A few more tweaks, and add the file nfft.dat of efficient FFT lengths.

git-svn-id: svn+ssh://svn.code.sf.net/p/wsjt/wsjt/branches/wsjtx@4837 ab8295b8-cf94-4d9e-aec4-7959e3be5d79

lib/chkfft.txt

@@ -113,3 +113,8 @@ Test data for all transforms is gaussian random noise of zero mean and
 standard deviation 1.  Tabulated values of "rms" are the
 root-mean-square differences between the original data and the
 back-transfmred data.
+
+File nfft.dat contains all numbers between 2^3 and 2^23 that have
+no factor greater than 7, followed by their factors.  These numbers 
+are good choices for FFT lengths.
+

lib/four2a.f90

@@ -19,7 +19,7 @@ subroutine four2a(a,nfft,ndim,isign,iform)
 ! This version of four2a makes calls to the FFTW library to do the 
 ! actual computations.
 
-  parameter (NPMAX=100)                  !Max numberf of stored plans
+  parameter (NPMAX=2100)                 !Max numberf of stored plans
   parameter (NSMALL=16384)               !Max size of "small" FFTs
   complex a(nfft)                        !Array to be transformed
   complex aa(NSMALL)                     !Local copy of "small" a()