File:  [local] / rpl / lapack / lapack / dsbevd.f
Revision 1.2: download - view: text, annotated - select for diffs - revision graph
Wed Apr 21 13:45:24 2010 UTC (14 years, 1 month ago) by bertrand
Branches: MAIN
CVS tags: rpl-4_0_17, rpl-4_0_16, rpl-4_0_15, HEAD
En route pour la 4.0.15 !

    1:       SUBROUTINE DSBEVD( JOBZ, UPLO, N, KD, AB, LDAB, W, Z, LDZ, WORK,
    2:      $                   LWORK, IWORK, LIWORK, INFO )
    3: *
    4: *  -- LAPACK driver 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          JOBZ, UPLO
   11:       INTEGER            INFO, KD, LDAB, LDZ, LIWORK, LWORK, N
   12: *     ..
   13: *     .. Array Arguments ..
   14:       INTEGER            IWORK( * )
   15:       DOUBLE PRECISION   AB( LDAB, * ), W( * ), WORK( * ), Z( LDZ, * )
   16: *     ..
   17: *
   18: *  Purpose
   19: *  =======
   20: *
   21: *  DSBEVD computes all the eigenvalues and, optionally, eigenvectors of
   22: *  a real symmetric band matrix A. If eigenvectors are desired, it uses
   23: *  a divide and conquer algorithm.
   24: *
   25: *  The divide and conquer algorithm makes very mild assumptions about
   26: *  floating point arithmetic. It will work on machines with a guard
   27: *  digit in add/subtract, or on those binary machines without guard
   28: *  digits which subtract like the Cray X-MP, Cray Y-MP, Cray C-90, or
   29: *  Cray-2. It could conceivably fail on hexadecimal or decimal machines
   30: *  without guard digits, but we know of none.
   31: *
   32: *  Arguments
   33: *  =========
   34: *
   35: *  JOBZ    (input) CHARACTER*1
   36: *          = 'N':  Compute eigenvalues only;
   37: *          = 'V':  Compute eigenvalues and eigenvectors.
   38: *
   39: *  UPLO    (input) CHARACTER*1
   40: *          = 'U':  Upper triangle of A is stored;
   41: *          = 'L':  Lower triangle of A is stored.
   42: *
   43: *  N       (input) INTEGER
   44: *          The order of the matrix A.  N >= 0.
   45: *
   46: *  KD      (input) INTEGER
   47: *          The number of superdiagonals of the matrix A if UPLO = 'U',
   48: *          or the number of subdiagonals if UPLO = 'L'.  KD >= 0.
   49: *
   50: *  AB      (input/output) DOUBLE PRECISION array, dimension (LDAB, N)
   51: *          On entry, the upper or lower triangle of the symmetric band
   52: *          matrix A, stored in the first KD+1 rows of the array.  The
   53: *          j-th column of A is stored in the j-th column of the array AB
   54: *          as follows:
   55: *          if UPLO = 'U', AB(kd+1+i-j,j) = A(i,j) for max(1,j-kd)<=i<=j;
   56: *          if UPLO = 'L', AB(1+i-j,j)    = A(i,j) for j<=i<=min(n,j+kd).
   57: *
   58: *          On exit, AB is overwritten by values generated during the
   59: *          reduction to tridiagonal form.  If UPLO = 'U', the first
   60: *          superdiagonal and the diagonal of the tridiagonal matrix T
   61: *          are returned in rows KD and KD+1 of AB, and if UPLO = 'L',
   62: *          the diagonal and first subdiagonal of T are returned in the
   63: *          first two rows of AB.
   64: *
   65: *  LDAB    (input) INTEGER
   66: *          The leading dimension of the array AB.  LDAB >= KD + 1.
   67: *
   68: *  W       (output) DOUBLE PRECISION array, dimension (N)
   69: *          If INFO = 0, the eigenvalues in ascending order.
   70: *
   71: *  Z       (output) DOUBLE PRECISION array, dimension (LDZ, N)
   72: *          If JOBZ = 'V', then if INFO = 0, Z contains the orthonormal
   73: *          eigenvectors of the matrix A, with the i-th column of Z
   74: *          holding the eigenvector associated with W(i).
   75: *          If JOBZ = 'N', then Z is not referenced.
   76: *
   77: *  LDZ     (input) INTEGER
   78: *          The leading dimension of the array Z.  LDZ >= 1, and if
   79: *          JOBZ = 'V', LDZ >= max(1,N).
   80: *
   81: *  WORK    (workspace/output) DOUBLE PRECISION array,
   82: *                                         dimension (LWORK)
   83: *          On exit, if INFO = 0, WORK(1) returns the optimal LWORK.
