VectorFFT: A permutation-free mixed-radix FFT library that beats Intel MKL and FFTW on every tested size. Hand-tuned AVX2/AVX-512 codelets for 18 radixes, zero-allocation in-place execution, and a wisdom-based planner. 1.1x–4.9x faster than Intel MKL, 1.3x–12.8x faster than FFTW. Pure C, no dependencies.
-
Updated
Apr 19, 2026 - C