High Performance Simulations of the Cellular Potts Model

The lattice cellular Potts model is widely used to represent physical and biological cellular systems. Standard implementations of this model in the literature employ inefficient Monte Carlo protocols to select elements (labels) of the lattice storing the investigated system. Recently, we proposed a new parallel algorithm, the Random Walker algorithm (RW), that can generate speedups from 3 to 10 times in simulations of this model. In this paper we present a distributed implementation of the RW algorithm exploiting parallel architectures. Our tests in a computer cluster show speedups between 5.7 and 15.6 in relation to the distributed standard algorithm.