[Shootout-list] small change (that i can do) that would benefit 3 people
Bengt Kleberg
bengt.kleberg@ericsson.com
Sat, 07 May 2005 14:08:19 +0200
greetings,
if the shootout changed from using ''time for fixed n'' as the ranking
reason to ''time/n'' the following 3 people would ultimatly be happier:
Jon Harrop wrote:
...deleted
> Anyway, the time is hugely noisy because it is so short:
Isaac Gouy wrote:
...deleted
> Maybe we need to use a bimodal data range - 3 high workload (for C and
> Ada) and 3 low workload (for Ruby and Tcl).
Bengt Kleberg wrote:
...deleted
> ie, it is impossible to use the shootout (as presently implemented)
> to see when different languages/implementations ''fall off the edge''
> on a benchmark.
the change itself would not make any difference to us 3, nor would it
make any change to the current rankings. what would change is that
afterwards it is possible to solve all 3 problems by changing the values
of 'n' used for each test in the shootout. i would be happy to do this
change, too.
the change would only be for one entry in ''data.csv'' (ie, ndata.csv
would be left untouched).
i currently do this calculation with one line (90 characters) in a shell
script, so i do not forsee any problems due to increesed complexity in
minibench.
many languages would finally be allowed to take about 1 second on a
test, but the overall time to run a test would not increase since other
languages no longer need to run for minutes.
are there any reason not to do this change?
bengt