CFD Utilities Files

PURPOSE:

   SMOOTH2D is a driver for 2-D interpolation/smoothing routines. It
   applies the selected methods to one or more datasets read from a
   3-column-type file.  The dataset(s) may or may not be rectangular
   but some of the methods assume rectangularity and are expected to
   be used appropriately.  Some also just require pseudorectangularity.

   SMOOTH2D evaluates the fit at points which may be generated or read
   from a separate file (in a similar format except only two columns
   are needed - any additional columns are ignored).

   Results are written to a file in the same form as the original data.

   If (X,Y)s are generated, one slice or more in each direction is
   provided for, and the rectangular grid of interpolated data may be
   ordered in the output file by either rows or columns to assist 1-D
   plotting of slices.  This format is also suited to the 2-D graphics
   utility FMAP.

   Dataset format:           <Reads are list-directed>

   Title                     <Up to 80 characters>
   NX  NY   or   NPTS        <2 integers = (pseudo)rectangular; 1 = random>
   X(1)     Y(1)     F(1)    <If (pseudo)rectangular, the table may be
   X(2)     Y(2)     F(2)     entered either by columns or by rows.>
     :        :        :
     :        :        :
   X(N)     Y(N)     F(N)    <where N is either NX * NY or NPTS>
   NX  NY   or   NPTS        <repeat for further datasets in same file>
     :        :        :

   An optional format for binary, pseudorectangular datasets is indicated
   by the following code segment (basically all Xs by rows or columns, then
   all Ys the same way, and all in one unformatted record).  The
   corresponding function values are read similarly from a separate file:

   READ (LUNGRD, IOSTAT=IOS) NX, NY
   NPTS = NX * NY
   READ (LUNGRD, IOSTAT=IOS) (X(I), I=1,NPTS), (Y(I), I=1,NPTS)

   READ (LUNFUN, IOSTAT=IOS) NX, NY
   READ (LUNFUN, IOSTAT=IOS) (F(I), I=1,NPTS)

METHOD:

   Open file for normal evaluated results (smooth2d.out).
   Prompt for and open the input data file; read its title.
   Prompt for descriptive text to go to results file (default = title).
   Prompt for generating (X,Y)s for evaluation, or reading them.
   IF <read points> THEN
      Prompt for the evaluation file.  ! Same for all datasets.
      Open and read it.
   END IF
   Prompt for scaling the data so that distances between pts. are
   meaningful.
   PICK_METHOD:
      Select a 2-D method.
      IF <CR> or ^Z GO TO <DONE>
      Rewind the input data file; skip its title.
      GET_DATASET:
         Try to read a dataset (formatted|unformatted; pseudorect. or not)
         IF <EOF> GO TO PICK_METHOD.
         Determine data range (partly for scaling, partly for grid gen.).

         IF <generate points> THEN
            IF <first dataset and first method> THEN
               Display data range.
               Prompt for low, high, and increment in X and in Y.
            END IF
            Generate rectangular evaluation mesh (possibly just a slice),
            ordered by rows or columns according to a first-time prompt.
         END IF

         IF <first dataset> THEN
            Prompt for smoothing parameters (if applicable).
         END IF
         Fit requested surface.                |  These two steps may be
         Evaluate fit at requested points.     |  combined by some methods.
         Write results to output file.
      GO TO GET_DATASET.
   <DONE>

HISTORY:
  DAS   02/12/86   Initial adaptation from SMOOTH (Akima, Hardy,
                   and TPSPLN).
  DAS   02/14/86   Developed FACETS and installed it, after above
                   methods seemed to behave dismally.
  DAS   02/19/86   Allowed for <multiple methods/one dataset> in one
                   run; added DETAILS for self-descriptive output.
  DAS   03/14/86   Provided for generating grid of evaluation points.
  DAS   03/21/86   Provided for scaling the Xs to match the Ys so that
                   the "distances" used in some methods are meaningful.
  DAS   09/19/86   MXNSQR limits certain methods to fewer data points
                   than others, since some involve solution of N*N dense
                   systems.
  DAS   01/20/88   Installed BIMOND3=PBHMD/PBHEV along with RDTABLE
                   derived from part of BIMOND3).  Reading of regular
                   or scattered data (NX and NY, or just NPTS resp. on
                   line 2) is automated using ERR= keyword.  Also R1MACH
                   (which came with BIMOND3 and is called by PBHMD) is
                   now called by SMOOTH2D because it needs "BIG" for its
                   own purposes (data range finding).
  DAS   03/03/88   ERR= with list-directed reads isn't good enough:
                   looking for NX and NY where just NPTS is present
                   means NY is set to (the truncated value of) X(1).
                   Switched to counting tokens instead.
  DAS   06/12/89   Provided for multiple datasets in one input file.
  DAS   03/09/90   MXNSQR raised reluctantly from 100 to 300 for one user.
  DAS   03/20/90   The Akima method was not saving its results when
                   some points (probably only a few) were not evaluated
                   properly (IER=999; extrapolation bug in the method).
                   Results are now saved anyway - bulk of them may be OK.
  DAS   08/21/90   Added INMESH/AKIMA combination for case of pseudo-
                   rectangular dataset.  Also provided unformatted option
                   for both the data file and the file of target (X,Y)s.
                   Output results may also be unformatted now, and have
                   either NXE, NYE, or just NEVAL if regular or not,
                   although the unformatted case assumes regularity.
                   Had to eliminate RDTABLE (for truly rectangular data)
                   in order to handle pseudorectangular datasets.  (Test
                   in-line for true rectangularity now.)  Raised MXDPTS,
                   MXIPTS to 201*101 for a wing surface application.
  DAS   04/09/91   Added LCSINT2D method for data consisting of rows
                   with constant Y along each row.
   "    04/18/91   LCSINT2D had independent edge definition arguments added
                   (but they are not made use of here).  It also needed
                   first/last indices for the evaluation arrays, and
                   different interpolation controls for "rows" & "columns"
                   and "leading" & "trailing" edges.
   "    11/28/93   Added LCSFIT2D.
   "    11/29/93   Added PLBICUBE.  Parametric cases are different enough
                   that some data options such as rescaling may not be
                   valid.
   "    05/23/94   Installed TOMS Algorithm 624 (REORDR, TRMESH, EVAL624).

AUTHOR:  David Saunders, NASA Ames/Sterling Software, Palo Alto, CA