B-Calm: An Open-Source Multi-Gpu-Based 3d-FDTD With Multi-Pole Dispersion for Plasmonics

Abstract

Abstract-Numerical calculations based on finite-difference timedomain (FDTD) simulations for metallic nanostructures in a broad optical spectrum require an accurate modeling of the permittivity of dispersive materials. In this paper, we present the algorithms behind B-CALM (Belgium-CAlifornia Light Machine), an open-source 3D-FDTD solver simultaneously operating on multiple Graphical Processing Units (GPUs) and efficiently utilizing multi-pole dispersion models while hiding latency in inter-GPU memory transfers. Our architecture shows a reduction in computing times for multi-pole dispersion models and an almost linear speed-up with respect to the amount of used GPUs. We benchmark B-CALM by computing the absorption efficiency of a metallic nanosphere in a broad spectral range with a six-pole Lorentz model and compare it with Mie theory and with a widely used Central Processing Unit (CPU)-based FDTD simulator.

The Investigation of Efficiency of Physical Phenomena Modelling Using Differential Equations on Distributed Systems

“…In the first part of the research an algorithm to solve the 3D FDTD problem with GPU using CUDA platform was proposed and analyzed. GPU was used widely before in this field and some recent researches (Wahl, Ly Gagnon, Debaes, Van Erps, Vermeulen, Miller, Thienpont 2013) show that this type of calculations can be automatically parallelized on several GPUs using special software packages. Other alternative could be the usage of implementation simplifying tools such as (Hoshino, Maruyama, Matsuoka, Takaki 2013), which in general has approximately 50% lower performance than CUDA, however, for some applications it can reach up to 98% of CUDA efficiency.…”

Multiple-Gpu-Based Frequency-Dependent Finite-Difference Time Domain Formulation Using Matlab Parallel Computing Toolbox

“…Different types of GPUs were used to compare their real computability. Researchers from Belgium used CUDA on up to four GPUs for a 3-D FDTD application with multi-pole dispersion for plasma [7,8], and found an almost linear speedup with respect to the number of GPUs.…”

“…There are two major approaches to programming multi-GPU based parallel FD-FDTD: open computing language (OpenCL) [3,4] and compute unified device architecture (CUDA) [5][6][7][8]. OpenCL is a framework and programming language that executes across heterogeneous platforms consisting of CPUs, GPUs, or other processors.…”

Applications of the Discrete Green's Function in the Finite-Difference Time-Domain Method

“…Recently, the discrete Green's function (DGF) [1][2][3][4] has been proven to be an efficient tool facilitating the finite-difference time-domain (FDTD) method [5][6][7][8][9][10][11]. DGF is a response of the FDTD grid to the Kronecker delta current source.…”