I think Matlab caches precomputed plans (FFTW "wisdom") for power-of-two sizes, maybe that is making the difference here. Also, for real arrays Matlab might be using FFTW's real-data FFT by default. Try it for a complex input array in both cases, and try it with and without precomputed plans in Julia.
- [julia-users] matlab vs julia slower fft baillot maxime
- [julia-users] Re: matlab vs julia slower fft Steven G. Johnson
- [julia-users] Re: matlab vs julia slower fft baillot maxime
- [julia-users] Re: matlab vs julia slower ff... baillot maxime
- [julia-users] Re: matlab vs julia slowe... Steven G. Johnson
- [julia-users] Re: matlab vs julia slowe... Steven G. Johnson
- [julia-users] Re: matlab vs julia ... baillot maxime
