论文信息 - An Implementation of Parallel 3-D FFT with 2-D Decomposition on a Massively Parallel Cluster of Multi-core Processors

An Implementation of Parallel 3-D FFT with 2-D Decomposition on a Massively Parallel Cluster of Multi-core Processors

In this paper, we propose an implementation of a parallel three-dimensional fast Fourier transform (FFT) with two-dimensional decomposition on a massively parallel cluster of multi-core processors. The proposed parallel three-dimensional FFT algorithm is based on the multicolumn FFT algorithm. We show that a two-dimensional decomposition effectively improves performance by reducing the communication time for larger numbers of MPI processes. We successfully achieved a performance of over 401 GFlops on 256 nodes of Appro Xtreme-X3 (648 nodes, 147.2 GFlops/node, 95.4 TFlops peak performance) for 2563-point FFT.

Daisuke Takahashi

[1] Ramesh C. Agarwal,et al. An efficient parallel algorithm for the 3-D FFT NAS parallel benchmark , 1994, Proceedings of IEEE Scalable High Performance Computing Conference.

[2] Steven G. Johnson,et al. The Design and Implementation of FFTW3 , 2005, Proceedings of the IEEE.

[3] Bin Fang,et al. Performance of the 3D FFT on the 6D network torus QCDOC parallel supercomputer , 2007, Comput. Phys. Commun..

[4] J. Tukey,et al. An algorithm for the machine calculation of complex Fourier series , 1965 .

[5] Andy Brass,et al. Two and three dimensional FFTs on highly parallel computers , 1986, Parallel Comput..

[6] Daisuke Takahashi. Efficient implementation of parallel three-dimensional FFT on clusters of PCs , 2003 .

[7] C. Loan. Computational Frameworks for the Fast Fourier Transform , 1992 .

[8] Daisuke Takahashi. A Hybrid MPI/OpenMP Implementation of a Parallel 3-D FFT on SMP Clusters , 2005, PPAM.

[9] Robert S. Germain,et al. Scalable framework for 3D FFTs on the Blue Gene/L supercomputer: Implementation and early performance measurements , 2005, IBM J. Res. Dev..