SLATEC Routines --- DBSKNU ---

*DECK DBSKNU
      SUBROUTINE DBSKNU (X, FNU, KODE, N, Y, NZ)
C***BEGIN PROLOGUE  DBSKNU
C***SUBSIDIARY
C***PURPOSE  Subsidiary to DBESK
C***LIBRARY   SLATEC
C***TYPE      DOUBLE PRECISION (BESKNU-S, DBSKNU-D)
C***AUTHOR  Amos, D. E., (SNLA)
C***DESCRIPTION
C
C     Abstract  **** A DOUBLE PRECISION routine ****
C         DBSKNU computes N member sequences of K Bessel functions
C         K/SUB(FNU+I-1)/(X), I=1,N for non-negative orders FNU and
C         positive X. Equations of the references are implemented on
C         small orders DNU for K/SUB(DNU)/(X) and K/SUB(DNU+1)/(X).
C         Forward recursion with the three term recursion relation
C         generates higher orders FNU+I-1, I=1,...,N. The parameter
C         KODE permits K/SUB(FNU+I-1)/(X) values or scaled values
C         EXP(X)*K/SUB(FNU+I-1)/(X), I=1,N to be returned.
C
C         To start the recursion FNU is normalized to the interval
C         -0.5.LE.DNU.LT.0.5. A special form of the power series is
C         implemented on 0.LT.X.LE.X1 while the Miller algorithm for the
C         K Bessel function in terms of the confluent hypergeometric
C         function U(FNU+0.5,2*FNU+1,X) is implemented on X1.LT.X.LE.X2.
C         For X.GT.X2, the asymptotic expansion for large X is used.
C         When FNU is a half odd integer, a special formula for
C         DNU=-0.5 and DNU+1.0=0.5 is used to start the recursion.
C
C         The maximum number of significant digits obtainable
C         is the smaller of 14 and the number of digits carried in
C         DOUBLE PRECISION arithmetic.
C
C         DBSKNU assumes that a significant digit SINH function is
C         available.
C
C     Description of Arguments
C
C         INPUT      X,FNU are DOUBLE PRECISION
C           X      - X.GT.0.0D0
C           FNU    - Order of initial K function, FNU.GE.0.0D0
C           N      - Number of members of the sequence, N.GE.1
C           KODE   - A parameter to indicate the scaling option
C                    KODE= 1  returns
C                             Y(I)=       K/SUB(FNU+I-1)/(X)
C                                  I=1,...,N
C                        = 2  returns
C                             Y(I)=EXP(X)*K/SUB(FNU+I-1)/(X)
C                                  I=1,...,N
C
C         OUTPUT     Y is DOUBLE PRECISION
C           Y      - A vector whose first N components contain values
C                    for the sequence
C                    Y(I)=       K/SUB(FNU+I-1)/(X), I=1,...,N or
C                    Y(I)=EXP(X)*K/SUB(FNU+I-1)/(X), I=1,...,N
C                    depending on KODE
C           NZ     - Number of components set to zero due to
C                    underflow,
C                    NZ= 0   , normal return
C                    NZ.NE.0 , first NZ components of Y set to zero
C                              due to underflow, Y(I)=0.0D0,I=1,...,NZ
C
C     Error Conditions
C         Improper input arguments - a fatal error
C         Overflow - a fatal error
C         Underflow with KODE=1 - a non-fatal error (NZ.NE.0)
C
C***SEE ALSO  DBESK
C***REFERENCES  N. M. Temme, On the numerical evaluation of the modified
C                 Bessel function of the third kind, Journal of
C                 Computational Physics 19, (1975), pp. 324-337.
C***ROUTINES CALLED  D1MACH, DGAMMA, I1MACH, XERMSG
C***REVISION HISTORY  (YYMMDD)
C   790201  DATE WRITTEN
C   890531  Changed all specific intrinsics to generic.  (WRB)
C   890911  Removed unnecessary intrinsics.  (WRB)
C   891214  Prologue converted to Version 4.0 format.  (BAB)
C   900315  CALLs to XERROR changed to CALLs to XERMSG.  (THJ)
C   900326  Removed duplicate information from DESCRIPTION section.
C           (WRB)
C   900328  Added TYPE section.  (WRB)
C   900727  Added EXTERNAL statement.  (WRB)
C   910408  Updated the AUTHOR and REFERENCES sections.  (WRB)
C   920501  Reformatted the REFERENCES section.  (WRB)
C***END PROLOGUE  DBSKNU