Kokkos-FFT

Kokkos-FFT is a library which started development in December 2023, as a part of the CExA project, and is now an official part of the Kokkos project (and thus of HPSF). Its initial goals were to offer access to a Kokkos-aware performance-driven and portable FFT, with an API as simple as possible. It has been done by building an API inspired by NumPy FFT, which dispatch work to existing FFT libs (such as cufft, hipfft, FFTW, or onemkl), choosing the appropriate one depending on the current Kokkos backend. This API is designed to be safe to use, with as much compile-time and runtime checks as possible. Kokkos-FFT takes advantage of Kokkos to offer support for multi dimensional arrays with up to 8 dimensions. It also supports real to real, real to complex and complex to real transformations. Kokkos-FFT is thoroughly tested and offers an extensive documentation, it is also easy to build, most of the compilation flags coming directly from the Kokkos installation you link against. It recently reached an important milestone with the release of the 1.0 version, making it production ready. This presentation will serve to: - showcase the API, its ease of use, and the way it integrates in a Kokkos project, - present performances results, - present ongoing work regarding the distribution of FFT across several nodes using MPI and other collective communication libraries.

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-03-02
DOI
https://doi.org/10.5281/zenodo.18833648
Primary Topic
Parallel Computing and Optimization Techniques
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

Kokkos-FFT

Yuuichi Asahi, Paul Gannay, The CExA team
Zenodo (CERN European Organization for Nuclear Research)
Parallel Computing and Optimization Techniques
article

Kokkos-FFT

Yuuichi Asahi, Paul Gannay, The CExA team
article en

Abstract

Kokkos-FFT is a library which started development in December 2023, as a part of the CExA project, and is now an official part of the Kokkos project (and thus of HPSF). Its initial goals were to offer access to a Kokkos-aware performance-driven and portable FFT, with an API as simple as possible. It has been done by building an API inspired by NumPy FFT, which dispatch work to existing FFT libs (such as cufft, hipfft, FFTW, or onemkl), choosing the appropriate one depending on the current Kokkos backend. This API is designed to be safe to use, with as much compile-time and runtime checks as possible. Kokkos-FFT takes advantage of Kokkos to offer support for multi dimensional arrays with up to 8 dimensions. It also supports real to real, real to complex and complex to real transformations. Kokkos-FFT is thoroughly tested and offers an extensive documentation, it is also easy to build, most of the compilation flags coming directly from the Kokkos installation you link against. It recently reached an important milestone with the release of the 1.0 version, making it production ready. This presentation will serve to: - showcase the API, its ease of use, and the way it integrates in a Kokkos project, - present performances results, - present ongoing work regarding the distribution of FFT across several nodes using MPI and other collective communication libraries.

Zenodo (CERN European Organization for Nuclear Research)
Commissariat à l'Énergie Atomique et aux Énergies Alternatives (FR), CEA Paris-Saclay - Etablissement de Saclay (FR), CEA Paris-Saclay (FR)
Industry, innovation and infrastructure
Openalex Percentile: Top 53%
Parallel Computing and Optimization Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.