fastsum - fast NFFT-based summation

  This library of C functions computes approximations of sums of the form

            N
            -
  f(y )  =  >  alpha   K(y - x )       (j=1,...,M)
     j      -       k     j   k
           k=1

  where x_k, y_j are arbitrary nodes in [-1/4+eps_B/2, 1/4-eps_B/2]^d, alpha_k
  are complex coefficients and K: R^d -> C is a kernel function which is C^oo
  except of the origin. Such kernels include log|x| and 1/|x|^b (b \in N) or
  the parameter dependent Hardy's multiquadric (x^2+c^2)^(1/2) and generalized
  multiquadrics (x^2+c^2)^(-b/2) (b \in N).

  It is based on the fast Fourier transform for nonequispaced nodes (NFFT) and
  therefore depends on the NFFT C-library. New kernels are easily incorporated
  by defining an appropriate C-function (see kernels.c for some examples).

  fastsum_test.c is an example for the usage of the library. The MATLAB script
  file fastsum_test.m calls the MATLAB function fastsum.m, which is a simple
  example for the usage in MATLAB.



  References:

  [1] D. Potts, and G. Steidl. Fast summation at nonequispaced knots by NFFTs.
      SIAM J. Sci. Comput., 24:2013-2037, 2003.

  [2] D. Potts, G. Steidl, and A. Nieslony. Fast convolution with radial
      kernels at nonequispaced knots. Numer. Math., 98:329-351, 2004.

  [3] M. Fenn, and G. Steidl. Fast NFFT-based summation of radial functions.
      Sampl. Theory Signal Image Process., 3:1-28, 2004.

  [4] M. Fenn, and D. Potts. Fast summation based on fast trigonometric
      transforms at non-equispaced nodes. Numer. Linear Algebra. Appl.,
      12:161-169, 2005.

