[Shootout-list] Re: [Gd-chatter] examples/shootout strcat.dylan, NONE, 1.1

Bengt Kleberg bengt.kleberg@ericsson.com
Mon, 18 Oct 2004 12:56:43 +0200


Isaac Gouy wrote:
...deleted
>>adjust n to make the fastest language run in (1, 2...somethingelse) 
>>seconds. just guessing that ''somethingelse'' should be 5 is better
>>than what we have now, but we could try to find a good value of 
>>''somethingelse''.
> 
> 
> - when all the test programs have been re-written so that increasing N
> doesn't change the nature of the test.
> 

i agree. if the test requirement is such that increasing n changes the 
nature of the test it is not a good thing (tm) to increase n.



> - when someone donates cpu time required to run test continually
> 

i agree. if we want to run a test continually we would need more cpu time.


> - when this becomes a Shootout between long running app servers

i disagree. 1 second is not a good value for long running app servers.


>>we still need to have several values of n and the graphs. they are 
>>indespensable. imho.
> 
> 
> imo They have a very limited benefit. 
> 
> They demonstrate that cpu times did indeed increase with increasing
> work - if cpu times didn't increase then we start to be suspicious of
> the implementation and/or measurements. That's all.
> 
> There are too few different values to say anything else.

i agree. the current shootout graphs have too few values.
i will rephrase it:
	graphs are indispensable, if properly used.


bengt