High performance FDTD algorithm for GPGPU supercomputers

Abstract

An implementation of FDTD method for solution of optical and other electrodynamic problems of high computational cost is described. The implementation is based on the LRnLA algorithm DiamondTorre, which is developed specifically for GPGPU hardware. The specifics of the DiamondTorre algorithms for staggered grid (Yee cell) and many-GPU devices are shown. The algorithm is implemented in the software for real physics calculation. The software performance is estimated through algorithms parameters and computer model. The real performance is tested on one GPU device, as well as on the many-GPU cluster. The performance of up to 0.65 • 1012 cell updates per second for 3D domain with 0.3 • 1012 Yee cells total is achieved.