Re: [Boost.GIL] GSoC 2020 - FFT implementation

From: Frank Mori Hess (fmh6jj_at_[hidden])
Date: 2020-02-06 03:45:09

On Tue, Feb 4, 2020 at 3:18 PM stefan via Boost <boost_at_[hidden]> wrote:
> Boost.GIL. Given the amount of work (and expertise !) that goes into
> existing third-party FFT libraries (such as FFTW: http://fftw.org/), I
> think it's far more useful to make sure Boost.GIL can readily use (or be
> used with) such libraries.

One thing that bothered me when I used FFTW years ago in a C++
program, is they don't care much about types as long as they have the
same memory layout.  The fftw_complex type changes its definition in
fftw3.h depending on whether complex.h is included before fftw3.h or
not.  This seems to be asking for trouble with strict aliasing rules
and the C++ one definition rule.