Annotation of rpl/lapack/lapack/dlasq1.f, revision 1.19

1.11      bertrand    1: *> \brief \b DLASQ1 computes the singular values of a real square bidiagonal matrix. Used by sbdsqr.
1.8       bertrand    2: *
                      3: *  =========== DOCUMENTATION ===========
                      4: *
1.16      bertrand    5: * Online html documentation available at
                      6: *            http://www.netlib.org/lapack/explore-html/
1.8       bertrand    7: *
                      8: *> \htmlonly
1.16      bertrand    9: *> Download DLASQ1 + dependencies
                     10: *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/dlasq1.f">
                     11: *> [TGZ]</a>
                     12: *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/dlasq1.f">
                     13: *> [ZIP]</a>
                     14: *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/dlasq1.f">
1.8       bertrand   15: *> [TXT]</a>
1.16      bertrand   16: *> \endhtmlonly
1.8       bertrand   17: *
                     18: *  Definition:
                     19: *  ===========
                     20: *
                     21: *       SUBROUTINE DLASQ1( N, D, E, WORK, INFO )
1.16      bertrand   22: *
1.8       bertrand   23: *       .. Scalar Arguments ..
                     24: *       INTEGER            INFO, N
                     25: *       ..
                     26: *       .. Array Arguments ..
                     27: *       DOUBLE PRECISION   D( * ), E( * ), WORK( * )
                     28: *       ..
1.16      bertrand   29: *
1.8       bertrand   30: *
                     31: *> \par Purpose:
                     32: *  =============
                     33: *>
                     34: *> \verbatim
                     35: *>
                     36: *> DLASQ1 computes the singular values of a real N-by-N bidiagonal
                     37: *> matrix with diagonal D and off-diagonal E. The singular values
                     38: *> are computed to high relative accuracy, in the absence of
                     39: *> denormalization, underflow and overflow. The algorithm was first
                     40: *> presented in
                     41: *>
                     42: *> "Accurate singular values and differential qd algorithms" by K. V.
                     43: *> Fernando and B. N. Parlett, Numer. Math., Vol-67, No. 2, pp. 191-230,
                     44: *> 1994,
                     45: *>
                     46: *> and the present implementation is described in "An implementation of
                     47: *> the dqds Algorithm (Positive Case)", LAPACK Working Note.
                     48: *> \endverbatim
                     49: *
                     50: *  Arguments:
                     51: *  ==========
                     52: *
                     53: *> \param[in] N
                     54: *> \verbatim
                     55: *>          N is INTEGER
                     56: *>        The number of rows and columns in the matrix. N >= 0.
                     57: *> \endverbatim
                     58: *>
                     59: *> \param[in,out] D
                     60: *> \verbatim
                     61: *>          D is DOUBLE PRECISION array, dimension (N)
                     62: *>        On entry, D contains the diagonal elements of the
                     63: *>        bidiagonal matrix whose SVD is desired. On normal exit,
                     64: *>        D contains the singular values in decreasing order.
                     65: *> \endverbatim
                     66: *>
                     67: *> \param[in,out] E
                     68: *> \verbatim
                     69: *>          E is DOUBLE PRECISION array, dimension (N)
                     70: *>        On entry, elements E(1:N-1) contain the off-diagonal elements
                     71: *>        of the bidiagonal matrix whose SVD is desired.
                     72: *>        On exit, E is overwritten.
                     73: *> \endverbatim
                     74: *>
                     75: *> \param[out] WORK
                     76: *> \verbatim
                     77: *>          WORK is DOUBLE PRECISION array, dimension (4*N)
                     78: *> \endverbatim
                     79: *>
                     80: *> \param[out] INFO
                     81: *> \verbatim
                     82: *>          INFO is INTEGER
                     83: *>        = 0: successful exit
                     84: *>        < 0: if INFO = -i, the i-th argument had an illegal value
                     85: *>        > 0: the algorithm failed
                     86: *>             = 1, a split was marked by a positive value in E
                     87: *>             = 2, current block of Z not diagonalized after 100*N
                     88: *>                  iterations (in inner while loop)  On exit D and E
                     89: *>                  represent a matrix with the same singular values
                     90: *>                  which the calling subroutine could use to finish the
                     91: *>                  computation, or even feed back into DLASQ1
1.16      bertrand   92: *>             = 3, termination criterion of outer while loop not met
1.8       bertrand   93: *>                  (program created more than N unreduced blocks)
                     94: *> \endverbatim
                     95: *
                     96: *  Authors:
                     97: *  ========
                     98: *
1.16      bertrand   99: *> \author Univ. of Tennessee
                    100: *> \author Univ. of California Berkeley
                    101: *> \author Univ. of Colorado Denver
                    102: *> \author NAG Ltd.
1.8       bertrand  103: *
                    104: *> \ingroup auxOTHERcomputational
1.1       bertrand  105: *
1.8       bertrand  106: *  =====================================================================
                    107:       SUBROUTINE DLASQ1( N, D, E, WORK, INFO )
1.1       bertrand  108: *
1.19    ! bertrand  109: *  -- LAPACK computational routine --
1.1       bertrand  110: *  -- LAPACK is a software package provided by Univ. of Tennessee,    --
                    111: *  -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
                    112: *
                    113: *     .. Scalar Arguments ..
