Automatic performance estimating

[donRumata03]

It would be great to be able to estimate optimization algorithm performance automatically.
Now the creator has no ideas which algorithm is better is worth continuing developing.
This might prevent situations like this one #1

[donRumata03]

It probably makes sense to estimate performance this way:

Given some points (specific for a function) and a score for each, we compute total verdict (taking the number of fitness function computations, used to achieve the point, into account).

[donRumata03]

To compute the number of iterations required to achieve a given point (target function value), it makes sense to use binary or n-ary search on logariphmated space between a really big and a small number (probably, 0) of max_iterations transmitted to the algorithm. The boolean predicate for the x-ary search shows if the function succeed to achieve the given value in this iterations. If it didn't even success to find this value in max amount of iterations, do something with the creator of the algorithm!