A parallel implementation of a cellular automata based model for coffee percolation

A cellular automata (CA) model for fluid-dynamic simulations is presented as a strategic tool for monitoring physical properties which are difficult to be measured by an experiment. The use of an arbitrary number of rest particles let the FHP-N model recover the Galileian invariance, which was missing in the standard FHP model. The code has been implemented on a Gray T3E supercomputer. The overall speed-up performance is very good, being superlinear, as far as particles propagation and average procedure are regarded. The efficiency, i,e,, a measure of the overhead of the parallel code versus the sequential one is constant up to 32 processors, and it is over 60% when 128 PEs are involved. The simulation results agree qualitatively with the experimental ones, thus confirming the model to be an effective tool for simulating a wide range of situations. (C) 2001 Elsevier Science B,V. All rights reserved.