[Shootout-list] Re: New test in CVS: implicitode

[Bengt Kleberg]

Brent Fulgham wrote:
...deleted
> We tend to run the benchmarks a number of times (3,
> to be precise) for a variety of input parameters
> (N = 1000, 2000, etc.) so that we can get a feel for
> program behavior as the problem size increases.

one important thing that i was told last time, when someone finally 
explained the input parameter range ''problem'' to me, was the following:

if a program fails to produce a correct result for the largest argument 
it will not be possible to give it a rating in the shootout. ie, it is 
impossible to use the shootout (as presently implemented) to see when 
different languages/implementations ''fall off the edge'' on a benchmark.


if there comes a time when somebody else but me becomes interested in 
limitiations of languages/implementations, i would like to collaborate 
upon removing this problem with the shootouts current implementation.


bengt