slaebz.f man page

subroutine slaebz (IJOB, NITMAX, N, MMAX, MINP, NBMIN, ABSTOL, RELTOL, PIVMIN, D, E, E2, NVAL, AB, C, MOUT, NAB, WORK, IWORK, INFO)
SLAEBZ computes the number of eigenvalues of a real symmetric tridiagonal matrix which are less than or equal to a given value, and performs other tasks required by the routine sstebz.

SLAEBZ computes the number of eigenvalues of a real symmetric tridiagonal matrix which are less than or equal to a given value, and performs other tasks required by the routine sstebz.  

Purpose:

 SLAEBZ contains the iteration loops which compute and use the
 function N(w), which is the count of eigenvalues of a symmetric
 tridiagonal matrix T less than or equal to its argument  w.  It
 performs a choice of two types of loops:

 IJOB=1, followed by
 IJOB=2: It takes as input a list of intervals and returns a list of
         sufficiently small intervals whose union contains the same
         eigenvalues as the union of the original intervals.
         The input intervals are (AB(j,1),AB(j,2)], j=1,...,MINP.
         The output interval (AB(j,1),AB(j,2)] will contain
         eigenvalues NAB(j,1)+1,...,NAB(j,2), where 1 <= j <= MOUT.

 IJOB=3: It performs a binary search in each input interval
         (AB(j,1),AB(j,2)] for a point  w(j)  such that
         N(w(j))=NVAL(j), and uses  C(j)  as the starting point of
         the search.  If such a w(j) is found, then on output
         AB(j,1)=AB(j,2)=w.  If no such w(j) is found, then on output
         (AB(j,1),AB(j,2)] will be a small interval containing the
         point where N(w) jumps through NVAL(j), unless that point
         lies outside the initial interval.

 Note that the intervals are in all cases half-open intervals,
 i.e., of the form  (a,b] , which includes  b  but not  a .

 To avoid underflow, the matrix should be scaled so that its largest
 element is no greater than  overflow**(1/2) * underflow**(1/4)
 in absolute value.  To assure the most accurate computation
 of small eigenvalues, the matrix should be scaled to be
 not much smaller than that, either.

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

 Note: the arguments are, in general, *not* checked for unreasonable
 values.

Parameters:

IJOB

          IJOB is INTEGER
          Specifies what is to be done:
          = 1:  Compute NAB for the initial intervals.
          = 2:  Perform bisection iteration to find eigenvalues of T.
          = 3:  Perform bisection iteration to invert N(w), i.e.,
                to find a point which has a specified number of
                eigenvalues of T to its left.
          Other values will cause SLAEBZ to return with INFO=-1.

NITMAX

          NITMAX is INTEGER
          The maximum number of "levels" of bisection to be
          performed, i.e., an interval of width W will not be made
          smaller than 2^(-NITMAX) * W.  If not all intervals
          have converged after NITMAX iterations, then INFO is set
          to the number of non-converged intervals.

N

          N is INTEGER
          The dimension n of the tridiagonal matrix T.  It must be at
          least 1.

MMAX

          MMAX is INTEGER
          The maximum number of intervals.  If more than MMAX intervals
          are generated, then SLAEBZ will quit with INFO=MMAX+1.

MINP

          MINP is INTEGER
          The initial number of intervals.  It may not be greater than
          MMAX.

NBMIN

          NBMIN is INTEGER
          The smallest number of intervals that should be processed
          using a vector loop.  If zero, then only the scalar loop
          will be used.

ABSTOL

          ABSTOL is REAL
          The minimum (absolute) width of an interval.  When an
          interval is narrower than ABSTOL, or than RELTOL times the
          larger (in magnitude) endpoint, then it is considered to be
          sufficiently small, i.e., converged.  This must be at least
          zero.

RELTOL

          RELTOL is REAL
          The minimum relative width of an interval.  When an interval
          is narrower than ABSTOL, or 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.

PIVMIN

          PIVMIN is REAL
          The minimum absolute value of a "pivot" in the Sturm
          sequence loop.
          This must be at least  max |e(j)**2|*safe_min  and at
          least safe_min, where safe_min is at least
          the smallest number that can divide one without overflow.

D

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

E

          E is REAL array, dimension (N)
          The offdiagonal elements of the tridiagonal matrix T in
          positions 1 through N-1.  E(N) is arbitrary.

E2

          E2 is REAL array, dimension (N)
          The squares of the offdiagonal elements of the tridiagonal
          matrix T.  E2(N) is ignored.

NVAL

          NVAL is INTEGER array, dimension (MINP)
          If IJOB=1 or 2, not referenced.
          If IJOB=3, the desired values of N(w).  The elements of NVAL
          will be reordered to correspond with the intervals in AB.
          Thus, NVAL(j) on output will not, in general be the same as
          NVAL(j) on input, but it will correspond with the interval
          (AB(j,1),AB(j,2)] on output.

