FFTW3 for new instruction sets

Link to (obsolete) GitHub repository.. The code is now available on a branch in the official FFTW3 repository.
Some single-thread, double-precision results on an Intel "Haswell" 4570S.
The results were generated with benchfft-3.1.

FFTW3/AVX2 1D Double Precision Complex

FFTW3/AVX2 1D Double Precision Complex (non power of 2)

FFTW3/AVX2 1D Double Precision Real

FFTW3/AVX2 1D Double Precision Real (non power of 2)

FFTW3/AVX2 2D Double Precision Complex

FFTW3/AVX2 2D Double Precision Complex (non power of 2)

FFTW3/AVX2 2D Double Precision Real

FFTW3/AVX2 2D Double Precision Real (non power of 2)

FFTW3/AVX2 3D Double Precision Complex

FFTW3/AVX2 3D Double Precision Complex (non power of 2)

FFTW3/AVX2 3D Double Precision Real

FFTW3/AVX2 3D Double Precision Real (non power of 2)

home

Last modified: Fri Mar 7 11:21:22 CET 2014