sorbdb.f man page

sorbdb.f —

Synopsis

Functions/Subroutines

subroutine sorbdb (TRANS, SIGNS, M, P, Q, X11, LDX11, X12, LDX12, X21, LDX21, X22, LDX22, THETA, PHI, TAUP1, TAUP2, TAUQ1, TAUQ2, WORK, LWORK, INFO)
SORBDB

Function/Subroutine Documentation

subroutine sorbdb (characterTRANS, characterSIGNS, integerM, integerP, integerQ, real, dimension( ldx11, * )X11, integerLDX11, real, dimension( ldx12, * )X12, integerLDX12, real, dimension( ldx21, * )X21, integerLDX21, real, dimension( ldx22, * )X22, integerLDX22, real, dimension( * )THETA, real, dimension( * )PHI, real, dimension( * )TAUP1, real, dimension( * )TAUP2, real, dimension( * )TAUQ1, real, dimension( * )TAUQ2, real, dimension( * )WORK, integerLWORK, integerINFO)

SORBDB  

Purpose:

 SORBDB simultaneously bidiagonalizes the blocks of an M-by-M
 partitioned orthogonal matrix X:

                                 [ B11 | B12 0  0 ]
     [ X11 | X12 ]   [ P1 |    ] [  0  |  0 -I  0 ] [ Q1 |    ]**T
 X = [-----------] = [---------] [----------------] [---------]   .
     [ X21 | X22 ]   [    | P2 ] [ B21 | B22 0  0 ] [    | Q2 ]
                                 [  0  |  0  0  I ]

 X11 is P-by-Q. Q must be no larger than P, M-P, or M-Q. (If this is
 not the case, then X must be transposed and/or permuted. This can be
 done in constant time using the TRANS and SIGNS options. See SORCSD
 for details.)

 The orthogonal matrices P1, P2, Q1, and Q2 are P-by-P, (M-P)-by-
 (M-P), Q-by-Q, and (M-Q)-by-(M-Q), respectively. They are
 represented implicitly by Householder vectors.

 B11, B12, B21, and B22 are Q-by-Q bidiagonal matrices represented
 implicitly by angles THETA, PHI.
Parameters:

TRANS

          TRANS is CHARACTER
          = 'T':      X, U1, U2, V1T, and V2T are stored in row-major
                      order;
          otherwise:  X, U1, U2, V1T, and V2T are stored in column-
                      major order.

SIGNS

          SIGNS is CHARACTER
          = 'O':      The lower-left block is made nonpositive (the
                      "other" convention);
          otherwise:  The upper-right block is made nonpositive (the
                      "default" convention).

M

          M is INTEGER
          The number of rows and columns in X.

P

          P is INTEGER
          The number of rows in X11 and X12. 0 <= P <= M.

Q

          Q is INTEGER
          The number of columns in X11 and X21. 0 <= Q <=
          MIN(P,M-P,M-Q).

X11

          X11 is REAL array, dimension (LDX11,Q)
          On entry, the top-left block of the orthogonal matrix to be
          reduced. On exit, the form depends on TRANS:
          If TRANS = 'N', then
             the columns of tril(X11) specify reflectors for P1,
             the rows of triu(X11,1) specify reflectors for Q1;
          else TRANS = 'T', and
             the rows of triu(X11) specify reflectors for P1,
             the columns of tril(X11,-1) specify reflectors for Q1.

LDX11

          LDX11 is INTEGER
          The leading dimension of X11. If TRANS = 'N', then LDX11 >=
          P; else LDX11 >= Q.

X12

          X12 is REAL array, dimension (LDX12,M-Q)
          On entry, the top-right block of the orthogonal matrix to
          be reduced. On exit, the form depends on TRANS:
          If TRANS = 'N', then
             the rows of triu(X12) specify the first P reflectors for
             Q2;
          else TRANS = 'T', and
             the columns of tril(X12) specify the first P reflectors
             for Q2.