   84: *
   85: *  LWORK   (input) INTEGER
   86: *          The dimension of the array WORK.
   87: *          IF N <= 1,                LWORK must be at least 1.
   88: *          If JOBZ  = 'N' and N > 2, LWORK must be at least 2*N.
   89: *          If JOBZ  = 'V' and N > 2, LWORK must be at least
   90: *                         ( 1 + 5*N + 2*N**2 ).
   91: *
   92: *          If LWORK = -1, then a workspace query is assumed; the routine
   93: *          only calculates the optimal sizes of the WORK and IWORK
   94: *          arrays, returns these values as the first entries of the WORK
   95: *          and IWORK arrays, and no error message related to LWORK or
   96: *          LIWORK is issued by XERBLA.
   97: *
   98: *  IWORK   (workspace/output) INTEGER array, dimension (MAX(1,LIWORK))
   99: *          On exit, if INFO = 0, IWORK(1) returns the optimal LIWORK.
  100: *
  101: *  LIWORK  (input) INTEGER
  102: *          The dimension of the array LIWORK.
  103: *          If JOBZ  = 'N' or N <= 1, LIWORK must be at least 1.
  104: *          If JOBZ  = 'V' and N > 2, LIWORK must be at least 3 + 5*N.
  105: *
  106: *          If LIWORK = -1, then a workspace query is assumed; the
  107: *          routine only calculates the optimal sizes of the WORK and
  108: *          IWORK arrays, returns these values as the first entries of
  109: *          the WORK and IWORK arrays, and no error message related to
  110: *          LWORK or LIWORK is issued by XERBLA.
  111: *
  112: *  INFO    (output) INTEGER
  113: *          = 0:  successful exit
  114: *          < 0:  if INFO = -i, the i-th argument had an illegal value
  115: *          > 0:  if INFO = i, the algorithm failed to converge; i
  116: *                off-diagonal elements of an intermediate tridiagonal
  117: *                form did not converge to zero.
  118: *
  119: *  =====================================================================
  120: *
  121: *     .. Parameters ..
  122:       DOUBLE PRECISION   ZERO, ONE
  123:       PARAMETER          ( ZERO = 0.0D+0, ONE = 1.0D+0 )
  124: *     ..
  125: *     .. Local Scalars ..
  126:       LOGICAL            LOWER, LQUERY, WANTZ
  127:       INTEGER            IINFO, INDE, INDWK2, INDWRK, ISCALE, LIWMIN,
  128:      $                   LLWRK2, LWMIN
  129:       DOUBLE PRECISION   ANRM, BIGNUM, EPS, RMAX, RMIN, SAFMIN, SIGMA,
  130:      $                   SMLNUM
  131: *     ..
  132: *     .. External Functions ..
  133:       LOGICAL            LSAME
  134:       DOUBLE PRECISION   DLAMCH, DLANSB
  135:       EXTERNAL           LSAME, DLAMCH, DLANSB
  136: *     ..
  137: *     .. External Subroutines ..
  138:       EXTERNAL           DGEMM, DLACPY, DLASCL, DSBTRD, DSCAL, DSTEDC,
  139:      $                   DSTERF, XERBLA
  140: *     ..
  141: *     .. Intrinsic Functions ..
  142:       INTRINSIC          SQRT
  143: *     ..
  144: *     .. Executable Statements ..
  145: *
  146: *     Test the input parameters.
