Hi Anton, The problem is not so much to implement FFT by itself. We implemented a wrapper to FFTW (the Fastest Fourier Transform in the West) since some time already. The problem is what type of object (time series...) you want to send to these routines. The simplest way is to send just arrays of doubles or complex doubles. But for this, one just needs to use what already exists (CERNLIB, fftw,...) in C or Fortran. On the other hand, each application has a given type of objects (vectors) on which to do an fft. So it seems to me the problem is not only fft but all signal processing. We have to do a consistent set which will include some type of generic vectors, routines, etc..., keeping in mind that every user will want to adapt it to his own types. Furthermore, the more you put slices, the slower you are (user time series -> root time series -> fft -> root frequency series -> user frequency series). Well, it's not as trivial as it seems. Any ideas ? Damir FOKIN@tsl.uu.se wrote: > > I think FFT is one of the most standard and "must be" tools apart from > fitting facility. There are many implementations, free or commercial. > I think one should be in some CERN lib or other lib spread around CERN. > NAG C lib is a commercial package as far as I know. > > I think we need FFT in the standard ROOT package, i.e. implemented and > supported. Could someone think about it? > > Regards, > Anton -- ===================================================================== | Damir Buskulic | Universite de Savoie/LAPP | | | Chemin de Bellevue, B.P. 110 | | Tel : +33 (0)450091600 | F-74941 Annecy-le-Vieux Cedex | | e-mail: buskulic@lapp.in2p3.fr | FRANCE | ===================================================================== mailto:buskulic@lapp.in2p3.fr
This archive was generated by hypermail 2b29 : Tue Jan 02 2001 - 11:50:36 MET