Computational Electromagnetics

SUMMARY

Development and Implementation of Computational Electromagnetics Techniques on Massively Parallel Computing Architectures

PARTICIPATING INSTITUTIONS

NPAC, Syracuse University
Syracuse Research Corporation

KEY CONTACTS

Gang Cheng | gcheng@npac.syr.edu | 315-443-2083

Chung-chi Cha | 315-426-3221

INDUSTRIAL/REAL WORLD IMPACT

Migrate parallel computing technology to current CEM techniques
Develop and demonstrate a major increase in CEM capability, particularly, the RCS(Radar Cross Section) prediction of a full-scale aircraft in the 1 GHz range

PROJECT DESCRIPTION

The primary objective of this project is to demonstrate the potential of the method of moments(MOM) technology, when impelmented using parallel computing, for RCS modeling of full scale airborne vehicles. We are developing explicit message-passing parallel algorithms of a 3D MOM code, PARAMOM, which was developed previously at Syracuse Research Corporation(SRC) for accurate and efficient predictions of RCS. Our approach is to develop the PARAMOM code on multiple MIMD platforms, including CM5, Intel machines and IBM SP-1, to provide a concrete demonstration of the portability and scalability of the parallel implementation and to provide performance comparisons between these machines.

The most distinctive feature of PARAMOM(which stands for "Parametric Patch Method of Moment") code is its use of basis functions that conforms to parametric surfaces with curvature. The ability of PARAMOM to accurately account for surface curvature with the use of parametric surface patches has in many cases shown significant advantage over conventional techniques that employ flat facets as modeling elements.

The features of the initial parallel algorithm design will be scalable data decomposition, use of stable and efficient parallel linear solvers for large dense matrices, effective use of memory , and optimized balance of computation and communication for this application.

We choose to adopt in our implementation the scalable and portable parallel linear algebra package ScaLAPACK developed at ORNL and would also investigate the use of vendor-supplied package, e.g. ProSolver-DES on the Intel machine and CMSSL dense solver on CM5, to achieve optimum performance.

PROJECT STATUS