  147: *
  148:       WANTZ = LSAME( JOBZ, 'V' )
  149:       LOWER = LSAME( UPLO, 'L' )
  150:       LQUERY = ( LWORK.EQ.-1 .OR. LIWORK.EQ.-1 )
  151: *
  152:       INFO = 0
  153:       IF( N.LE.1 ) THEN
  154:          LIWMIN = 1
  155:          LWMIN = 1
  156:       ELSE
  157:          IF( WANTZ ) THEN
  158:             LIWMIN = 3 + 5*N
  159:             LWMIN = 1 + 5*N + 2*N**2
  160:          ELSE
  161:             LIWMIN = 1
  162:             LWMIN = 2*N
  163:          END IF
  164:       END IF
  165:       IF( .NOT.( WANTZ .OR. LSAME( JOBZ, 'N' ) ) ) THEN
  166:          INFO = -1
  167:       ELSE IF( .NOT.( LOWER .OR. LSAME( UPLO, 'U' ) ) ) THEN
  168:          INFO = -2
  169:       ELSE IF( N.LT.0 ) THEN
  170:          INFO = -3
  171:       ELSE IF( KD.LT.0 ) THEN
  172:          INFO = -4
  173:       ELSE IF( LDAB.LT.KD+1 ) THEN
  174:          INFO = -6
  175:       ELSE IF( LDZ.LT.1 .OR. ( WANTZ .AND. LDZ.LT.N ) ) THEN
  176:          INFO = -9
  177:       END IF
  178: *
  179:       IF( INFO.EQ.0 ) THEN
  180:          WORK( 1 ) = LWMIN
  181:          IWORK( 1 ) = LIWMIN
  182: *
  183:          IF( LWORK.LT.LWMIN .AND. .NOT.LQUERY ) THEN
  184:             INFO = -11
  185:          ELSE IF( LIWORK.LT.LIWMIN .AND. .NOT.LQUERY ) THEN
  186:             INFO = -13
  187:          END IF
  188:       END IF
  189: *
  190:       IF( INFO.NE.0 ) THEN
  191:          CALL XERBLA( 'DSBEVD', -INFO )
  192:          RETURN
  193:       ELSE IF( LQUERY ) THEN
  194:          RETURN
  195:       END IF
  196: *
  197: *     Quick return if possible
  198: *
  199:       IF( N.EQ.0 )
  200:      $   RETURN
  201: *
  202:       IF( N.EQ.1 ) THEN
  203:          W( 1 ) = AB( 1, 1 )
  204:          IF( WANTZ )
  205:      $      Z( 1, 1 ) = ONE
  206:          RETURN
  207:       END IF
  208: *
  209: *     Get machine constants.
  210: *
  211:       SAFMIN = DLAMCH( 'Safe minimum' )
  212:       EPS = DLAMCH( 'Precision' )
  213:       SMLNUM = SAFMIN / EPS
  214:       BIGNUM = ONE / SMLNUM
  215:       RMIN = SQRT( SMLNUM )
  216:       RMAX = SQRT( BIGNUM )
  217: *
  218: *     Scale matrix to allowable range, if necessary.
  219: *
  220:       ANRM = DLANSB( 'M', UPLO, N, KD, AB, LDAB, WORK )
  221:       ISCALE = 0
  222:       IF( ANRM.GT.ZERO .AND. ANRM.LT.RMIN ) THEN
  223:          ISCALE = 1
  224:          SIGMA = RMIN / ANRM
  225:       ELSE IF( ANRM.GT.RMAX ) THEN
  226:          ISCALE = 1
  227:          SIGMA = RMAX / ANRM
  228:       END IF
  229:       IF( ISCALE.EQ.1 ) THEN
  230:          IF( LOWER ) THEN
  231:             CALL DLASCL( 'B', KD, KD, ONE, SIGMA, N, N, AB, LDAB, INFO )
  232:          ELSE
  233:             CALL DLASCL( 'Q', KD, KD, ONE, SIGMA, N, N, AB, LDAB, INFO )
  234:          END IF
  235:       END IF
  236: *
  237: *     Call DSBTRD to reduce symmetric band matrix to tridiagonal form.
  238: *
  239:       INDE = 1
  240:       INDWRK = INDE + N
  241:       INDWK2 = INDWRK + N*N
  242:       LLWRK2 = LWORK - INDWK2 + 1
  243:       CALL DSBTRD( JOBZ, UPLO, N, KD, AB, LDAB, W, WORK( INDE ), Z, LDZ,
  244:      $             WORK( INDWRK ), IINFO )
  245: *
  246: *     For eigenvalues only, call DSTERF.  For eigenvectors, call SSTEDC.
  247: *
  248:       IF( .NOT.WANTZ ) THEN
  249:          CALL DSTERF( N, W, WORK( INDE ), INFO )
  250:       ELSE
  251:          CALL DSTEDC( 'I', N, W, WORK( INDE ), WORK( INDWRK ), N,
  252:      $                WORK( INDWK2 ), LLWRK2, IWORK, LIWORK, INFO )
  253:          CALL DGEMM( 'N', 'N', N, N, N, ONE, Z, LDZ, WORK( INDWRK ), N,
  254:      $               ZERO, WORK( INDWK2 ), N )
  255:          CALL DLACPY( 'A', N, N, WORK( INDWK2 ), N, Z, LDZ )
  256:       END IF
  257: *
  258: *     If matrix was scaled, then rescale eigenvalues appropriately.
  259: *
  260:       IF( ISCALE.EQ.1 )
  261:      $   CALL DSCAL( N, ONE / SIGMA, W, 1 )
  262: *
  263:       WORK( 1 ) = LWMIN
  264:       IWORK( 1 ) = LIWMIN
  265:       RETURN
  266: *
  267: *     End of DSBEVD
  268: *
  269:       END

CVSweb interface <joel.bertrand@systella.fr>