[Shootout-list] fannkuch

[Brent Fulgham]

--- Bengt Kleberg <bengt.kleberg@ericsson.com> wrote:
> the shortest allowed run time in the shootout is
> 0.01 seconds. there is no statment in the faq about 
> how this was decided upon, nor how it was 
> measured to be sufficient.

This measurement limit is based on the resolution of
the timers available.  We found that anything below
about 0.01 second was not reliable.

> i have a c program that can be used for this. i do
> not have the skill to test the perl timeing system. 
> would it be a good idea to run the c 
> program? or should somebody that knows perl recode
> it in perl? it is simple and short.

I think your C program is great.  The only problem
is that it would involve a large shift in the way
the shootout runs, changing the measurements from
"How long does it take to do some work X", to
"How much work can be done in some time Y".

I am not opposed to adding additional X's to our
set of data runs -- this is just making the existing
infrastructure do more iterations.

I am not sure I want to go down the road of trying
to compute Y at this time since it would involve a
lot of work, and I'm not sure I see what would be
gained.

> i think it would be a good thing to check the basics
> before adding iceing (like new tests and new 
> metrics).

Agreed.

-Brent