Dual-Level GPU Acceleration for Time-Domain Physical Optics (TDPO): Facet-Observer Hybrid Parallelism in Electromagnetic Scattering Analysis

The growing complexity of electromagnetic scattering scenarios demands efficient time-domain solvers for large-scale problems. Time-Domain Physical Optics (TDPO) provides a viable framework for transient electromagnetic analysis. But its computational cost becomes prohibitive when handling complex targets with fine discretization or wide-angle observation configurations. This paper proposes a dual-level GPU acceleration strategy that introduces facet-observer hybrid parallelism to overcome the scalability limitations of conventional TDPO implementations. By decomposing the computational workload into facet-level parallelism and observer-level parallelism, the method achieves better utilization of GPU computational resources. Several straightforward numerical cases are presented to demonstrate the accuracy and efficiency of the proposed method.