The computational performance and power consumption of the parallel FDTD on a smartphone platform

Abstract

© 2015 ACES. The use of the FDTD in Android applications heralds the use of mobile phone platforms for performing electromagnetic modeling tasks. The Samsung S4 and Alpha smartphones computations are powered by a pair of multi-core Advanced RISC Machines (ARM) processors, supported by the Android operating system, which comprises a self-contained platform, which can be exploited for numerical simulation applications. In this paper, the parallelized two dimensional FDTD is implemented on the Samsung Smartphone using threading and SIMD techniques. The computational efficiency and power consumption of the parallelized FDTD on this platform are compared to that for other systems, such as Intel's i5 processor, and NVIDIA's GTX 480 GPU. A comparison is made of the power consumption of the different techniques that can be used to parallelize the FDTD on a conventional multicore processor. In addition to parallelizing the FDTD using threading, the feasibility of accelerating the FDTD with the SIMD registers inherent in the phone's ARM processor is also examined.