[plt-scheme] fft

From: Bill Wood (william.wood3 at comcast.net)
Date: Thu Mar 16 17:35:09 EST 2006

On Thu, 2006-03-16 at 14:12 -0600, Corey Sweeney wrote:
   . . .
> I'm about to have a need soon to do fourier transformations.  My web
> searches havn't been finding much.  Does anyone know of anything that
> can do fft from scheme? (preferably plt)

The major question is, how performant a solution do you need?  If you're
doing proof-of-concept type things, high-level solutions are likely to
be fine.  For medium-scale performance constraints, something like FFTW
would be good.  If you need performance just short of DSP boards or
other hardware solutions, then custom software that can tune the number
of points and radix to get optimal performance is the way to go. If this
last is what you need, you may have to roll your own.  You might look up
the work of Robert Johnson and Jeremy Johnson on high-performance FFT
algorithms that are tuned to hardware characteristics.  Bob Johnson has
a patent on a hardware design for "mixed radix multidimensional FFTs";
he and Jeremy also did some of the fundamental math.

If you get into the math, you'll want to look at Richard Tolimieri,
Myoung An and Chao Lu, _Mathematics of Multidimensional Fourier
Transform Algorithms_, 2nd ed., Springer-Verlag New York, 1997. (ISBN
0-387-98260-4).

 -- Bill Wood






Posted on the users mailing list.