Achieving Fast Computer-Generated Hologram Calculations via Parallelization

Computer-Generated Holography (CGH) plays an important role in the development of three-dimensional display. However, the enormous computational time for CGH generation hinders the practicality of the CGH—depending on a specific application, it takes several hours, even more to complete CGH computations. To enhance the hologram computation speed, we present a parallel computing approach to accelerating long-running computations of the CGH application to provide high resolution digital holograms. A typical sequential version of a CGH program is parallelized by extracting independent tasks. Message Passing Interface (MPI) is used to communicate with parallel tasks and an entire computation will be assigned among different computational nodes. For all the experiments, we will use a supercomputer which is a 60-node heterogeneous CPU/GPU cluster. To validate our approach, we have applied our parallel version of the CGH to generate a digital hologram from a 3D object. The experimental results demonstrate that a variety of parallelization strategies could be a novel way to increase computation speed improvement for computationally-intensive CGH applications. 1288 Hyun Jun Choi and Dong Kwan Kim