AB

          AB is REAL array, dimension (MMAX,2)
          The endpoints of the intervals.  AB(j,1) is  a(j), the left
          endpoint of the j-th interval, and AB(j,2) is b(j), the
          right endpoint of the j-th interval.  The input intervals
          will, in general, be modified, split, and reordered by the
          calculation.

C

          C is REAL array, dimension (MMAX)
          If IJOB=1, ignored.
          If IJOB=2, workspace.
          If IJOB=3, then on input C(j) should be initialized to the
          first search point in the binary search.

MOUT

          MOUT is INTEGER
          If IJOB=1, the number of eigenvalues in the intervals.
          If IJOB=2 or 3, the number of intervals output.
          If IJOB=3, MOUT will equal MINP.

NAB

          NAB is INTEGER array, dimension (MMAX,2)
          If IJOB=1, then on output NAB(i,j) will be set to N(AB(i,j)).
          If IJOB=2, then on input, NAB(i,j) should be set.  It must
             satisfy the condition:
             N(AB(i,1)) <= NAB(i,1) <= NAB(i,2) <= N(AB(i,2)),
             which means that in interval i only eigenvalues
             NAB(i,1)+1,...,NAB(i,2) will be considered.  Usually,
             NAB(i,j)=N(AB(i,j)), from a previous call to SLAEBZ with
             IJOB=1.
             On output, NAB(i,j) will contain
             max(na(k),min(nb(k),N(AB(i,j)))), where k is the index of
             the input interval that the output interval
             (AB(j,1),AB(j,2)] came from, and na(k) and nb(k) are the
             the input values of NAB(k,1) and NAB(k,2).
          If IJOB=3, then on output, NAB(i,j) contains N(AB(i,j)),
             unless N(w) > NVAL(i) for all search points  w , in which
             case NAB(i,1) will not be modified, i.e., the output
             value will be the same as the input value (modulo
             reorderings -- see NVAL and AB), or unless N(w) < NVAL(i)
             for all search points  w , in which case NAB(i,2) will
             not be modified.  Normally, NAB should be set to some
             distinctive value(s) before SLAEBZ is called.

WORK

          WORK is REAL array, dimension (MMAX)
          Workspace.

IWORK

          IWORK is INTEGER array, dimension (MMAX)
          Workspace.

INFO

          INFO is INTEGER
          = 0:       All intervals converged.
          = 1--MMAX: The last INFO intervals did not converge.
          = MMAX+1:  More than MMAX intervals were generated.

Author:

Univ. of Tennessee

Univ. of California Berkeley

Univ. of Colorado Denver

NAG Ltd.

Date:

September 2012

Further Details:

      This routine is intended to be called only by other LAPACK
  routines, thus the interface is less user-friendly.  It is intended
  for two purposes:

  (a) finding eigenvalues.  In this case, SLAEBZ should have one or
      more initial intervals set up in AB, and SLAEBZ should be called
      with IJOB=1.  This sets up NAB, and also counts the eigenvalues.
      Intervals with no eigenvalues would usually be thrown out at
      this point.  Also, if not all the eigenvalues in an interval i
      are desired, NAB(i,1) can be increased or NAB(i,2) decreased.
      For example, set NAB(i,1)=NAB(i,2)-1 to get the largest
      eigenvalue.  SLAEBZ is then called with IJOB=2 and MMAX
      no smaller than the value of MOUT returned by the call with
      IJOB=1.  After this (IJOB=2) call, eigenvalues NAB(i,1)+1
      through NAB(i,2) are approximately AB(i,1) (or AB(i,2)) to the
      tolerance specified by ABSTOL and RELTOL.

  (b) finding an interval (a',b'] containing eigenvalues w(f),...,w(l).
      In this case, start with a Gershgorin interval  (a,b).  Set up
      AB to contain 2 search intervals, both initially (a,b).  One
      NVAL element should contain  f-1  and the other should contain  l
      , while C should contain a and b, resp.  NAB(i,1) should be -1
      and NAB(i,2) should be N+1, to flag an error if the desired
      interval does not lie in (a,b).  SLAEBZ is then called with
      IJOB=3.  On exit, if w(f-1) < w(f), then one of the intervals --
      j -- will have AB(j,1)=AB(j,2) and NAB(j,1)=NAB(j,2)=f-1, while
      if, to the specified tolerance, w(f-k)=...=w(f+r), k > 0 and r
      >= 0, then the interval will have  N(AB(j,1))=NAB(j,1)=f-k and
      N(AB(j,2))=NAB(j,2)=f+r.  The cases w(l) < w(l+1) and
      w(l-r)=...=w(l+k) are handled similarly.

Definition at line 318 of file slaebz.f.