Parallel algorithm for surface panel analysis. Southampton, UK, University of Southampton, 65pp.
(Ship Science Reports, (63) ).
This report describes the development of a paralled algorithm to carry out surface panel analysis. The algorithm is described to run across variable sized arrays of transputers. All software was written in Occam2. The parallelism strategy was based on geometric parallelism for both the data and code. A flexible geometry definition allows a wide range of flow geometries to be investigated. A perturbation potential formulation is used with an iterative Kutta condition to give zero pressure loading at the trailing edge. An investigation was carried out to compare the relative merits of iterative and direct methods for solving the influence coefficient matrix. An iterative block Jacobi was found to give the best performance when the block size was equal to the number of panels in a chordwise strip. The overall performance of the PArallel LIfting SUrface PANel code (PALISUPAN) was measured and although an implicit method it gave a high code efficiency.
Actions (login required)