Parallel computing with panel methods : I . A feasibility study

Realistic applications of panel methods of computational13; aerodynamics represent the aerodynamic configuration by many thousands of panels, resulting in massive computational loads on sequential computers with correspondingly large computation times. By distributing the loads over several processors, parallel processing13; technology provides a useful strategy to reduce these large computation times to reasonably small levels. The present paper examines accordingly what changes must be introduced in the panel codes for use with parallel computers .13; It emerges that efficient complete parallelization can be13; achieved in the calculation of panel influence coefficient matrices, which is one of the two major compute-intensive components of any panel method application. However, the solution of linear and nonlinear dense systems of equations for the singularity strengths, which is the other major compute-intensive part of a panel method, and the solution of small bandwidth systems of equations arising in input preparation for generation of panel grids on the configuration surface, pose a serious critical challenge as the existing good serial algorithms do not lend themselves to ready easy parallelization, for optimal efficiency in a multi processor parallel computing environment, indicating a need to conduct more research on the design of efficient parallel algorithms