SHIFT : Scalable High-level Interface for Fourier Transforms

Introduction

SHIFT is a lightweight Python library designed to simplify Fourier transforms and spectral calculations on regular grids. Built around SciPy's FFT routines, SHIFT provides a convenient interface for computing Fourier, discrete cosine (DCT), and discrete sine (DST) transforms in one, two, and three dimensions, alongside utilities for constructing grids and their corresponding Fourier-space modes.

A central motivation behind SHIFT is to make parallel Fourier transforms easier to install, use, and integrate into scientific applications. Unlike implementations requiring FFTW and additional compiled extensions, SHIFT uses SciPy's FFT backend and supports distributed transforms through mpi4py, employing an MPI slab decomposition. This avoids the need for a separate FFTW installation while retaining compatibility with standard scientific Python environments.

In addition to spectral transforms, SHIFT provides utilities for MPI communication and distributed array operations, making it suitable for integration into larger scientific computing workflows.

NOTE: SHIFT was originally conceived as a framework for Fourier transforms in Cartesian, polar, and spherical coordinate systems. Current development focuses on Cartesian spectral transforms and MPI-based scientific computing. Experimental polar and spherical transform implementations are available in the sphere branch, but are not part of the supported release and should be used with caution. Further development of these capabilities is planned.

Dependencies

• numba
• numpy
• scipy
• mpi4py [Optional: enables MPI distributed processes]

Installation

SHIFT can be installed via pip:

pip install shift-fft

Or by cloning the github repository:

git clone https://github.com/knaidoo29/SHIFT.git
cd SHIFT
pip install .

The latter of which can be used to install specific branches.

Once this is done you should be able to call SHIFT from python:

import shift

To use the MPI functionality please take a look at the documentation which instructs users how to use the SHIFT MPI object and how to run these distributed jobs successfully without errors or MPI related hanging.

Documentation

In depth documentation and tutorials are provided here.

License

SHIFT is released under the MIT License. See LICENSE for details.

Citation

If you use SHIFT in your work, please cite:

  @software{naidoo_shift_2025,
    author       = {Naidoo, Krishna},
    title        = {SHIFT: a scalable MPI library for computing fast Fourier transforms in python},
    year         = 2025,
    publisher    = {Zenodo},
    doi          = {10.5281/zenodo.17093446},
    url          = {https://doi.org/10.5281/zenodo.17093446}}

Support

If you have any issues with the code or want to suggest ways to improve it please open a new issue (here) or (if you don't have a github account) email krishna.naidoo.11@ucl.ac.uk.