[Shootout-list] Keeping slower benchmarks

[Jacob Lee]

On Sun, 2005-04-17 at 22:51 -0700, Isaac Gouy wrote:
> --- Jacob Lee <artdent@freeshell.org> wrote:
> > On Sun, 2005-04-17 at 17:56 -0700, Isaac Gouy wrote:
> > > So people looking at the website aren't going to have a problem
> > > when they see multiple java implementations?
> > >=20
> > > What problem do you have when you see multiple java implementations
> > > (apart from the concern with "rules")?
> >=20
> > The disadvantage to having multiple implementations is in hurting the
> > language in the overall rankings. I grant that the overall rankings
> > are not meant to be very significant; that being said, the fastest
> > implementation should be the one that goes into the overall ranking.
>=20
> What makes you think that the fastest implementation isn't the one that
> goes into the scorecard?
>=20
> What makes you think that the fastest implementation isn't the one that
> goes into the ranking page?
>=20
> Check the Java threads-flow 'Full CPU Time'
> http://shootout.alioth.debian.org/benchmark.php?test=3Dall=E2=8C=A9=3Dj=
ava&sort=3Dfullcpu#rank
>=20
> Now count it down from the top
> http://shootout.alioth.debian.org/benchmark.php?test=3Dmessage=E2=8C=A9=
=3Dall&sort=3Dfullcpu#bench
>=20
>=20
> > If the first implementation is subpar or gives an error, a newer
> > version should replace that instead of being labeled as=20
> > implementation #2 or #3.
>=20
> FAQ
> http://shootout.alioth.debian.org/faq.php?sort=3Dfullcpu#id

Hmm, I hadn't looked at the threads-flow one. For the Java regex,
however, the implementation that gives an error is the one used in the
ranking, rather than the fixed implementation (labeled Java #2). The
same is true for the Python reverse-complement. I guess this is just a
bug, then: I'll go ahead and file a bug report unless you reply stating
otherwise. Thanks.