Parallel FFT Libraries
Evangelos Brachos · 2011
The focus of this project is the area of the fast Fourier transform (FFT) libraries (FFTW 3.3, P3DFFT 2.4 and 2DECOMP&FFT 1.1.319). During this project we benchmark the libraries on Hector phase 2b supercomputer. The main goal of this project is to examine the performance of the libraries which are built on the top of the FFTW’s latest version and compare their results with the FFTW itself. Moreover, the ability of the P3DFFT and the 2DECOMP&FFT to implement the 2D decomposition, let us experiment with the scaling of these libraries in thousands of cores. The findings from this experiment provide evidence of the FFTW library’s excellent scaling by using the 1D decomposition and the ability of the other two libraries to minimise the time required for the Fourier transform computation, by using the 2D decomposition. The main conclusions drawn from this study were the ability of the P3DFFT and the 2DECOMP&FFT libraries to scale up to thousands of cores and the affects of the Gemini Interconnect on their performance.