Parallel implementation of the two-dimensional Fourier transform for high-speed derivation of computer generated holograms
Abstract
We use parallel computation of a two-dimensional fast Fourier transform (2D FFT) to speed-up the derivation of a phase-only computer generated hologram (CGH). To calculate the 2D FFT in parallel, we implemented the algorithm on a cluster computer. Rather than splitting the core FFT routine in parallel, we simplify the approach to split the input 2D data into various compute nodes where a complete FFT routine is performed for the distributed data. This minimizes the communication overhead between cluster nodes and can therefore be a potential routine of a Gerchberg-Saxton algorithm to derive CGH. We show the improvement in terms of speed for calculating forward 2D FFT of a precalculated phase-only CGH. The computation time is approximately dependent on the inverse relationship with respect to the number of nodes.