Parallel 3D-FFT Spherical Decomposition Implementation

摘要

In order to introduce more efficient high performance computing (HPC) applications we need to provide new solution paradigms for the complex heterogeneous systems. Most of the HPC applications, as in the molecular dynamics simulation, implement the parallel FFT algorithm that consumes a major portion of the application execution time. Parallel FFT algorithms are always subject for performance enhancements. This work provides a new methodology for reducing the FFT algorithm data input. In addition, this paper aims to provide a prototype implementation for parallel 3D FFT spherical cutoff data input reduction.