[Shootout-list] benchmark design, hard limit for how many 'n'?

[Bengt Kleberg]

greetings

could someone tell me if there is a hard limit for how many 'n', and if 
so, what it is?


while looking at the number of 'n's for a test i noticed an interesting 
thing. some time ago, when i argued that 3 'n' values where too few, i 
was told that the basic design of the shootout ''kernel'' (minibench 
script) mandates this low number of 'n's.
i argued for making the number of 'n's bigger, but nobody else was 
interested. and atleast one person was very much against it.

so here i was looking at 'n' for ackerman today. and it is not 3, it is 
5. perhaps the hard limit of 3 was not so hard as explained?
what is the limit?


bengt