LDX12

          LDX12 is INTEGER
          The leading dimension of X12. If TRANS = 'N', then LDX12 >=
          P; else LDX11 >= M-Q.

X21

          X21 is REAL array, dimension (LDX21,Q)
          On entry, the bottom-left block of the orthogonal matrix to
          be reduced. On exit, the form depends on TRANS:
          If TRANS = 'N', then
             the columns of tril(X21) specify reflectors for P2;
          else TRANS = 'T', and
             the rows of triu(X21) specify reflectors for P2.

LDX21

          LDX21 is INTEGER
          The leading dimension of X21. If TRANS = 'N', then LDX21 >=
          M-P; else LDX21 >= Q.

X22

          X22 is REAL array, dimension (LDX22,M-Q)
          On entry, the bottom-right block of the orthogonal matrix to
          be reduced. On exit, the form depends on TRANS:
          If TRANS = 'N', then
             the rows of triu(X22(Q+1:M-P,P+1:M-Q)) specify the last
             M-P-Q reflectors for Q2,
          else TRANS = 'T', and
             the columns of tril(X22(P+1:M-Q,Q+1:M-P)) specify the last
             M-P-Q reflectors for P2.

LDX22

          LDX22 is INTEGER
          The leading dimension of X22. If TRANS = 'N', then LDX22 >=
          M-P; else LDX22 >= M-Q.

THETA

          THETA is REAL array, dimension (Q)
          The entries of the bidiagonal blocks B11, B12, B21, B22 can
          be computed from the angles THETA and PHI. See Further
          Details.

PHI

          PHI is REAL array, dimension (Q-1)
          The entries of the bidiagonal blocks B11, B12, B21, B22 can
          be computed from the angles THETA and PHI. See Further
          Details.

TAUP1

          TAUP1 is REAL array, dimension (P)
          The scalar factors of the elementary reflectors that define
          P1.

TAUP2

          TAUP2 is REAL array, dimension (M-P)
          The scalar factors of the elementary reflectors that define
          P2.

TAUQ1

          TAUQ1 is REAL array, dimension (Q)
          The scalar factors of the elementary reflectors that define
          Q1.

TAUQ2

          TAUQ2 is REAL array, dimension (M-Q)
          The scalar factors of the elementary reflectors that define
          Q2.

WORK

          WORK is REAL array, dimension (LWORK)

LWORK

          LWORK is INTEGER
          The dimension of the array WORK. LWORK >= M-Q.

          If LWORK = -1, then a workspace query is assumed; the routine
          only calculates the optimal size of the WORK array, returns
          this value as the first entry of the WORK array, and no error
          message related to LWORK is issued by XERBLA.

INFO

          INFO is INTEGER
          = 0:  successful exit.
          < 0:  if INFO = -i, the i-th argument had an illegal value.
Author:

Univ. of Tennessee

Univ. of California Berkeley

Univ. of Colorado Denver

NAG Ltd.

Date:

November 2013

Further Details:

  The bidiagonal blocks B11, B12, B21, and B22 are represented
  implicitly by angles THETA(1), ..., THETA(Q) and PHI(1), ...,
  PHI(Q-1). B11 and B21 are upper bidiagonal, while B21 and B22 are
  lower bidiagonal. Every entry in each bidiagonal band is a product
  of a sine or cosine of a THETA with a sine or cosine of a PHI. See
  [1] or SORCSD for details.

  P1, P2, Q1, and Q2 are represented as products of elementary
  reflectors. See SORCSD for details on generating P1, P2, Q1, and Q2
  using SORGQR and SORGLQ.

References:

[1] Brian D. Sutton. Computing the complete CS decomposition. Numer. Algorithms, 50(1):33-65, 2009.

Definition at line 286 of file sorbdb.f.

Author

Generated automatically by Doxygen for LAPACK from the source code.

Referenced By

sorbdb(3) is an alias of sorbdb.f(3).

Sat Nov 16 2013 Version 3.4.2 LAPACK