106. Extreme Fidelity Computational Electromagnetic Analysis in the Supercomputer Era
Authors: Brian MacKie-Mason (University of New Mexico)Zhen Peng (University of New Mexico)Christopher Kung (Engility Corporation)
Abstract: Ever-increasing fidelity and accuracy needs for advanced electromagnetic applications have pushed problem sizes toward extreme scales. It puts a high premium on parallel and scalable algorithms with optimal computational complexity. This poster displays the research into high-performance, geometry-aware domain decomposition (DD) methods for the solution of time-harmonic Maxwell’s Equations. The technique ingredients include a volume-based optimized Schwarz finite element DD method, and a surface-based interior penalty boundary element DD method.
The work has three benefits: (i) it results in robust, cost-effective preconditioning techniques that reduce the condition number of very large systems of equations; (ii) it provides a flexible and natural way to set up the mathematical models, to create the problem geometries and to discretize the computational domain; (iii) it leads to parallel and scalable algorithms to reduce the time complexity of extreme-scale simulations. The capability of the algorithms is illustrated through real-world applications on HPC systems.
Two-page extended abstract: pdf