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

"Steven G. Johnson" <stevenj@[EMAIL PROTECTED]
> wrote in message 
news:3346d2d1-b620-48c0-9bd5-0ba6e8e8b9dd@[EMAIL PROTECTED]
 Oct 10, 1:15 pm, Eric Jacobsen <eric.jacob...@[EMAIL PROTECTED]
> wrote:
> Since processing power and memory are cheap these days everybody just
> plugs in FFTs. When computing or memory resources are hard to come
> by, though, the FHT can have some real advantages.
>That's an old myth, but it's not true.
>From a theoretical perspective, the best known FFT algorithms
>specialized for real inputs have slightly fewer operations than the
>best known FHT algorithms.  (Claims that FHTs save a factor of two
>over FFTs for real data are bogus, and have been widely debunked in
>the literature.)

Your FFTW vested interests are starting to show through.

>From a practical perspective, the structure of FHT algorithms is
>roughly similar to that of real-data FFT algorithms, so there don't
>seem to be any significant implementation advantages one way or the
>other. So, it comes down to who has optimized their code more, and
>there are far more heavily optimized FFT libraries readily available
>than FHT libraries.

Well, seeing as the FHT uses half the memory of FFT, then I guess it's
about 
time someone did a heavily optimised FHT don't you think? Although I
suspect 
people have, and it's not likely to get simply posted on the interweb is
it.