A GPU approach to FDTD for radio coverage prediction

The benefits of using Finite-Difference alike methods for coverage prediction comprise highly accurate electromagnetic simulations that serve as a reliable input for wireless networks planning and optimization algorithms. These algorithms usually require several thousands of iterations in order to find the optimal network configuration, so to obtain results within reasonable computation times, the applied propagation models must be as fast as possible. In this study an implementation-oriented analysis on the suitability of using Graphics Processing Units (GPU) to perform Finite-Difference Time-Domain simulations is carried out. We believe that the recently released Compute Unified Device Architecture (CUDA) technology has opened the door for computational intensive algorithms such as FDTD to be considered for the first time as a precise and fast propagation model to predict radio coverage.