On parallel processing in computational fluid dynamics

Two popular finite difference techniques in computational fluid dynamics are parallelized and tested on a Sequent 20-processor Symmetry computer. Several parallel implementations are investigated in the context of the simple heat and the more complicated curvilinear grid mapping equations. Significant speedups are demonstrated with the largest ones resulting from more complex equations, larger size grids, and more stringent convergence criteria. A novel but intuitive approach in which processor loads are balanced between successive major computational tasks is presented, and shown to yield speedups which are closer to theoretical