[Shootout-list] fannkuch (timer resolution; HZ=1000?)
Bengt Kleberg
bengt.kleberg@ericsson.com
Tue, 24 May 2005 07:40:16 +0200
On 2005-05-23 15:55, Jon Harrop wrote:
...deleted
>
> Alas you can't do this because the startup time for each benchmark is likely
> to be quite different (at least for some languages) because different
> benchmarks will exercise different parts of the run-time which may incur
> their own startup.
my line of thought was that since the startup time might be different
for different benchmarks it would be ''better'' to measure the startup
time for this particular benchmark. as opposed to using the startup time
for a specific benchmark (hello).
> Otherwise, yes, we could just have subtracted the startup time.
this makes me understand the error of my ways. if somebody have all the
work done in the startup then their runtimes would be very small if the
startup is subtracted.
> A better solution would be to study the results for various "n".
i agree. i only mentioned this thought (different statup times for
different benchmarks) to see if it would be helpful to the people
interested in one-number-fits-all.
bengt