dlarrd.f man page

subroutine dlarrd (RANGE, ORDER, N, VL, VU, IL, IU, GERS, RELTOL, D, E, E2, PIVMIN, NSPLIT, ISPLIT, M, W, WERR, WL, WU, IBLOCK, INDEXW, WORK, IWORK, INFO)
DLARRD computes the eigenvalues of a symmetric tridiagonal matrix to suitable accuracy.

DLARRD computes the eigenvalues of a symmetric tridiagonal matrix to suitable accuracy.  

Purpose:

 DLARRD computes the eigenvalues of a symmetric tridiagonal
 matrix T to suitable accuracy. This is an auxiliary code to be
 called from DSTEMR.
 The user may ask for all eigenvalues, all eigenvalues
 in the half-open interval (VL, VU], or the IL-th through IU-th
 eigenvalues.

 To avoid overflow, the matrix must be scaled so that its
 largest element is no greater than overflow**(1/2) * underflow**(1/4) in absolute value, and for greatest
 accuracy, it should not be much smaller than that.

 See W. Kahan "Accurate Eigenvalues of a Symmetric Tridiagonal
 Matrix", Report CS41, Computer Science Dept., Stanford
 University, July 21, 1966.

Parameters:

RANGE

          RANGE is CHARACTER*1
          = 'A': ("All")   all eigenvalues will be found.
          = 'V': ("Value") all eigenvalues in the half-open interval
                           (VL, VU] will be found.
          = 'I': ("Index") the IL-th through IU-th eigenvalues (of the
                           entire matrix) will be found.

ORDER

          ORDER is CHARACTER*1
          = 'B': ("By Block") the eigenvalues will be grouped by
                              split-off block (see IBLOCK, ISPLIT) and
                              ordered from smallest to largest within
                              the block.
          = 'E': ("Entire matrix")
                              the eigenvalues for the entire matrix
                              will be ordered from smallest to
                              largest.

N

          N is INTEGER
          The order of the tridiagonal matrix T.  N >= 0.

VL

          VL is DOUBLE PRECISION

VU

          VU is DOUBLE PRECISION
          If RANGE='V', the lower and upper bounds of the interval to
          be searched for eigenvalues.  Eigenvalues less than or equal
          to VL, or greater than VU, will not be returned.  VL < VU.
          Not referenced if RANGE = 'A' or 'I'.

IL

          IL is INTEGER

IU

          IU is INTEGER
          If RANGE='I', the indices (in ascending order) of the
          smallest and largest eigenvalues to be returned.
          1 <= IL <= IU <= N, if N > 0; IL = 1 and IU = 0 if N = 0.
          Not referenced if RANGE = 'A' or 'V'.

GERS

          GERS is DOUBLE PRECISION array, dimension (2*N)
          The N Gerschgorin intervals (the i-th Gerschgorin interval
          is (GERS(2*i-1), GERS(2*i)).

RELTOL

          RELTOL is DOUBLE PRECISION
          The minimum relative width of an interval.  When an interval
          is narrower than RELTOL times the larger (in
          magnitude) endpoint, then it is considered to be
          sufficiently small, i.e., converged.  Note: this should
          always be at least radix*machine epsilon.

D

          D is DOUBLE PRECISION array, dimension (N)
          The n diagonal elements of the tridiagonal matrix T.

E

          E is DOUBLE PRECISION array, dimension (N-1)
          The (n-1) off-diagonal elements of the tridiagonal matrix T.

E2

          E2 is DOUBLE PRECISION array, dimension (N-1)
          The (n-1) squared off-diagonal elements of the tridiagonal matrix T.

PIVMIN

          PIVMIN is DOUBLE PRECISION
          The minimum pivot allowed in the Sturm sequence for T.

NSPLIT

          NSPLIT is INTEGER
          The number of diagonal blocks in the matrix T.
          1 <= NSPLIT <= N.

ISPLIT

          ISPLIT is INTEGER array, dimension (N)
          The splitting points, at which T breaks up into submatrices.
          The first submatrix consists of rows/columns 1 to ISPLIT(1),
          the second of rows/columns ISPLIT(1)+1 through ISPLIT(2),
          etc., and the NSPLIT-th consists of rows/columns
          ISPLIT(NSPLIT-1)+1 through ISPLIT(NSPLIT)=N.
          (Only the first NSPLIT elements will actually be used, but
          since the user cannot know a priori what value NSPLIT will
          have, N words must be reserved for ISPLIT.)

