Changes between Version 383 and Version 384 of doc/app/initialization_parameters


Ignore:
Timestamp:
May 22, 2018 9:40:06 AM (7 years ago)
Author:
Giersch
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • doc/app/initialization_parameters

    v383 v384  
    8548541.:  The '''Temperton'''-method from Clive Temperton (ECMWF) which is computationally very fast and switched on with '''fft_method''' = '' 'temperton-algorithm'.'' The number of horizontal grid points ([#nx nx]+1, [#ny ny]+1) to be used with this method must be composed of prime factors 2, 3 and 5.\\\\
    8558552.:  The '''Singleton'''-method which is very slow but has no restrictions concerning the number of grid points to be used with, switched on with '''fft_method''' = '' 'singleton-algorithm'.''\\\\
    856 Furthermore, the fftw (see http://www.fftw.org ) can be used by setting '''fft_method''' = '' 'fftw' ''. This also requires to set the cpp-preprocessor switch {{{-D__fftw}}} in the {{{%cpp_options}}}-line of the configuration file, and to give the paths to the respective fftw include file and library by a {{{%fftw_inc}}} and {{{%fftw_lib}}} line, respectively. The fftw-routines are more than two times faster than the Temperton-fft.
     856Furthermore, the fftw (see http://www.fftw.org ) can be used by setting '''fft_method''' = '' 'fftw' ''. This also requires to set the cpp-preprocessor switch {{{-D__fftw}}} in the {{{%cpp_options}}}-line of the configuration file, and to give the paths to the respective fftw include file and library by a {{{%fftw_inc}}} and {{{%fftw_lib}}} line, respectively. The fftw-routines are more than two times faster than the Temperton-fft.\\
     857Finally, '''fft_method''' can also be set to '' 'system-specific' '' to account for fft routines from IBM or NEC. This also requires to set the cpp-preprocessor switch {{{-D__ibm}}} or {{{-D__nec}}} in the {{{%cpp_options}}}-line of the configuration file
    857858}}}
    858859|----------------