1       SUBROUTINE ZLAQGB( M, N, KL, KU, AB, LDAB, R, C, ROWCND, COLCND,
  2      $                   AMAX, EQUED )
  3 *
  4 *  -- LAPACK auxiliary routine (version 3.2) --
  5 *  -- LAPACK is a software package provided by Univ. of Tennessee,    --
  6 *  -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
  7 *     November 2006
  8 *
  9 *     .. Scalar Arguments ..
 10       CHARACTER          EQUED
 11       INTEGER            KL, KU, LDAB, M, N
 12       DOUBLE PRECISION   AMAX, COLCND, ROWCND
 13 *     ..
 14 *     .. Array Arguments ..
 15       DOUBLE PRECISION   C( * ), R( * )
 16       COMPLEX*16         AB( LDAB, * )
 17 *     ..
 18 *
 19 *  Purpose
 20 *  =======
 21 *
 22 *  ZLAQGB equilibrates a general M by N band matrix A with KL
 23 *  subdiagonals and KU superdiagonals using the row and scaling factors
 24 *  in the vectors R and C.
 25 *
 26 *  Arguments
 27 *  =========
 28 *
 29 *  M       (input) INTEGER
 30 *          The number of rows of the matrix A.  M >= 0.
 31 *
 32 *  N       (input) INTEGER
 33 *          The number of columns of the matrix A.  N >= 0.
 34 *
 35 *  KL      (input) INTEGER
 36 *          The number of subdiagonals within the band of A.  KL >= 0.
 37 *
 38 *  KU      (input) INTEGER
 39 *          The number of superdiagonals within the band of A.  KU >= 0.
 40 *
 41 *  AB      (input/output) COMPLEX*16 array, dimension (LDAB,N)
 42 *          On entry, the matrix A in band storage, in rows 1 to KL+KU+1.
 43 *          The j-th column of A is stored in the j-th column of the
 44 *          array AB as follows:
 45 *          AB(ku+1+i-j,j) = A(i,j) for max(1,j-ku)<=i<=min(m,j+kl)
 46 *
 47 *          On exit, the equilibrated matrix, in the same storage format
 48 *          as A.  See EQUED for the form of the equilibrated matrix.
 49 *
 50 *  LDAB    (input) INTEGER
 51 *          The leading dimension of the array AB.  LDA >= KL+KU+1.
 52 *
 53 *  R       (input) DOUBLE PRECISION array, dimension (M)
 54 *          The row scale factors for A.
 55 *
 56 *  C       (input) DOUBLE PRECISION array, dimension (N)
 57 *          The column scale factors for A.
 58 *
 59 *  ROWCND  (input) DOUBLE PRECISION
 60 *          Ratio of the smallest R(i) to the largest R(i).
 61 *
 62 *  COLCND  (input) DOUBLE PRECISION
 63 *          Ratio of the smallest C(i) to the largest C(i).
 64 *
 65 *  AMAX    (input) DOUBLE PRECISION
 66 *          Absolute value of largest matrix entry.
 67 *
 68 *  EQUED   (output) CHARACTER*1
 69 *          Specifies the form of equilibration that was done.
 70 *          = 'N':  No equilibration
 71 *          = 'R':  Row equilibration, i.e., A has been premultiplied by
 72 *                  diag(R).
 73 *          = 'C':  Column equilibration, i.e., A has been postmultiplied
 74 *                  by diag(C).
 75 *          = 'B':  Both row and column equilibration, i.e., A has been
 76 *                  replaced by diag(R) * A * diag(C).
 77 *
 78 *  Internal Parameters
 79 *  ===================
 80 *
 81 *  THRESH is a threshold value used to decide if row or column scaling
 82 *  should be done based on the ratio of the row or column scaling
 83 *  factors.  If ROWCND < THRESH, row scaling is done, and if
 84 *  COLCND < THRESH, column scaling is done.
 85 *
 86 *  LARGE and SMALL are threshold values used to decide if row scaling
 87 *  should be done based on the absolute size of the largest matrix
 88 *  element.  If AMAX > LARGE or AMAX < SMALL, row scaling is done.
 89 *
 90 *  =====================================================================
 91 *
 92 *     .. Parameters ..
 93       DOUBLE PRECISION   ONE, THRESH
 94       PARAMETER          ( ONE = 1.0D+0, THRESH = 0.1D+0 )
 95 *     ..
 96 *     .. Local Scalars ..
 97       INTEGER            I, J
 98       DOUBLE PRECISION   CJ, LARGE, SMALL
 99 *     ..
100 *     .. External Functions ..
101       DOUBLE PRECISION   DLAMCH
102       EXTERNAL           DLAMCH
103 *     ..
104 *     .. Intrinsic Functions ..
105       INTRINSIC          MAXMIN
106 *     ..
107 *     .. Executable Statements ..
108 *
109 *     Quick return if possible
110 *
111       IF( M.LE.0 .OR. N.LE.0 ) THEN
112          EQUED = 'N'
113          RETURN
114       END IF
115 *
116 *     Initialize LARGE and SMALL.
117 *
118       SMALL = DLAMCH( 'Safe minimum' ) / DLAMCH( 'Precision' )
119       LARGE = ONE / SMALL
120 *
121       IF( ROWCND.GE.THRESH .AND. AMAX.GE.SMALL .AND. AMAX.LE.LARGE )
122      $     THEN
123 *
124 *        No row scaling
125 *
126          IF( COLCND.GE.THRESH ) THEN
127 *
128 *           No column scaling
129 *
130             EQUED = 'N'
131          ELSE
132 *
133 *           Column scaling
134 *
135             DO 20 J = 1, N
136                CJ = C( J )
137                DO 10 I = MAX1, J-KU ), MIN( M, J+KL )
138                   AB( KU+1+I-J, J ) = CJ*AB( KU+1+I-J, J )
139    10          CONTINUE
140    20       CONTINUE
141             EQUED = 'C'
142          END IF
143       ELSE IF( COLCND.GE.THRESH ) THEN
144 *
145 *        Row scaling, no column scaling
146 *
147          DO 40 J = 1, N
148             DO 30 I = MAX1, J-KU ), MIN( M, J+KL )
149                AB( KU+1+I-J, J ) = R( I )*AB( KU+1+I-J, J )
150    30       CONTINUE
151    40    CONTINUE
152          EQUED = 'R'
153       ELSE
154 *
155 *        Row and column scaling
156 *
157          DO 60 J = 1, N
158             CJ = C( J )
159             DO 50 I = MAX1, J-KU ), MIN( M, J+KL )
160                AB( KU+1+I-J, J ) = CJ*R( I )*AB( KU+1+I-J, J )
161    50       CONTINUE
162    60    CONTINUE
163          EQUED = 'B'
164       END IF
165 *
166       RETURN
167 *
168 *     End of ZLAQGB
169 *
170       END