On Wednesday, April 24, 2002, at 01:24 , Bruce Van Allen wrote:
[..]
> Another approach might be to increase the number of repetitions performed 
> by Benchmark. A factor of 10^3 raises milliseconds to seconds, etc. 
> Sub-seconds would be important in high-volume applications, so more 
> repetitions might be more realistic anyway...

first off thanks to folks.

as a Generic Rule of thumb I run them on the order of about
a million iterations just to make me feel all warm and fuzzy
about generic reasonable approximations for the RUN/sec value
to get even close to 'statistically interesting'.

My premise - some of it up at:

http://www.wetware.com/drieux/CS/lang/Perl/Beginners/BenchMarks/

was to provide the 'benchmark' test cases for various standard
ways of doing 'how do I do foo' - and let the 'numbers speak
for themselves'.....

and it just struck me that I may be 'asking the wrong question'.

since clearly Benchmark.pm has not really been hacked on since
'99 - and in the five years from Tim Bunce's original work -
no one has seen a clear need to run it with HiRes....

The problem - as I see it now - is how to differenciate 'trivia'
contest bench marking from 'relevant' benchmarking....


ciao
drieux

---

Reply via email to