Collaborating CPU and GPU for the electromagnetic simulations with the FDTD algorithm

Abstract

Summary The co‐processors with the capability of powerful floating‐point operations have been used to study the electromagnetic (EM) simulations with the finite‐difference time‐domain (FDTD) method, which usually consumes huge computational resources. This work studied the implementation and optimization of the three‐dimensional FDTD algorithm with the uniaxial perfectly matched layer boundary conditions on GPU clusters. In this study, CPUs and GPUs collaborate to solve the different parts of the computation domain concurrently, and a set of techniques are used to optimize the execution efficiency of GPUs, such as the use of GPU texture memory, the pinned host memory, and the asynchronization of data transfer between CPU and GPU. For the CPU‐GPU collaborative mode, the problem size solved is not limited by the size of global memory on GPU because the computational capability of CPU is also used for the calculation of EM field. To keep the load balance between CPU and GPU, 85% of the computation load is assigned to GPU for NVIDIA Tesla K40m, and the collaborative efficiency is found to be less than 75%. Copyright © 2016 John Wiley & Sons, Ltd.

Nvidia CUDA parallel processing of large FDTD meshes in a desktop computer

“…Otros trabajos se centran en CPUs y GPUs que colaboran para resolver distintas partes del dominio de simulación de forma concurrente [7], y también se implementan entornos de cluster multi-GPU para conseguir mayor velocidad e investigar numéricamente el rendimiento [8], [9]. Se encuentran planteamientos MPI (message passing interface) -OpenMP [10]-GPU para calcular RCS con FDTD, que consiguen reducir el tiempo de cálculo de 3dias a 0.8h (RCS de un F-117) [11].…”