The number of units (n
) is really up to the end user. I really can't conceive of n
ever being less than 100 or more than a million. It depends on how complex the simulation needs to be. I would guess the "average" user would opt for around 1,000 to 10,000 units. An important point is that I am planning to have some units grouped together moving as one "super-unit" which would reduce the burden of the fog of war calculation. If I had to guess, I would say that in the limit as n
goes to infinity I could probably expect a 10 fold drop in the number of units I have to run calculations on. It is possible that I may not have performance issues, but I would like to have a fairly efficient algorithm to start with.
I don't need cube partitions since all the units are positioned on the surface of a sphere. Would it be best to divide up the units into lists representing portions of the earth's surface area? It is easy to divide up the sphere surface into rectangles (like this: http://upload.wikimedia.org/wikipedia/commons/3/38/Sphere-wireframe.png
) , but it might be best to just lump all of the polar regions into two lists because the rectangles get small at the poles and there will probably be little activity there. For a unit in a given rectangle, I would then see if any units in the adjacent rectangles can see it. I'm new to this, so perhaps you can think of a more efficient way to divide up the units.