Heterogeneous CPU-GPU Accelerated Parallel Subgridding FDTD Algorithm

The subgridding method can efficiently deal with the tremendous mesh involved in the Finite-Difference Time-Domain simulation. Here, we introduce a heterogeneous CPU-GPU parallel technique to further accelerate the subgridding FDTD algorithm. According to the proposed parallel strategy, GPU is used to calculate dense grid regions, and CPUs only take into account the coarse mesh region. The algorithm’s efficiency is benchmarked by simulating a frequency-selective structure. The numerical results reveal that the execution speed of our proposed method is significantly improved, while the computational accuracy is also guaranteed.