[Shootout-list] demoting marginal languages
Brent Fulgham
bfulg@pacbell.net
Tue, 28 Sep 2004 22:36:22 -0700
On 2004-09-28 03:33:23 -0700 Bengt Kleberg
<bengt.kleberg@ericsson.com> wrote:
> 1 we have yet to decide if a good performance means a low score or a
> high
> score. or has that already been done?
Good performance means "low use of resources": Shorter CPU cycles,
less memory,
shorter lines-of-code is better.
However, these data are then handled per me earlier e-mail, which
places "low use" scores
at the high-end of the range (so they bubble up to the top)
Therefore, a high score means a "more efficient" or "better"
implementation.
> 2 yes, i did intend for the missing test to be awarded maximum bad
> score,
> giving this language a loss to a language with complete tests. this
> would
> save us the trouble of having to invent a separate 'Provisonal'
> category.
Right now, it gets a score of zero if it is not implemented, or if the
program fails to complete
in the alloted time (currently 5 minutes of CPU time).
> in a way it is. we could have one test for all string handling. that
> test
> would then be subject to debate about how much of it that should be
> set
> aside for the different string handling primitives that exist.
> do we want
> 10% execution time for string add
> 20% execution time for string search
> 20% execution time for string regexp
> perhaps string reverse search should be counted in string search?
> etc? hours
> of endless fun awaits :-)
Ugh! Can't we just have simple tests that do one thing? :-)
> the good thing with the very more difficult to administrate approach
> of
> having several string tests is that we could then weight the various
> results
> much more easily.
yes -- plus for we lazy people, we don't have to change anything...
-Brent