FFTxi

FFTxi is an open-source software library that provides efficient implementations of fast Fourier transforms for real-time and offline signal processing. It supports one-, two-, and three-dimensional FFTs, real-to-complex and complex-to-real transforms, windowing, power spectrum estimation, and batch processing. The design emphasizes portability, performance, and ease of integration into audio, scientific, and engineering workflows.

Development began in 2015 by a collaborative community of researchers and developers and was released under the MIT license. The project aims to offer a robust, portable alternative to vendor- and platform-specific FFT implementations, with a focus on predictable performance across a range of hardware platforms.

FFTxi features architecture that includes a C99 core API with a plan-based interface; hardware acceleration via SIMD backends (SSE/AVX on x86 and NEON on ARM); optional GPU backends using CUDA and OpenCL; and auto-tuning to select optimal algorithms for a given data size and hardware. It also provides multi-threading and streaming support.

In practice, FFTxi is used for spectral analysis, convolution, filter design, and real-time audio processing, as well as scientific computing tasks such as spectrum estimation and image processing. Its wrappers and bindings allow use from Python, C++, and MATLAB, enabling integration into existing workflows.

The project maintains documentation, example code, and a contributor guide, with a permissive license and an open governance model. It is used in academic research and by hobbyists, and continues to evolve with community contributions and performance-oriented optimizations.