M

          M is INTEGER
          The actual number of eigenvalues found. 0 <= M <= N.
          (See also the description of INFO=2,3.)

W

          W is DOUBLE PRECISION array, dimension (N)
          On exit, the first M elements of W will contain the
          eigenvalue approximations. DLARRD computes an interval
          I_j = (a_j, b_j] that includes eigenvalue j. The eigenvalue
          approximation is given as the interval midpoint
          W(j)= ( a_j + b_j)/2. The corresponding error is bounded by
          WERR(j) = abs( a_j - b_j)/2

WERR

          WERR is DOUBLE PRECISION array, dimension (N)
          The error bound on the corresponding eigenvalue approximation
          in W.

WL

          WL is DOUBLE PRECISION

WU

          WU is DOUBLE PRECISION
          The interval (WL, WU] contains all the wanted eigenvalues.
          If RANGE='V', then WL=VL and WU=VU.
          If RANGE='A', then WL and WU are the global Gerschgorin bounds
                        on the spectrum.
          If RANGE='I', then WL and WU are computed by DLAEBZ from the
                        index range specified.

IBLOCK

          IBLOCK is INTEGER array, dimension (N)
          At each row/column j where E(j) is zero or small, the
          matrix T is considered to split into a block diagonal
          matrix.  On exit, if INFO = 0, IBLOCK(i) specifies to which
          block (from 1 to the number of blocks) the eigenvalue W(i)
          belongs.  (DLARRD may use the remaining N-M elements as
          workspace.)

INDEXW

          INDEXW is INTEGER array, dimension (N)
          The indices of the eigenvalues within each block (submatrix);
          for example, INDEXW(i)= j and IBLOCK(i)=k imply that the
          i-th eigenvalue W(i) is the j-th eigenvalue in block k.

WORK

          WORK is DOUBLE PRECISION array, dimension (4*N)

IWORK

          IWORK is INTEGER array, dimension (3*N)

INFO

          INFO is INTEGER
          = 0:  successful exit
          < 0:  if INFO = -i, the i-th argument had an illegal value
          > 0:  some or all of the eigenvalues failed to converge or
                were not computed:
                =1 or 3: Bisection failed to converge for some
                        eigenvalues; these eigenvalues are flagged by a
                        negative block number.  The effect is that the
                        eigenvalues may not be as accurate as the
                        absolute and relative tolerances.  This is
                        generally caused by unexpectedly inaccurate
                        arithmetic.
                =2 or 3: RANGE='I' only: Not all of the eigenvalues
                        IL:IU were found.
                        Effect: M < IU+1-IL
                        Cause:  non-monotonic arithmetic, causing the
                                Sturm sequence to be non-monotonic.
                        Cure:   recalculate, using RANGE='A', and pick
                                out eigenvalues IL:IU.  In some cases,
                                increasing the PARAMETER "FUDGE" may
                                make things work.
                = 4:    RANGE='I', and the Gershgorin interval
                        initially used was too small.  No eigenvalues
                        were computed.
                        Probable cause: your machine has sloppy
                                        floating-point arithmetic.
                        Cure: Increase the PARAMETER "FUDGE",
                              recompile, and try again.

Internal Parameters:

  FUDGE   DOUBLE PRECISION, default = 2
          A "fudge factor" to widen the Gershgorin intervals.  Ideally,
          a value of 1 should work, but on machines with sloppy
          arithmetic, this needs to be larger.  The default for
          publicly released versions should be large enough to handle
          the worst machine around.  Note that this has no effect
          on accuracy of the solution.

Contributors:

W. Kahan, University of California, Berkeley, USA
Beresford Parlett, University of California, Berkeley, USA
Jim Demmel, University of California, Berkeley, USA
Inderjit Dhillon, University of Texas, Austin, USA
Osni Marques, LBNL/NERSC, USA
Christof Voemel, University of California, Berkeley, USA

Author:

Univ. of Tennessee

Univ. of California Berkeley

Univ. of Colorado Denver

NAG Ltd.

Date:

September 2012

Definition at line 319 of file dlarrd.f.