Investigating the Scalability of FFT Algorithms in Contemporary Parallel Computing Environments

Abstract

Parallel programming models are quite challenging and emerging topic in the parallel computing era. These models allow a developer to port a sequential application onto a platform with number of processors so that the problem or application can be figured out easily. Adapting the applications in this mode using the Parallel programming models is often influenced by the type of the application, the type of the platform and many others. There are several parallel programming models developed and two main variants of parallel programming models classified are shared and distributed memory based parallel programming models. This article compares various techniques for the fast evaluation of Fast Fourier transform on parallel machines. In this work we present a model covering the essential features of communication systems for discussing and comparing their operational semantics. Our access is based on parallel FFT algorithms. Currently, many cores are the most suitable for the deployment of HPC (High performance Computing) infrastructures, due to their performance over cost ratio and scalability. These systems can be programmed using OpenMP(For Shared memory) and MPI(For distributed memory) and their hybrid model MPI+OpenMP (for cluster of shared memory) and  the  recent variation PGAS (Partitioned Global Address Space) languages, such as UPC (Unified parallel C),  promises more productivity and execution, providing  support for shared, distributed and their hybrid model in efficient manners. PGAS languages demonstrate very little operating cost as compared with MPI for problems that are inadequately parallel. Evaluation of sequential and MP based implementation of FFT is desirable, because FFT is one of the seven benchmarks for measuring performance of HPCC.