| Map overlay is one of the fundamental operations of Geographic Information Systems. Given its processor intensive nature, important benefit can be achieved by parallelizing it over several processors. The unequal distribution of data, however, makes load balancing difficult to achieve. In our research, we have used a non-uniform partitioning technique to distribute hierarchical data between the processors. We have achieved parallel efficiencies between 40 to 96% using up to 36 processors.; Moreover, we have developed a floating-point filter to solve numerical problems that affects the line intersection algorithm, a crucial component of the map overlay operation. The speed and robustness of our filter greatly improves the map overlay operation. |
