Implementation and Performance of the Parallel Ecological Simulations

The spatial and temporal aspects of population dynamics are pivotal to computational biology. We developed a spatially explicit model of epidemics that behaves like a large probabilistic cellular automaton. The cells of the automaton are discrete sites into which the habitat is partitioned. Probabilistic local state transitions are executed synchronously at all sites making the simulation suitable for parallel implementation on SIMD architec-tures. An interpretation of the simulation results requires computing global parameters of the habitat that are challenging to implement eeciently on SIMD machines. For example , pattern detection and measurements use sophisticated image processing algorithms. In this paper, we discuss the parallel implementation of the epidemiological model and analyze its performance. The achieved results indicate that the use of a massively parallel machine was necessary and eecient.