SLATEC Routines --- SSICS ---


*DECK SSICS
      SUBROUTINE SSICS (N, NELT, IA, JA, A, ISYM, NEL, IEL, JEL, EL, D,
     +   R, IWARN)
C***BEGIN PROLOGUE  SSICS
C***PURPOSE  Incompl. Cholesky Decomposition Preconditioner SLAP Set Up.
C            Routine to generate the Incomplete Cholesky decomposition,
C            L*D*L-trans, of a symmetric positive definite matrix, A,
C            which is stored in SLAP Column format.  The unit lower
C            triangular matrix L is stored by rows, and the inverse of
C            the diagonal matrix D is stored.
C***LIBRARY   SLATEC (SLAP)
C***CATEGORY  D2E
C***TYPE      SINGLE PRECISION (SSICS-S, DSICS-D)
C***KEYWORDS  INCOMPLETE CHOLESKY FACTORIZATION,
C             ITERATIVE PRECONDITION, LINEAR SYSTEM, SLAP SPARSE
C***AUTHOR  Greenbaum, Anne, (Courant Institute)
C           Seager, Mark K., (LLNL)
C             Lawrence Livermore National Laboratory
C             PO BOX 808, L-60
C             Livermore, CA 94550 (510) 423-3141
C             seager@llnl.gov
C***DESCRIPTION
C
C *Usage:
C     INTEGER N, NELT, IA(NELT), JA(NELT), ISYM
C     INTEGER NEL, IEL(NEL), JEL(NEL), IWARN
C     REAL    A(NELT), EL(NEL), D(N), R(N)
C
C     CALL SSICS( N, NELT, IA, JA, A, ISYM, NEL, IEL, JEL, EL, D, R,
C    $    IWARN )
C
C *Arguments:
C N      :IN       Integer.
C         Order of the Matrix.
C NELT   :IN       Integer.
C         Number of elements in arrays IA, JA, and A.
C IA     :INOUT    Integer IA(NELT).
C JA     :INOUT    Integer JA(NELT).
C A      :INOUT    Real A(NELT).
C         These arrays should hold the matrix A in the SLAP Column
C         format.  See "Description", below.
C ISYM   :IN       Integer.
C         Flag to indicate symmetric storage format.
C         If ISYM=0, all non-zero entries of the matrix are stored.
C         If ISYM=1, the matrix is symmetric, and only the lower
C         triangle of the matrix is stored.
C NEL    :OUT      Integer.
C         Number of non-zeros in the lower triangle of A.   Also
C         corresponds to the length of the IEL, JEL, EL arrays.
C IEL    :OUT      Integer IEL(NEL).
C JEL    :OUT      Integer JEL(NEL).
C EL     :OUT      Real     EL(NEL).
C         IEL, JEL, EL contain the unit lower triangular factor  of the
C         incomplete decomposition   of the A  matrix  stored  in  SLAP
C         Row format.   The Diagonal of   ones   *IS*   stored.     See
C         "Description", below for more details about the SLAP Row fmt.
C D      :OUT      Real D(N)
C         Upon return this array holds D(I) = 1./DIAG(A).
C R      :WORK     Real R(N).
C         Temporary real workspace needed for the factorization.
C IWARN  :OUT      Integer.
C         This is a warning variable and is zero if the IC factoriza-
C         tion goes well.  It is set to the row index corresponding to
C         the last zero pivot found.  See "Description", below.
C
C *Description
C       =================== S L A P Column format ==================
C       This routine  requires that  the matrix A  be stored in  the
C       SLAP Column format.  In this format the non-zeros are stored
C       counting down columns (except for  the diagonal entry, which
C       must appear first in each  "column")  and are stored  in the
C       real array A.  In other words, for each column in the matrix
C       put the diagonal entry in A.  Then put in the other non-zero
C       elements going down   the  column (except  the diagonal)  in
C       order.  The IA array holds the row  index for each non-zero.
C       The JA array holds the offsets into the IA, A arrays for the
C       beginning of   each    column.    That  is,    IA(JA(ICOL)),
C       A(JA(ICOL)) points to the beginning of the ICOL-th column in
C       IA and  A.  IA(JA(ICOL+1)-1),  A(JA(ICOL+1)-1) points to the
C       end  of   the ICOL-th  column.  Note   that  we  always have
C       JA(N+1) = NELT+1, where  N  is the number of columns in  the
C       matrix and  NELT   is the number of non-zeros in the matrix.
