advantages of the proposed matrix norm calculation is that
>> > it is resistant to LAPACK acceleration. So you'll have to get your
>> > hands dirty if you want to get some performance.
>>
>> given the recent discussion on testing compilers and runtimes, not 3rd
>> party
>> libs.
>
> Well, can you tell there's a "third party lib" here?
>
> http://shootout.alioth.debian.org/sandbox/benchmark.php?test=matrix&lang=python&id=0&sort=fullcpu
>
> That's LAPACK.
>
No, with a casual glance and not knowing exactly what the base
implementation of Python is, frankly I can't tell.
Maybe a better - if cumbersome - term for "3rd party libraries" would be
"extraneous libraries not included in the base implementation" or some such.
However, if common base distributions of Python, or commonly used pre-built
wrapper libraries make it that easy to use LAPACK then I'm Ok with that,
because that represents a good "real-world/general-case" use of the
language/implementation.
The challenge for at least some of the Shootout tests should be to minimize
advantages like that - just like your matrix-norm test does.
My opinion boils down to: The better a test is at benchmarking base
language/implementation facilities (and not "extraneous libraries" or
hardware), the more suitable the test is for the Shootout, or maybe at the
least those types of tests should have a higher default weight.
- Dave
>
> Sebastien Loisel
>
>
>
> --
> No virus found in this incoming message.
> Checked by AVG Anti-Virus.
> Version: 7.0.308 / Virus Database: 266.8.4 - Release Date: 3/27/2005
>
>