Amemory-hierarchy-based optimization of MECA (ModifiedEquivalent Current Approximation) for the analysis ofelectrically large dielectric and lossy structures

摘要

Asymptotic high frequency computational techniques are widely used in modeling scenarios with electrically large scatterers. A well-known technique of this type, Physical Optics (PO), has been extensively tested in several contexts [1], [2]. Two recent works [3], [4] present a new PO analysis of dielectric and lossy geometries. In this method, called MECA, the interface is discretized into planar triangular facets. Then, the Snell reflection coefficients establish the relation between incident and reflected waves. This modified PO method successfully models problems which may be prohibitive for' full-wave simulation. For the present work, we developed an OpenMP parallel version of MECA in order to reduce even more the execution time. In our parallel code, we implemented memory-hierarchy-based optimization techniques [5], [6]. As seen in the results of this paper, these techniques can improve the computational; performance when calculating the scattered fields.