C
C       Here is an example of the  SLAP Column  storage format for a
C       5x5 Matrix (in the A and IA arrays '|'  denotes the end of a
C       column):
C
C           5x5 Matrix      SLAP Column format for 5x5 matrix on left.
C                              1  2  3    4  5    6  7    8    9 10 11
C       |11 12  0  0 15|   A: 11 21 51 | 22 12 | 33 53 | 44 | 55 15 35
C       |21 22  0  0  0|  IA:  1  2  5 |  2  1 |  3  5 |  4 |  5  1  3
C       | 0  0 33  0 35|  JA:  1  4  6    8  9   12
C       | 0  0  0 44  0|
C       |51  0 53  0 55|
C
C       ==================== S L A P Row format ====================
C
C       This routine requires  that the matrix A  be  stored  in the
C       SLAP  Row format.   In this format  the non-zeros are stored
C       counting across  rows (except for the diagonal  entry, which
C       must appear first in each "row") and  are stored in the real
C       array A.  In other words, for each row in the matrix put the
C       diagonal entry in  A.   Then   put  in the   other  non-zero
C       elements   going  across the  row (except   the diagonal) in
C       order.   The  JA array  holds   the column   index for  each
C       non-zero.   The IA  array holds the  offsets into  the JA, A
C       arrays  for   the   beginning  of   each  row.   That    is,
C       JA(IA(IROW)),  A(IA(IROW)) points  to  the beginning  of the
C       IROW-th row in JA and A.   JA(IA(IROW+1)-1), A(IA(IROW+1)-1)
C       points to the  end of the  IROW-th row.  Note that we always
C       have IA(N+1) =  NELT+1, where  N  is  the number of rows  in
C       the matrix  and NELT  is the  number   of  non-zeros in  the
C       matrix.
C
C       Here is an example of the SLAP Row storage format for a  5x5
C       Matrix (in the A and JA arrays '|' denotes the end of a row):
C
C           5x5 Matrix         SLAP Row format for 5x5 matrix on left.
C                              1  2  3    4  5    6  7    8    9 10 11
C       |11 12  0  0 15|   A: 11 12 15 | 22 21 | 33 35 | 44 | 55 51 53
C       |21 22  0  0  0|  JA:  1  2  5 |  2  1 |  3  5 |  4 |  5  1  3
C       | 0  0 33  0 35|  IA:  1  4  6    8  9   12
C       | 0  0  0 44  0|
C       |51  0 53  0 55|
C
C       With the SLAP  format some  of  the   "inner  loops" of this
C       routine should vectorize  on  machines with hardware support
C       for vector   gather/scatter  operations.  Your compiler  may
C       require a compiler directive to  convince it that  there are
C       no  implicit  vector  dependencies.  Compiler directives for
C       the Alliant    FX/Fortran and CRI   CFT/CFT77 compilers  are
C       supplied with the standard SLAP distribution.
C
C       The IC factorization does not always exist for SPD matrices.
C       In the event that a zero pivot is found it is set  to be 1.0
C       and the factorization proceeds.   The integer variable IWARN
C       is set to the last row where the Diagonal was fudged.  This
C       eventuality hardly ever occurs in practice.
C
C***SEE ALSO  SCG, SSICCG
C***REFERENCES  1. Gene Golub and Charles Van Loan, Matrix Computations,
C                  Johns Hopkins University Press, Baltimore, Maryland,
C                  1983.
C***ROUTINES CALLED  XERMSG
C***REVISION HISTORY  (YYMMDD)
C   871119  DATE WRITTEN
C   881213  Previous REVISION DATE
C   890915  Made changes requested at July 1989 CML Meeting.  (MKS)
C   890922  Numerous changes to prologue to make closer to SLATEC
C           standard.  (FNF)
C   890929  Numerous changes to reduce SP/DP differences.  (FNF)
C   900805  Changed XERRWV calls to calls to XERMSG.  (RWC)
C   910411  Prologue converted to Version 4.0 format.  (BAB)
C   920511  Added complete declaration section.  (WRB)
C   920929  Corrected format of reference.  (FNF)
C   930701  Updated CATEGORY section.  (FNF, WRB)
C***END PROLOGUE  SSICS