                    114:       INTEGER            INFO, N
                    115: *     ..
                    116: *     .. Array Arguments ..
                    117:       DOUBLE PRECISION   D( * ), E( * ), WORK( * )
                    118: *     ..
                    119: *
                    120: *  =====================================================================
                    121: *
                    122: *     .. Parameters ..
                    123:       DOUBLE PRECISION   ZERO
                    124:       PARAMETER          ( ZERO = 0.0D0 )
                    125: *     ..
                    126: *     .. Local Scalars ..
                    127:       INTEGER            I, IINFO
                    128:       DOUBLE PRECISION   EPS, SCALE, SAFMIN, SIGMN, SIGMX
                    129: *     ..
                    130: *     .. External Subroutines ..
                    131:       EXTERNAL           DCOPY, DLAS2, DLASCL, DLASQ2, DLASRT, XERBLA
                    132: *     ..
                    133: *     .. External Functions ..
                    134:       DOUBLE PRECISION   DLAMCH
                    135:       EXTERNAL           DLAMCH
                    136: *     ..
                    137: *     .. Intrinsic Functions ..
                    138:       INTRINSIC          ABS, MAX, SQRT
                    139: *     ..
                    140: *     .. Executable Statements ..
                    141: *
                    142:       INFO = 0
                    143:       IF( N.LT.0 ) THEN
1.14      bertrand  144:          INFO = -1
1.1       bertrand  145:          CALL XERBLA( 'DLASQ1', -INFO )
                    146:          RETURN
                    147:       ELSE IF( N.EQ.0 ) THEN
                    148:          RETURN
                    149:       ELSE IF( N.EQ.1 ) THEN
                    150:          D( 1 ) = ABS( D( 1 ) )
                    151:          RETURN
                    152:       ELSE IF( N.EQ.2 ) THEN
                    153:          CALL DLAS2( D( 1 ), E( 1 ), D( 2 ), SIGMN, SIGMX )
                    154:          D( 1 ) = SIGMX
                    155:          D( 2 ) = SIGMN
                    156:          RETURN
                    157:       END IF
                    158: *
                    159: *     Estimate the largest singular value.
                    160: *
                    161:       SIGMX = ZERO
                    162:       DO 10 I = 1, N - 1
                    163:          D( I ) = ABS( D( I ) )
                    164:          SIGMX = MAX( SIGMX, ABS( E( I ) ) )
                    165:    10 CONTINUE
                    166:       D( N ) = ABS( D( N ) )
                    167: *
                    168: *     Early return if SIGMX is zero (matrix is already diagonal).
                    169: *
                    170:       IF( SIGMX.EQ.ZERO ) THEN
                    171:          CALL DLASRT( 'D', N, D, IINFO )
                    172:          RETURN
                    173:       END IF
                    174: *
                    175:       DO 20 I = 1, N
                    176:          SIGMX = MAX( SIGMX, D( I ) )
                    177:    20 CONTINUE
                    178: *
                    179: *     Copy D and E into WORK (in the Z format) and scale (squaring the
                    180: *     input data makes scaling by a power of the radix pointless).
                    181: *
                    182:       EPS = DLAMCH( 'Precision' )
                    183:       SAFMIN = DLAMCH( 'Safe minimum' )
                    184:       SCALE = SQRT( EPS / SAFMIN )
                    185:       CALL DCOPY( N, D, 1, WORK( 1 ), 2 )
                    186:       CALL DCOPY( N-1, E, 1, WORK( 2 ), 2 )
                    187:       CALL DLASCL( 'G', 0, 0, SIGMX, SCALE, 2*N-1, 1, WORK, 2*N-1,
                    188:      $             IINFO )
1.16      bertrand  189: *
1.1       bertrand  190: *     Compute the q's and e's.
                    191: *
                    192:       DO 30 I = 1, 2*N - 1
                    193:          WORK( I ) = WORK( I )**2
                    194:    30 CONTINUE
                    195:       WORK( 2*N ) = ZERO
                    196: *
                    197:       CALL DLASQ2( N, WORK, INFO )
                    198: *
                    199:       IF( INFO.EQ.0 ) THEN
                    200:          DO 40 I = 1, N
                    201:             D( I ) = SQRT( WORK( I ) )
                    202:    40    CONTINUE
                    203:          CALL DLASCL( 'G', 0, 0, SCALE, SIGMX, N, 1, D, N, IINFO )
1.8       bertrand  204:       ELSE IF( INFO.EQ.2 ) THEN
                    205: *
                    206: *     Maximum number of iterations exceeded.  Move data from WORK
                    207: *     into D and E so the calling subroutine can try to finish
                    208: *
                    209:          DO I = 1, N
                    210:             D( I ) = SQRT( WORK( 2*I-1 ) )
                    211:             E( I ) = SQRT( WORK( 2*I ) )
                    212:          END DO
                    213:          CALL DLASCL( 'G', 0, 0, SCALE, SIGMX, N, 1, D, N, IINFO )
                    214:          CALL DLASCL( 'G', 0, 0, SCALE, SIGMX, N, 1, E, N, IINFO )
1.1       bertrand  215:       END IF
                    216: *
                    217:       RETURN
                    218: *
                    219: *     End of DLASQ1
                    220: *
                    221:       END

CVSweb interface <joel.bertrand@systella.fr>