UPML-FDTD Parallel Computing on GPU

Ma Weiwei,Sun Dong,Wu Xianliang
DOI: https://doi.org/10.1109/icmmt.2012.6230262
2012-01-01
Abstract:Finite-Difference Time-Domain method (FDTD) which is known with its simple and flexible is widely used for the calculating of electromagnetic fields. However, it costs a lot of time in simulating the electrical-large object. To solve the problem, parallel method for UPML-FDTD algorithm was put forward using GPU based-on Compute Unified Device Architecture (CUDA) in this paper. The algorithm is further optimized by using texture memory. Comparing with the calculation with traditional CPU, The result of simulation shows that this algorithm has enough precision and the remarkable increased efficiency was acquired in different sizes of Yee cells.
What problem does this paper attempt to address?