Vectorized directional sweep based parallelization method for factored alternating direction implicit (ADI) schemes

A new parallelization method is proposed for factored alternating direction implicit (ADI) schemes based on the vectorized global domain directional s weep. This approach, when combined with multi-partitioning domain decomposi tion, significantly reduces the frequency of necessary communication calls and minimizes processor idling during the sweeping processes. The present p arallelization approach is applied to a number of vectorized two-dimensiona l compressible Navier-Stokes solvers. The codes vary in complexity from lam inar to algebraic turbulence closure model and finally the strongly coupled Navier-Stokes and K-epsilon equations. Implementation is conducted using P VM (Parallel Virtual Machine) message passing tool on the Gray T3D massivel y parallel processing (MPP) machine. The implemented parallel codes are ass essed in terms of accuracy and parallel performance.