Efficient Parallelization of Eulerian-Lagrangian Approach for Disperse Multiphase Flow Calculations on MIMD Computer Architectures

The parallelization of the solution algorithm for the set of continuity, Navier–Stokes and turbulence model equations is carried out by parallelization in space, that means by application of the domain decomposition or grid partitioning method. Using a block-structured grid the flow domain is partitioned in a number of subdomains. Usually the number of grid blocks exceeds the number of processors, so that each processor of the parallel machine (PM) has to handle a few blocks. The grid-block-to-processor assignment is given by a heuristicaly determined block–processor allocation table and remains static and unchanged over the time of fluid flow calculation process. The gas flow calculation is then performed by individual processor nodes on the grid partitions stored in their local memory. Flow characteristics along the grid block boundaries which are common to two different nodes have to be exchanged during the solution process by inter–processor communication. Details of the parallel solution method for the gas flow can be found in [1].