Re: "Discrete Hartley transform" vs " Discrete Fouuier transform"

On Oct 19, 10:41=A0am, "VelociChicken" <b...@[EMAIL PROTECTED]
> wrote:
> Your FFTW vested interests are starting to show through.

I'm not sure why you think FFTW gives me a vested interest on way or
the other on this.  We would be perfectly happy to use an FHT in FFTW
if it had any clear advantage.  The DHT is *already* one of the
transforms sup****ted by the FFTW API.  In fact, we do use a DHT to do
real-data transforms of prime sizes, although I later found out that
there is also a way to do this directly using a real-data FFT
(published by Burrus in the 1980s).

The problem is that proponents of FHTs on this thread have been
repeatedly making claims that are either false on their face, or
completely unsup****ted by any evidence and contradicted by 20 years of
attempts.

> Well, seeing as the FHT uses half the memory of FFT, then I guess it's
ab=
out
> time someone did a heavily optimised FHT don't you think?

This is utterly false, although it is a persistent myth. An FFT
algorithm specialized for real inputs has the same memory requirements
as an FHT.

Regards,
Steven G. Johnson