d2ydx2= tspline(tension, y, x)  
  
       -or-     yp= tspline(tension, d2ydx2, y, x, xp)  
       -or-     yp= tspline(tension, y, x, xp)  
     computes a tensioned spline curve passing through the points (X, Y).  
  
     The first argument, TENSION, is a positive number which determines  
     the "tension" in the spline.  In a cubic spline, the second derivative  
     of the spline function varies linearly between the points X.  In the  
     tensioned spline, the curvature is concentrated near the points X,  
     falling off at a rate proportional to the tension.  Between the points  
     of X, the function varies as:  
           y= C1*exp(k*x) + C2*exp(-k*x) + C3*x + C4  
     The parameter k is proportional to the TENSION; for k->0, the function  
     reduces to the cubic spline (a piecewise cubic function), while for  
     k->infinity, the function reduces to the piecewise linear function  
     connecting the points.  The TENSION argument may either be a scalar  
     value, in which case, k will be TENSION*(numberof(X)-1)/(max(X)-min(X))  
     in every interval of X, or TENSION may be an array of length one less  
     than the length of X, in which case the parameter k will be  
     abs(TENSION/X(dif)), possibly varying from one interval to the next.  
     You can use a variable tension to flatten "bumps" in one interval  
     without affecting nearby intervals.  Internally, tspline forces  
     k*X(dif) to lie between 0.01 and 100.0 in every interval, independent  
     of the value of TENSION.  Typically, the most dramatic variation  
     occurs between TENSION of 1.0 and 10.0.  
  
     With three arguments, Y and X, spline returns the derivatives D2YDX2 at  
     the points, an array of the same length as X and Y.  The D2YDX2 values  
     are chosen so that the tensioned spline function returned by the five  
     argument call will have a continuous first derivative.  
  
     The X array must be strictly monotonic; it may either increase or  
     decrease.  
  
     The values Y and the derivatives D2YDX2 uniquely determine a tensioned  
     spline function, whose value is returned in the five argument form.  
     In this form, tspline is analogous to the piecewise linear interpolator  
     interp; usually you will regard it as a continuous function of its  
     fifth (or fourth) argument, XP.  
  
     The XP array may have any dimensionality; the result YP will have  
     the same dimensions as XP.  
  
     The D2YDX2 argument will normally have been computed by a previous call  
     to the three argument tspline function.  If you will be computing the  
     values of the spline function for many sets of XP, use this five  
     argument form.  
  
     If you only want the tspline evaluated at a single set of XP, use the  
     four argument form.  This is equivalent to:  
          yp= tspline(tension, tspline(tension,y,x), y, x, xp)  
  
     The keywords DYDX1 and DYDX0 can be used to set the values of the  
     returned DYDX(1) and DYDX(0) -- the first and last values of the  
     slope, respectively.  If either is not specified or nil, the slope at  
     that end will be chosen so that the second derivative is zero there.  
  
     The function tspline (tensioned spline) gives an interpolation  
     function which lies between spline and interp, at the cost of  
     requiring you to specify another parameter (the tension).  
  

unknown type function, documented at include/spline.i   line 195