Kokkos-FFT key features

We are pleased to announce that Kokkos-FFT v1.0 has just been released, and has reached the production level. Let us recap its key features [1]:

• Integrated naturally in the Kokkos ecosystem on top of Kokkos Views and Kokkos ExecutionSpace.

• A simple interface like numpy.fft with in-place and out-of-place transforms:
  Only accepts Kokkos Views to make APIs simple and safe.

• 1D, 2D, 3D standard and real FFT functions (similar to numpy.fft) over 1D to 8D Kokkos Views:
  Batched plans are automatically used if View dimension is larger than FFT dimension.

• A reusable FFT plan which wraps the vendor libraries for various Kokkos backends: FFTW for Host Backends (Serial, OpenMP, and Threads), cuFFT for Cuda backend, hipFFT (rocFFT) for HIP backend, and oneMKL for SYCL backend are automatically enabled based on the enabled Kokkos backends.

• Compile-time and run-time checks for invalid usage (e.g. View extents mismatch).

Major changes in release v1.0

By releasing Kokkos-FFT v1.0, we consider KokkosFFT production-ready and the experimental warning has been suppressed. Here are the major changes:

• To align with Kokkos 5.0, we have set C++20 as a minimal requirement (minimum requirement of Kokkos is still 4.6, but recommended to use Kokkos 5.0 or later).

• We have improved the documentations.

• You can expect the APIs to remain stable; we follow semantic versioning and will not break current APIs under KokkosFFT namespace, until the next major release.

• We have created a Spack recipe to ease installation.

• You can join us in the #kokkos-fft channel for support on the Kokkos Slack Workspace.

Future developments

We are planning to add the following functionalities. Contributions to the project are highly welcomed (see developer guide).

• Multi-GPU support with MPI. We have started the integration of a prototype [2] into the Kokkos-FFT repository (see the thumbnail for the x component of vorticity plot from a distributed Taylor–Green vortex simulation).

• Device callable batched capability of FFTs like Kokkos Kernels

• Supporting callbacks if backend library supports that

References

[1] Y. Asahi, T. Padioleau, P. Zehner, J. Bigot and D Lebrun-Grandie, kokkos-fft: A shared-memory FFT for the Kokkos ecosystem, Journal of Open Source Software (JOSS), 10(111), 8391, https://doi.org/10.21105/joss.08391

[2] Yuuichi Asahi, Trévis Morvany, Thomas Padioleau, and Julien Bigot. 2025. Development of a performance portable distributed FFT interface on top of the Kokkos ecosystem. In Proceedings of the SC ’25 Workshops of the International Conference for High Performance Computing, Networking, Storage and Analysis (SC Workshops ’25). Association for Computing Machinery, New York, NY, USA, 1233–1242. doi:10.1145/3731599.3767494