Now let's look into the communication overhead required by parallel computing.
Assume that we have n-squared course or processors,
each of which handles its own square of our complete domain.
The complete domain should consist of l * l cells.
What would be a reasonable model for the time taken by the computation?
A constant C by the number of course or processors and squared
times the side length of the total domain plus another constant
times n times l?
Or constant times l-squared divided by n-squared plus another constant times (n - 1)?
Or the constant times n times l-squared minus a constant times n times l?
A constant times l-squared over n-squared plus a constant times (n - 1) times l-squared?
Pick one.