About Double Precision 

     Many Numerical Recipes readers have requested that the Numerical
     Recipes routines be made available in DOUBLE PRECISION, in
     addition to the default REAL. Doing so is not as simple as one might
     think, because of the distinction between "roundoff error" and
     "truncation error" (Please see the main book if you are not familiar
     with these concepts.).  While mechanically converting all REAL's to
     DOUBLE PRECISION's reduces a routine's roundoff error, it does not
     reduce any truncation error that may be intrinsic to the algorithm.
     Sometimes, a routine contains "accuracy parameters" which can be
     adjusted to reduce the truncation error to the new, desired level.
     Occasionally, however, the truncation error cannot be so easily
     reduced; then, a whole new algorithm is needed.  Clearly such new
     algorithms are beyond the scope of a simple "conversion".

     If, despite the above cautions, you want to proceed with converting
     some routines to DOUBLE PRECISION, you can use an automated conversion
     tool that is included in this distribution, in the "misc" subdirectory.

     There are two main files, nrtodp.c and nrtodp.dat.  You should compile
     the program nrtodp from nrtodp.c and install it (see the help 
     document nrtodp.compile). Then copy the file nrtodp.dat to your local
     library directory, preferably the same one that contains the other
     Numerical Recipes library files.  For a quick start with nrtodp,
     simply type nrtodp (with no arguments) at any Unix prompt.
     Instructions will be displayed.  Here are further details:

     The syntax of the nrtodp command is
         nrtodp path/ file [file...]
     The action of the command is to perform a pro-forma conversion of
     Numerical Recipes FORTRAN files to DOUBLE PRECISION.  Note that this
     routine works only on Numerical Recipes routines (Recipes and
     Example programs), not on general FORTRAN programs, because it assumes
     certain typographical and stylistic conventions that the Numerical
     Recipes routines strictly follow.

     The first argument following nrtodp is a path string that is used
     literally as the start of the output file name.  It should end with
     the character "/".

     The second and following arguments are input file names.  They must
     correspond to files in the current (working) directory.  Output
     file names are the same as input file names, but with a prepended
     character "d".

     Example:
          nrtodp /nr/double/ lubksb.f
     creates an output file /nr/double/dlubksb.f.  The subdirectory
     /nr/double/ must already exist.  For output to the current directory,
     use ./ for path.

     The data file nrtodp.dat must be present either in the library
     directory specified when compiling nrtodp, or the current directory.
     A copy in the current directory will override one in the library
     directory.

     You might wish to examine the file nrtodp.dat with a text editor.
     That file begins with comment lines that list the automatic
     substitutions made by the nrtodp program.  The rest of the file
     is a list of known exceptions and special cases, organized by Numerical
     Recipes routine name.  These special cases adjust the routines'
     accuracy parameters, when it is possible to do so.  If you want
     to make different substitutions, or if you discover errors in the
     conversions, you can edit the nrtodp.dat file yourself.  We would
     be grateful to hear from you regarding your corrections.

     NOTE: nrtodp is an UNSUPPORTED product, provided for your convenience.