High-Performance Hybrid CPU–GPU Parallel Finite Element Solver for Large-Scale Electromagnetic Simulation and Accelerated RCS Computation

This paper presents a scalable finite element solver for efficient simulation of large-scale electromagnetic wave phenomena. It is developed specifically for high-frequency applications and is designed to accurately capture complex field variations using advanced numerical techniques. A key feature of the framework is its hybrid parallel architecture which combines domain decomposition across multiple CPU cores with GPU acceleration to maximize computational speed and memory efficiency. The solver also includes a dedicated GPU-based module for fast post-processing of Radar Cross Section (RCS) data, markedly reducing overall computational cost. Performance evaluations on electrically large scattering problems show that the solver can handle systems with tens of millions of unknowns while maintaining high accuracy and excellent parallel scalability. The proposed method provides a practical and powerful solution for high-fidelity electromagnetic simulations and is well suited for various applications in antenna and propagation engineering.