[Shootout-list] Directions of various benchmarks
Bengt Kleberg
bengt.kleberg@ericsson.com
Wed, 25 May 2005 07:42:01 +0200
On 2005-05-24 18:41, John Skaller wrote:
...deleted
> I have a variation. Start with n=1 and add 1. End of story.
> If the test needs to be done for 10, 100, 1000, 10000 .. etc,
> then build that into the test. eg calculate m = 10 ^ n and use
> m as the control variable.
this is a good idea. i have not considered it since it would mean
changing all the source of the existing tests, on top of the test system
(minibench, or more exactly the ''range'' arguments to minibench). to be
a sufficiently small change, to get past those that do not want more
values in the graphs, i have tried to concentrate upon only changing one
place (the test system).
moreover, i find it difficult to inform the tested program to back off
if it goes over the time out limit. backing off might not be that
important, of course.
however, if this is more paletable than my idea i am all for it. ie, i
do not mind more work if it is an acceptable way forward.
bengt