Re: [boost] ffts & special functions

From: Neal D. Becker (nbecker_at_[hidden])
Date: 2002-09-19 07:07:24

>>>>> "Jason" == Jason D Schmidt <Jason> writes:

    Jason> Hi,
    Jason> I have some numerical code that might be in shape enough to think about
    Jason> having it reviewed.  Before I submit anything, I'd like to get an idea
    Jason> what people are looking for.  Here's what I have right now:

    Jason> ffts - transforms a std::valarray<T>; the size should be a power of 2;
    Jason> the size is a template parameter (for a good reason)

The interface should be generic.  Non power of 2 is much more
interesting.

    Jason> power spectral density, cross-spectral density, correlations, convolution
    Jason> - all simple algorithms built upon the fft; some are made faster through
    Jason> using tricks that do ffts at once or do a half-size real fft

I also have loads of this kind of stuff that I use daily.  I'd like to
see what you have if the interfaces are generic.

There are several kinds of genericity that are interesting in signal
processing applications.

1) generic container interfaces (this is easy)

2) In some cases it makes sense that algorithms can operate on either
   scalar or complex data

3) It is helpful if algorithms can operate on integer data (or
   complex<int>) as well as floating point.  This is sometimes
   feasible.