[Shootout-list] X per second scoring system, resume

Einar Karttunen ekarttun@cs.helsinki.fi
Thu, 30 Sep 2004 14:36:06 +0300


On 30.09 09:05, Bengt Kleberg wrote:
> one way to get this X-per-timeunit is (imho) quite simple. we take the 
> graph that is created for every test and extraxt X-per-timeunit from it. 
> (do i need to explain how?)

The only problem is that you will stumble upon problems when the
time per problem size is not linear. This can occur quite frequently.

I don't currently see much point in implementing a more complex timing 
system.

> another way to get X-per-timeunit is to re-write all our test to include 
> timers. it will also mean accepting that the timers might (depending 
> upon implementation) disturb the test result more or less. also, some 
> test (eg reverse) are very difficult to do like this.

Timers will make things harder as:
* Timers can add considerable overhead depending on the language
* Results will vary very much in garbage collected languages

- Einar Karttunen