[Shootout-list] Ray tracer
Isaac Gouy
igouy2@yahoo.com
Thu, 28 Apr 2005 22:40:57 -0700 (PDT)
--- skaller <skaller@users.sourceforge.net> wrote:
> On Fri, 2005-04-29 at 13:45, Isaac Gouy wrote:
> > What problem are these categories supposed to solve?
>
> Allow larger number of tests, hence more reliable results,
> whilst still allowing people to comprehend: the amount
> of available time someone would have examining the
> Shootout being a point you raised yourself.
Has anyone suggested that having a larger variety of test would in some
sense give more reliable results?
Apart from those I know to be plain wrong - the results do seem to be
reliable. The problem is elsewhere. The problem is that sometimes
people wish to draw general conclusions from limited data.
> Whether it would do so, and/or be worth the effort,
> is another question. What do yo think?
As a design tool for checking which benchmarks overlap - yes.
As an excuse for adding another and then another benchmark - no.
Writing a set of shootout benchmarks shouldn't become some marathon
chore.
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com