Optimization of conjugate-pair split-radix FFT algorithm for SIMD platforms

Fast Fourier Transform represents a canonical example of DSP algorithm since it is in some form present in virtually every signal processing application. While most DSP platforms, based on the classical Harvard architecture, tend to prefer fixed-radix FFT algorithms due to their regular structure, general purpose platforms may benefit more from advanced FFT implementations, such as split-radix FFT. In this paper, we present an implementation of conjugate-pair variant of split-radix FFT, optimized for platforms with SIMD instruction set extension.