[Shootout-list] Re: Proposal: revised matrix multiplication test
Aaron Denney
wnoise@ofb.net
Thu, 4 Nov 2004 01:12:31 +0000 (UTC)
On 2004-11-03, Isaac Gouy <igouy2@yahoo.com> wrote:
>> Rather than this, I would prefer that all the work actually be
>> necessary.
>
> imo a better way to start addressing that issue is to replace tests
> which are worse than matrix - such as hash.
>
> Doug Bagley thought it was a bad test, and everyone since has agreed.
> Can we devise a replacement 'lookup' test which allows acceptable
> solutions in both functional and imperative languages?
Feel free to come up with a better replacement for that. I've
considered it, but have had a tough time seeing a way of testing that
that I like.
>> Alternatives:
>> It's not strictly necessary that we read in the matrix.
>
> Unless we really want IO to be a big part of the test
> shouldn't we avoid having any IO in a test?
A minimal amount shouldn't hurt our resolving power. If I can
come up with a good easily computable set of matrices, I'll suggest it.
(Hmm. Properly scaled Walsh-Hadamard matrices could be the right
thing.)
--
Aaron Denney
-><-