[Shootout-list] begging questions about CRAPS

Brandon J. Van Every vanevery@indiegamedesign.com
Wed, 20 Oct 2004 00:09:56 -0700


Bengt Kleberg wrote:
>
> 2 i think the whole idea of having one score that should
> somehow reflect
> multiple results is very difficult to get right. i am not up to it.

Well, it seems a lot of people's energy has lately gone into the
concurrency and threading stuff.  At least, there were very long
discussions that I did not follow.  Maybe those people could define a
composite threading score.  Except that it seems you guys can't agree on
anything, so I guess it's going nowhere.  Or did I fall asleep before
consensus was reached?

I think it's probably up to a given 'interest group' to define a chunk
of tests that they care about and implement a proper composite score for
it.  I would probably belong to a 'CPU performance' interest group, like
how to get through low-level algorithms the fastest.  Array, list, and
hash benchmarks, that sort of thing.  I could care less about threads,
process creation, file IO, or OO.

I especially hate line counts and other nonsense about what language has
a 'better' syntax.  The best way to determine that is by looking at the
example snippets and deciding for oneself.

I'm seeing this list as an instance of the generalized 'herding cats'
problem.  I don't see any basic interest in agreeing on anything in
order to reach a goal.  Rather, I see individuals with their own agendas
who aren't going to budge from them.  Nobody has a stake in any
particular outcome, it's all just driven by people's moods and tastes.
Occasionally the interests of 2 people overlap on something.  Does that
result in any kind of cooperation?  Or do we mainly just discuss all
these things because we like the discussion for its own sake?

My own stakes in the Shootout have changed since I got on this list.  My
interest in advocating for OCaml is pretty much gone.  2 reasons for
that: (1) I realized the Shootout isn't well focused for showcasing
OCaml.  All that's really needed is a simple benchmark comparing OCaml,
C++, Java, and C#.  (2) I realized I couldn't get the OCaml community to
do any advocacy, marketing, or biz stuff.  (3) I've postponed any OCaml
developments and am currently looking at Scheme (ala Bigloo).  C FFIs
and tools are the driving issues here.  I might be better off in
Scheme-land.

So for me what's left is 2 generally good ideas:
- increasing timing accuracy for startup / shutdown
- better groupings of scoring categories

I don't acutally *need* to accomplish either of these things.  They're
funzies.  I'd like to do them if there's community energy to deal with
them.  But if there is to be a long exercise in opposition, I'd skip
'em.  Generally I gird my loins for problems I *need* to tackle.


Cheers,                         www.indiegamedesign.com
Brandon Van Every               Seattle, WA

20% of the world is real.
80% is gobbledygook we make up inside our own heads.