Optimization of Iterative Physical Optics Method through CPU-GPU Heterogeneous Technology

With the rapid advancement of radar technology, electromagnetic simulation has found widespread applications across various fields. When solving for the Radar Cross Section (RCS) of electrically large scattering bodies using the Iterative Physical Optics Method (IPO), the number of unknowns is significant, leading to substantial memory usage and computation time. To address this issue, this paper constructs an IPO method based on parameter space for computing the RCS of electrically large radar targets and introduces CUDA parallel computing (Compute Unified Device Architecture) technology to achieve parallel computation on heterogeneous CPU-GPU platforms. Compared with commercial software, a speedup of 137.79× is achieved on an NVIDIA GeForce RTX 3050 GPU. The exemplary results demonstrate the feasibility and efficiency of the CPU-GPU heterogeneous parallel IPO algorithm, enabling the resolution of scattering problems for electrically large targets that previously could only be tackled on high-performance computers or computer clusters.