Introduction    Download    Features    Manual    Publications    Acknowledgements   Feedback

Introduction

OpenFFT is an open source parallel package for computing three-dimensional Fast Fourier Transforms (3-D FFTs) of both real and complex numbers of arbitrary input size. It originates from OpenMX (Open source package for Material eXplorer). OpenFFT adopts a communication-optimal domain decomposition method that is adaptive and capable of localizing data when transposing from one dimension to another for reducing the total volume of communication. It is written in C and MPI, with support for Fortran through the Fortran interface, and employs FFTW3 for computing 1-D FFTs. 

OpenFFT is developed by Truong Vinh Truong Duy and Taisuke Ozaki at the University of Tokyo. 

Download

OpenFFT version 1.1 is the latest vesion of OpenFFT, and is downloadable from here. Please note that the API of  OpenFFT1.1 is incompatible with that of OpenFFT1.0 due to the additions of the auto-tuning feature and r2c interface. Upgrade to OpenFFT1.1 from OpenFFT1.0 is highly recommended. 

RELEASE NOTES:

Features

Manual

The manual of the latest version: manual1.1.html (HTML), OpenFFT1.1_Guide.pdf (PDF). 

Publications

[1] T.V.T. Duy and T. Ozaki, "A decomposition method with minimum communication amount for parallelization of multi-dimensional FFTs", Computer Physics Communications, Vol. 185, Issue 1, pp. 153-164, 2014. DOI: 10.1016/j.cpc.2013.08.028

[2] T.V.T. Duy and T. Ozaki, "A three-dimensional domain decomposition method for large-scale DFT electronic structure calculations", Computer Physics Communications, Vol. 185, Issue 3, pp. 777-789, 2014. DOI: 10.1016/j.cpc.2013.11.008

[3]  Best Research Poster Award for "OpenFFT: An Open-Source Package for 3-D FFTs with Minimal Volume of Communication", the 29th International Supercomputing Conference (ISC'14), 2014.

Acknowledgements

This package has its origins in OpenMX (Open source package for Material eXplorer), and has been funded by CMSI (Computational Materials Science Initiative) of the HPCI Strategic Program (SPIRE) of the Ministry of Education, Culture, Sports, Science and Technology of Japan. We are thankful to Japan Advanced Institute of Science and Technology (JAIST) for the computational resources. We also thank Prof. Katsumi Hagita of National Defense Academy of Japan for helpful discussions and contribution to the r2c interface.  

Feedback

Please feel free to drop us a line at duytvt@issp.u-tokyo.ac.jp (Truong Vinh Truong Duy) or t-ozaki@issp.u-tokyo.ac.jp (Taisuke Ozaki) for questions, comments, suggestions, and bug reports.   

Last updated: November 11, 2014.