mpiFFT4py performs FFTs in parallel in Python. It is developed to be able to do FFTs in parallel on a three-dimensional computational box (a structured grid), but there are also routines for doing the FFTs on a 2D mesh. It implements both the slab and the pencil decompositions.
mpiFFT4py requires numpy for basic array oparations, pyfftw for efficient FFTs and mpi4py for MPI communications. However, if pyfftw is not found, then the slower numpy.fft is used instead. cython is used to optimize a few routines. Install using regular python distutils
python setup.py install --prefix="Path on the PYTHONPATH"
To install in place do
python setup.py build_ext --inplace
To install using Anaconda, you may either compile it yourselves using (from the main directory)
conda config --add channels conda-forge
conda build conf/conda
conda install mpiFFT4py --use-local
or use precompiled binaries in theconda-forge or the spectralDNS channel on Anaconda cloud
conda install -c conda-forge mpifft4py
or conda config --add channels conda-forge conda install -c spectralDNS mpifft4py
There are binaries compiled for both OSX and linux, and several versions of Python. Note that the spectralDNS channel contains bleeding-edge versions of the Software, whereas conda-forge is more stable.
mpiFFT4py is developed by
- Mikael Mortensen
mpiFFT4py is licensed under the GNU GPL, version 3 or (at your option) any later version. mpiFFT4py is Copyright (2014-2016) by the authors.
The latest version of this software can be obtained from
https://github.com/spectralDNS/mpiFFT4py
Please report bugs and other issues through the issue tracker at: