File:  [local] / rpl / lapack / lapack / dggesx.f
Revision 1.1.1.1 (vendor branch): download - view: text, annotated - select for diffs - revision graph
Tue Jan 26 15:22:46 2010 UTC (14 years, 3 months ago) by bertrand
Branches: JKB
CVS tags: start, rpl-4_0_14, rpl-4_0_13, rpl-4_0_12, rpl-4_0_11, rpl-4_0_10


Commit initial.

    1:       SUBROUTINE DGGESX( JOBVSL, JOBVSR, SORT, SELCTG, SENSE, N, A, LDA,
    2:      $                   B, LDB, SDIM, ALPHAR, ALPHAI, BETA, VSL, LDVSL,
    3:      $                   VSR, LDVSR, RCONDE, RCONDV, WORK, LWORK, IWORK,
    4:      $                   LIWORK, BWORK, INFO )
    5: *
    6: *  -- LAPACK driver routine (version 3.2.1)                           --
    7: *  -- LAPACK is a software package provided by Univ. of Tennessee,    --
    8: *  -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
    9: *  -- April 2009                                                      --
   10: *
   11: *     .. Scalar Arguments ..
   12:       CHARACTER          JOBVSL, JOBVSR, SENSE, SORT
   13:       INTEGER            INFO, LDA, LDB, LDVSL, LDVSR, LIWORK, LWORK, N,
   14:      $                   SDIM
   15: *     ..
   16: *     .. Array Arguments ..
   17:       LOGICAL            BWORK( * )
   18:       INTEGER            IWORK( * )
   19:       DOUBLE PRECISION   A( LDA, * ), ALPHAI( * ), ALPHAR( * ),
   20:      $                   B( LDB, * ), BETA( * ), RCONDE( 2 ),
   21:      $                   RCONDV( 2 ), VSL( LDVSL, * ), VSR( LDVSR, * ),
   22:      $                   WORK( * )
   23: *     ..
   24: *     .. Function Arguments ..
   25:       LOGICAL            SELCTG
   26:       EXTERNAL           SELCTG
   27: *     ..
   28: *
   29: *  Purpose
   30: *  =======
   31: *
   32: *  DGGESX computes for a pair of N-by-N real nonsymmetric matrices
   33: *  (A,B), the generalized eigenvalues, the real Schur form (S,T), and,
   34: *  optionally, the left and/or right matrices of Schur vectors (VSL and
   35: *  VSR).  This gives the generalized Schur factorization
   36: *
   37: *       (A,B) = ( (VSL) S (VSR)**T, (VSL) T (VSR)**T )
   38: *
   39: *  Optionally, it also orders the eigenvalues so that a selected cluster
   40: *  of eigenvalues appears in the leading diagonal blocks of the upper
   41: *  quasi-triangular matrix S and the upper triangular matrix T; computes
   42: *  a reciprocal condition number for the average of the selected
   43: *  eigenvalues (RCONDE); and computes a reciprocal condition number for
   44: *  the right and left deflating subspaces corresponding to the selected
   45: *  eigenvalues (RCONDV). The leading columns of VSL and VSR then form
   46: *  an orthonormal basis for the corresponding left and right eigenspaces
   47: *  (deflating subspaces).
   48: *
   49: *  A generalized eigenvalue for a pair of matrices (A,B) is a scalar w
   50: *  or a ratio alpha/beta = w, such that  A - w*B is singular.  It is
   51: *  usually represented as the pair (alpha,beta), as there is a
   52: *  reasonable interpretation for beta=0 or for both being zero.
   53: *
   54: *  A pair of matrices (S,T) is in generalized real Schur form if T is
   55: *  upper triangular with non-negative diagonal and S is block upper
   56: *  triangular with 1-by-1 and 2-by-2 blocks.  1-by-1 blocks correspond
   57: *  to real generalized eigenvalues, while 2-by-2 blocks of S will be
   58: *  "standardized" by making the corresponding elements of T have the
   59: *  form:
   60: *          [  a  0  ]
   61: *          [  0  b  ]
   62: *
   63: *  and the pair of corresponding 2-by-2 blocks in S and T will have a
   64: *  complex conjugate pair of generalized eigenvalues.
   65: *
   66: *
   67: *  Arguments
   68: *  =========
   69: *
   70: *  JOBVSL  (input) CHARACTER*1
   71: *          = 'N':  do not compute the left Schur vectors;
   72: *          = 'V':  compute the left Schur vectors.
   73: *
   74: *  JOBVSR  (input) CHARACTER*1
   75: *          = 'N':  do not compute the right Schur vectors;
   76: *          = 'V':  compute the right Schur vectors.
   77: *
   78: *  SORT    (input) CHARACTER*1
   79: *          Specifies whether or not to order the eigenvalues on the
   80: *          diagonal of the generalized Schur form.
   81: *          = 'N':  Eigenvalues are not ordered;
   82: *          = 'S':  Eigenvalues are ordered (see SELCTG).
   83: *
   84: *  SELCTG  (external procedure) LOGICAL FUNCTION of three DOUBLE PRECISION arguments
   85: *          SELCTG must be declared EXTERNAL in the calling subroutine.
   86: *          If SORT = 'N', SELCTG is not referenced.
   87: *          If SORT = 'S', SELCTG is used to select eigenvalues to sort
   88: *          to the top left of the Schur form.
   89: *          An eigenvalue (ALPHAR(j)+ALPHAI(j))/BETA(j) is selected if
   90: *          SELCTG(ALPHAR(j),ALPHAI(j),BETA(j)) is true; i.e. if either
   91: *          one of a complex conjugate pair of eigenvalues is selected,
   92: *          then both complex eigenvalues are selected.
   93: *          Note that a selected complex eigenvalue may no longer satisfy
   94: *          SELCTG(ALPHAR(j),ALPHAI(j),BETA(j)) = .TRUE. after ordering,
   95: *          since ordering may change the value of complex eigenvalues
   96: *          (especially if the eigenvalue is ill-conditioned), in this
   97: *          case INFO is set to N+3.
   98: *
   99: *  SENSE   (input) CHARACTER*1
  100: *          Determines which reciprocal condition numbers are computed.
  101: *          = 'N' : None are computed;
  102: *          = 'E' : Computed for average of selected eigenvalues only;
  103: *          = 'V' : Computed for selected deflating subspaces only;
  104: *          = 'B' : Computed for both.
  105: *          If SENSE = 'E', 'V', or 'B', SORT must equal 'S'.
  106: *
  107: *  N       (input) INTEGER
  108: *          The order of the matrices A, B, VSL, and VSR.  N >= 0.
  109: *
  110: *  A       (input/output) DOUBLE PRECISION array, dimension (LDA, N)
  111: *          On entry, the first of the pair of matrices.
  112: *          On exit, A has been overwritten by its generalized Schur
  113: *          form S.
  114: *
  115: *  LDA     (input) INTEGER
  116: *          The leading dimension of A.  LDA >= max(1,N).
  117: *
  118: *  B       (input/output) DOUBLE PRECISION array, dimension (LDB, N)
  119: *          On entry, the second of the pair of matrices.
  120: *          On exit, B has been overwritten by its generalized Schur
  121: *          form T.
  122: *
  123: *  LDB     (input) INTEGER
  124: *          The leading dimension of B.  LDB >= max(1,N).
  125: *
  126: *  SDIM    (output) INTEGER
  127: *          If SORT = 'N', SDIM = 0.
  128: *          If SORT = 'S', SDIM = number of eigenvalues (after sorting)
  129: *          for which SELCTG is true.  (Complex conjugate pairs for which
  130: *          SELCTG is true for either eigenvalue count as 2.)
  131: *
  132: *  ALPHAR  (output) DOUBLE PRECISION array, dimension (N)
  133: *  ALPHAI  (output) DOUBLE PRECISION array, dimension (N)
  134: *  BETA    (output) DOUBLE PRECISION array, dimension (N)
  135: *          On exit, (ALPHAR(j) + ALPHAI(j)*i)/BETA(j), j=1,...,N, will
  136: *          be the generalized eigenvalues.  ALPHAR(j) + ALPHAI(j)*i
  137: *          and BETA(j),j=1,...,N  are the diagonals of the complex Schur
  138: *          form (S,T) that would result if the 2-by-2 diagonal blocks of
  139: *          the real Schur form of (A,B) were further reduced to
  140: *          triangular form using 2-by-2 complex unitary transformations.
  141: *          If ALPHAI(j) is zero, then the j-th eigenvalue is real; if
  142: *          positive, then the j-th and (j+1)-st eigenvalues are a
  143: *          complex conjugate pair, with ALPHAI(j+1) negative.
  144: *
  145: *          Note: the quotients ALPHAR(j)/BETA(j) and ALPHAI(j)/BETA(j)
  146: *          may easily over- or underflow, and BETA(j) may even be zero.
  147: *          Thus, the user should avoid naively computing the ratio.
  148: *          However, ALPHAR and ALPHAI will be always less than and
  149: *          usually comparable with norm(A) in magnitude, and BETA always
  150: *          less than and usually comparable with norm(B).
  151: *
  152: *  VSL     (output) DOUBLE PRECISION array, dimension (LDVSL,N)
  153: *          If JOBVSL = 'V', VSL will contain the left Schur vectors.
  154: *          Not referenced if JOBVSL = 'N'.
  155: *
  156: *  LDVSL   (input) INTEGER
  157: *          The leading dimension of the matrix VSL. LDVSL >=1, and
  158: *          if JOBVSL = 'V', LDVSL >= N.
  159: *
  160: *  VSR     (output) DOUBLE PRECISION array, dimension (LDVSR,N)
  161: *          If JOBVSR = 'V', VSR will contain the right Schur vectors.
  162: *          Not referenced if JOBVSR = 'N'.
  163: *
  164: *  LDVSR   (input) INTEGER
  165: *          The leading dimension of the matrix VSR. LDVSR >= 1, and
  166: *          if JOBVSR = 'V', LDVSR >= N.
  167: *
  168: *  RCONDE  (output) DOUBLE PRECISION array, dimension ( 2 )
  169: *          If SENSE = 'E' or 'B', RCONDE(1) and RCONDE(2) contain the
  170: *          reciprocal condition numbers for the average of the selected
  171: *          eigenvalues.
  172: *          Not referenced if SENSE = 'N' or 'V'.
  173: *
  174: *  RCONDV  (output) DOUBLE PRECISION array, dimension ( 2 )
  175: *          If SENSE = 'V' or 'B', RCONDV(1) and RCONDV(2) contain the
  176: *          reciprocal condition numbers for the selected deflating
  177: *          subspaces.
  178: *          Not referenced if SENSE = 'N' or 'E'.
  179: *
  180: *  WORK    (workspace/output) DOUBLE PRECISION array, dimension (MAX(1,LWORK))
  181: *          On exit, if INFO = 0, WORK(1) returns the optimal LWORK.
  182: *
  183: *  LWORK   (input) INTEGER
  184: *          The dimension of the array WORK.
  185: *          If N = 0, LWORK >= 1, else if SENSE = 'E', 'V', or 'B',
  186: *          LWORK >= max( 8*N, 6*N+16, 2*SDIM*(N-SDIM) ), else
  187: *          LWORK >= max( 8*N, 6*N+16 ).
  188: *          Note that 2*SDIM*(N-SDIM) <= N*N/2.
  189: *          Note also that an error is only returned if
  190: *          LWORK < max( 8*N, 6*N+16), but if SENSE = 'E' or 'V' or 'B'
  191: *          this may not be large enough.
  192: *
  193: *          If LWORK = -1, then a workspace query is assumed; the routine
  194: *          only calculates the bound on the optimal size of the WORK
  195: *          array and the minimum size of the IWORK array, returns these
  196: *          values as the first entries of the WORK and IWORK arrays, and
  197: *          no error message related to LWORK or LIWORK is issued by
  198: *          XERBLA.
  199: *
  200: *  IWORK   (workspace) INTEGER array, dimension (MAX(1,LIWORK))
  201: *          On exit, if INFO = 0, IWORK(1) returns the minimum LIWORK.
  202: *
  203: *  LIWORK  (input) INTEGER
  204: *          The dimension of the array IWORK.
  205: *          If SENSE = 'N' or N = 0, LIWORK >= 1, otherwise
  206: *          LIWORK >= N+6.
  207: *
  208: *          If LIWORK = -1, then a workspace query is assumed; the
  209: *          routine only calculates the bound on the optimal size of the
  210: *          WORK array and the minimum size of the IWORK array, returns
  211: *          these values as the first entries of the WORK and IWORK
  212: *          arrays, and no error message related to LWORK or LIWORK is
  213: *          issued by XERBLA.
  214: *
  215: *  BWORK   (workspace) LOGICAL array, dimension (N)
  216: *          Not referenced if SORT = 'N'.
  217: *
  218: *  INFO    (output) INTEGER
  219: *          = 0:  successful exit
  220: *          < 0:  if INFO = -i, the i-th argument had an illegal value.
  221: *          = 1,...,N:
  222: *                The QZ iteration failed.  (A,B) are not in Schur
  223: *                form, but ALPHAR(j), ALPHAI(j), and BETA(j) should
  224: *                be correct for j=INFO+1,...,N.
  225: *          > N:  =N+1: other than QZ iteration failed in DHGEQZ
  226: *                =N+2: after reordering, roundoff changed values of
  227: *                      some complex eigenvalues so that leading
  228: *                      eigenvalues in the Generalized Schur form no
  229: *                      longer satisfy SELCTG=.TRUE.  This could also
  230: *                      be caused due to scaling.
  231: *                =N+3: reordering failed in DTGSEN.
  232: *
  233: *  Further Details
  234: *  ===============
  235: *
  236: *  An approximate (asymptotic) bound on the average absolute error of
  237: *  the selected eigenvalues is
  238: *
  239: *       EPS * norm((A, B)) / RCONDE( 1 ).
  240: *
  241: *  An approximate (asymptotic) bound on the maximum angular error in
  242: *  the computed deflating subspaces is
  243: *
  244: *       EPS * norm((A, B)) / RCONDV( 2 ).
  245: *
  246: *  See LAPACK User's Guide, section 4.11 for more information.
  247: *
  248: *  =====================================================================
  249: *
  250: *     .. Parameters ..
  251:       DOUBLE PRECISION   ZERO, ONE
  252:       PARAMETER          ( ZERO = 0.0D+0, ONE = 1.0D+0 )
  253: *     ..
  254: *     .. Local Scalars ..
  255:       LOGICAL            CURSL, ILASCL, ILBSCL, ILVSL, ILVSR, LASTSL,
  256:      $                   LQUERY, LST2SL, WANTSB, WANTSE, WANTSN, WANTST,
  257:      $                   WANTSV
  258:       INTEGER            I, ICOLS, IERR, IHI, IJOB, IJOBVL, IJOBVR,
  259:      $                   ILEFT, ILO, IP, IRIGHT, IROWS, ITAU, IWRK,
  260:      $                   LIWMIN, LWRK, MAXWRK, MINWRK
  261:       DOUBLE PRECISION   ANRM, ANRMTO, BIGNUM, BNRM, BNRMTO, EPS, PL,
  262:      $                   PR, SAFMAX, SAFMIN, SMLNUM
  263: *     ..
  264: *     .. Local Arrays ..
  265:       DOUBLE PRECISION   DIF( 2 )
  266: *     ..
  267: *     .. External Subroutines ..
  268:       EXTERNAL           DGEQRF, DGGBAK, DGGBAL, DGGHRD, DHGEQZ, DLABAD,
  269:      $                   DLACPY, DLASCL, DLASET, DORGQR, DORMQR, DTGSEN,
  270:      $                   XERBLA
  271: *     ..
  272: *     .. External Functions ..
  273:       LOGICAL            LSAME
  274:       INTEGER            ILAENV
  275:       DOUBLE PRECISION   DLAMCH, DLANGE
  276:       EXTERNAL           LSAME, ILAENV, DLAMCH, DLANGE
  277: *     ..
  278: *     .. Intrinsic Functions ..
  279:       INTRINSIC          ABS, MAX, SQRT
  280: *     ..
  281: *     .. Executable Statements ..
  282: *
  283: *     Decode the input arguments
  284: *
  285:       IF( LSAME( JOBVSL, 'N' ) ) THEN
  286:          IJOBVL = 1
  287:          ILVSL = .FALSE.
  288:       ELSE IF( LSAME( JOBVSL, 'V' ) ) THEN
  289:          IJOBVL = 2
  290:          ILVSL = .TRUE.
  291:       ELSE
  292:          IJOBVL = -1
  293:          ILVSL = .FALSE.
  294:       END IF
  295: *
  296:       IF( LSAME( JOBVSR, 'N' ) ) THEN
  297:          IJOBVR = 1
  298:          ILVSR = .FALSE.
  299:       ELSE IF( LSAME( JOBVSR, 'V' ) ) THEN
  300:          IJOBVR = 2
  301:          ILVSR = .TRUE.
  302:       ELSE
  303:          IJOBVR = -1
  304:          ILVSR = .FALSE.
  305:       END IF
  306: *
  307:       WANTST = LSAME( SORT, 'S' )
  308:       WANTSN = LSAME( SENSE, 'N' )
  309:       WANTSE = LSAME( SENSE, 'E' )
  310:       WANTSV = LSAME( SENSE, 'V' )
  311:       WANTSB = LSAME( SENSE, 'B' )
  312:       LQUERY = ( LWORK.EQ.-1 .OR. LIWORK.EQ.-1 )
  313:       IF( WANTSN ) THEN
  314:          IJOB = 0
  315:       ELSE IF( WANTSE ) THEN
  316:          IJOB = 1
  317:       ELSE IF( WANTSV ) THEN
  318:          IJOB = 2
  319:       ELSE IF( WANTSB ) THEN
  320:          IJOB = 4
  321:       END IF
  322: *
  323: *     Test the input arguments
  324: *
  325:       INFO = 0
  326:       IF( IJOBVL.LE.0 ) THEN
  327:          INFO = -1
  328:       ELSE IF( IJOBVR.LE.0 ) THEN
  329:          INFO = -2
  330:       ELSE IF( ( .NOT.WANTST ) .AND. ( .NOT.LSAME( SORT, 'N' ) ) ) THEN
  331:          INFO = -3
  332:       ELSE IF( .NOT.( WANTSN .OR. WANTSE .OR. WANTSV .OR. WANTSB ) .OR.
  333:      $         ( .NOT.WANTST .AND. .NOT.WANTSN ) ) THEN
  334:          INFO = -5
  335:       ELSE IF( N.LT.0 ) THEN
  336:          INFO = -6
  337:       ELSE IF( LDA.LT.MAX( 1, N ) ) THEN
  338:          INFO = -8
  339:       ELSE IF( LDB.LT.MAX( 1, N ) ) THEN
  340:          INFO = -10
  341:       ELSE IF( LDVSL.LT.1 .OR. ( ILVSL .AND. LDVSL.LT.N ) ) THEN
  342:          INFO = -16
  343:       ELSE IF( LDVSR.LT.1 .OR. ( ILVSR .AND. LDVSR.LT.N ) ) THEN
  344:          INFO = -18
  345:       END IF
  346: *
  347: *     Compute workspace
  348: *      (Note: Comments in the code beginning "Workspace:" describe the
  349: *       minimal amount of workspace needed at that point in the code,
  350: *       as well as the preferred amount for good performance.
  351: *       NB refers to the optimal block size for the immediately
  352: *       following subroutine, as returned by ILAENV.)
  353: *
  354:       IF( INFO.EQ.0 ) THEN
  355:          IF( N.GT.0) THEN
  356:             MINWRK = MAX( 8*N, 6*N + 16 )
  357:             MAXWRK = MINWRK - N +
  358:      $               N*ILAENV( 1, 'DGEQRF', ' ', N, 1, N, 0 )
  359:             MAXWRK = MAX( MAXWRK, MINWRK - N +
  360:      $               N*ILAENV( 1, 'DORMQR', ' ', N, 1, N, -1 ) )
  361:             IF( ILVSL ) THEN
  362:                MAXWRK = MAX( MAXWRK, MINWRK - N +
  363:      $                  N*ILAENV( 1, 'DORGQR', ' ', N, 1, N, -1 ) )
  364:             END IF
  365:             LWRK = MAXWRK
  366:             IF( IJOB.GE.1 )
  367:      $         LWRK = MAX( LWRK, N*N/2 )
  368:          ELSE
  369:             MINWRK = 1
  370:             MAXWRK = 1
  371:             LWRK   = 1
  372:          END IF
  373:          WORK( 1 ) = LWRK
  374:          IF( WANTSN .OR. N.EQ.0 ) THEN
  375:             LIWMIN = 1
  376:          ELSE
  377:             LIWMIN = N + 6
  378:          END IF
  379:          IWORK( 1 ) = LIWMIN
  380: *
  381:          IF( LWORK.LT.MINWRK .AND. .NOT.LQUERY ) THEN
  382:             INFO = -22
  383:          ELSE IF( LIWORK.LT.LIWMIN  .AND. .NOT.LQUERY ) THEN
  384:             INFO = -24
  385:          END IF
  386:       END IF
  387: *
  388:       IF( INFO.NE.0 ) THEN
  389:          CALL XERBLA( 'DGGESX', -INFO )
  390:          RETURN
  391:       ELSE IF (LQUERY) THEN
  392:          RETURN
  393:       END IF
  394: *
  395: *     Quick return if possible
  396: *
  397:       IF( N.EQ.0 ) THEN
  398:          SDIM = 0
  399:          RETURN
  400:       END IF
  401: *
  402: *     Get machine constants
  403: *
  404:       EPS = DLAMCH( 'P' )
  405:       SAFMIN = DLAMCH( 'S' )
  406:       SAFMAX = ONE / SAFMIN
  407:       CALL DLABAD( SAFMIN, SAFMAX )
  408:       SMLNUM = SQRT( SAFMIN ) / EPS
  409:       BIGNUM = ONE / SMLNUM
  410: *
  411: *     Scale A if max element outside range [SMLNUM,BIGNUM]
  412: *
  413:       ANRM = DLANGE( 'M', N, N, A, LDA, WORK )
  414:       ILASCL = .FALSE.
  415:       IF( ANRM.GT.ZERO .AND. ANRM.LT.SMLNUM ) THEN
  416:          ANRMTO = SMLNUM
  417:          ILASCL = .TRUE.
  418:       ELSE IF( ANRM.GT.BIGNUM ) THEN
  419:          ANRMTO = BIGNUM
  420:          ILASCL = .TRUE.
  421:       END IF
  422:       IF( ILASCL )
  423:      $   CALL DLASCL( 'G', 0, 0, ANRM, ANRMTO, N, N, A, LDA, IERR )
  424: *
  425: *     Scale B if max element outside range [SMLNUM,BIGNUM]
  426: *
  427:       BNRM = DLANGE( 'M', N, N, B, LDB, WORK )
  428:       ILBSCL = .FALSE.
  429:       IF( BNRM.GT.ZERO .AND. BNRM.LT.SMLNUM ) THEN
  430:          BNRMTO = SMLNUM
  431:          ILBSCL = .TRUE.
  432:       ELSE IF( BNRM.GT.BIGNUM ) THEN
  433:          BNRMTO = BIGNUM
  434:          ILBSCL = .TRUE.
  435:       END IF
  436:       IF( ILBSCL )
  437:      $   CALL DLASCL( 'G', 0, 0, BNRM, BNRMTO, N, N, B, LDB, IERR )
  438: *
  439: *     Permute the matrix to make it more nearly triangular
  440: *     (Workspace: need 6*N + 2*N for permutation parameters)
  441: *
  442:       ILEFT = 1
  443:       IRIGHT = N + 1
  444:       IWRK = IRIGHT + N
  445:       CALL DGGBAL( 'P', N, A, LDA, B, LDB, ILO, IHI, WORK( ILEFT ),
  446:      $             WORK( IRIGHT ), WORK( IWRK ), IERR )
  447: *
  448: *     Reduce B to triangular form (QR decomposition of B)
  449: *     (Workspace: need N, prefer N*NB)
  450: *
  451:       IROWS = IHI + 1 - ILO
  452:       ICOLS = N + 1 - ILO
  453:       ITAU = IWRK
  454:       IWRK = ITAU + IROWS
  455:       CALL DGEQRF( IROWS, ICOLS, B( ILO, ILO ), LDB, WORK( ITAU ),
  456:      $             WORK( IWRK ), LWORK+1-IWRK, IERR )
  457: *
  458: *     Apply the orthogonal transformation to matrix A
  459: *     (Workspace: need N, prefer N*NB)
  460: *
  461:       CALL DORMQR( 'L', 'T', IROWS, ICOLS, IROWS, B( ILO, ILO ), LDB,
  462:      $             WORK( ITAU ), A( ILO, ILO ), LDA, WORK( IWRK ),
  463:      $             LWORK+1-IWRK, IERR )
  464: *
  465: *     Initialize VSL
  466: *     (Workspace: need N, prefer N*NB)
  467: *
  468:       IF( ILVSL ) THEN
  469:          CALL DLASET( 'Full', N, N, ZERO, ONE, VSL, LDVSL )
  470:          IF( IROWS.GT.1 ) THEN
  471:             CALL DLACPY( 'L', IROWS-1, IROWS-1, B( ILO+1, ILO ), LDB,
  472:      $                   VSL( ILO+1, ILO ), LDVSL )
  473:          END IF
  474:          CALL DORGQR( IROWS, IROWS, IROWS, VSL( ILO, ILO ), LDVSL,
  475:      $                WORK( ITAU ), WORK( IWRK ), LWORK+1-IWRK, IERR )
  476:       END IF
  477: *
  478: *     Initialize VSR
  479: *
  480:       IF( ILVSR )
  481:      $   CALL DLASET( 'Full', N, N, ZERO, ONE, VSR, LDVSR )
  482: *
  483: *     Reduce to generalized Hessenberg form
  484: *     (Workspace: none needed)
  485: *
  486:       CALL DGGHRD( JOBVSL, JOBVSR, N, ILO, IHI, A, LDA, B, LDB, VSL,
  487:      $             LDVSL, VSR, LDVSR, IERR )
  488: *
  489:       SDIM = 0
  490: *
  491: *     Perform QZ algorithm, computing Schur vectors if desired
  492: *     (Workspace: need N)
  493: *
  494:       IWRK = ITAU
  495:       CALL DHGEQZ( 'S', JOBVSL, JOBVSR, N, ILO, IHI, A, LDA, B, LDB,
  496:      $             ALPHAR, ALPHAI, BETA, VSL, LDVSL, VSR, LDVSR,
  497:      $             WORK( IWRK ), LWORK+1-IWRK, IERR )
  498:       IF( IERR.NE.0 ) THEN
  499:          IF( IERR.GT.0 .AND. IERR.LE.N ) THEN
  500:             INFO = IERR
  501:          ELSE IF( IERR.GT.N .AND. IERR.LE.2*N ) THEN
  502:             INFO = IERR - N
  503:          ELSE
  504:             INFO = N + 1
  505:          END IF
  506:          GO TO 60
  507:       END IF
  508: *
  509: *     Sort eigenvalues ALPHA/BETA and compute the reciprocal of
  510: *     condition number(s)
  511: *     (Workspace: If IJOB >= 1, need MAX( 8*(N+1), 2*SDIM*(N-SDIM) )
  512: *                 otherwise, need 8*(N+1) )
  513: *
  514:       IF( WANTST ) THEN
  515: *
  516: *        Undo scaling on eigenvalues before SELCTGing
  517: *
  518:          IF( ILASCL ) THEN
  519:             CALL DLASCL( 'G', 0, 0, ANRMTO, ANRM, N, 1, ALPHAR, N,
  520:      $                   IERR )
  521:             CALL DLASCL( 'G', 0, 0, ANRMTO, ANRM, N, 1, ALPHAI, N,
  522:      $                   IERR )
  523:          END IF
  524:          IF( ILBSCL )
  525:      $      CALL DLASCL( 'G', 0, 0, BNRMTO, BNRM, N, 1, BETA, N, IERR )
  526: *
  527: *        Select eigenvalues
  528: *
  529:          DO 10 I = 1, N
  530:             BWORK( I ) = SELCTG( ALPHAR( I ), ALPHAI( I ), BETA( I ) )
  531:    10    CONTINUE
  532: *
  533: *        Reorder eigenvalues, transform Generalized Schur vectors, and
  534: *        compute reciprocal condition numbers
  535: *
  536:          CALL DTGSEN( IJOB, ILVSL, ILVSR, BWORK, N, A, LDA, B, LDB,
  537:      $                ALPHAR, ALPHAI, BETA, VSL, LDVSL, VSR, LDVSR,
  538:      $                SDIM, PL, PR, DIF, WORK( IWRK ), LWORK-IWRK+1,
  539:      $                IWORK, LIWORK, IERR )
  540: *
  541:          IF( IJOB.GE.1 )
  542:      $      MAXWRK = MAX( MAXWRK, 2*SDIM*( N-SDIM ) )
  543:          IF( IERR.EQ.-22 ) THEN
  544: *
  545: *            not enough real workspace
  546: *
  547:             INFO = -22
  548:          ELSE
  549:             IF( IJOB.EQ.1 .OR. IJOB.EQ.4 ) THEN
  550:                RCONDE( 1 ) = PL
  551:                RCONDE( 2 ) = PR
  552:             END IF
  553:             IF( IJOB.EQ.2 .OR. IJOB.EQ.4 ) THEN
  554:                RCONDV( 1 ) = DIF( 1 )
  555:                RCONDV( 2 ) = DIF( 2 )
  556:             END IF
  557:             IF( IERR.EQ.1 )
  558:      $         INFO = N + 3
  559:          END IF
  560: *
  561:       END IF
  562: *
  563: *     Apply permutation to VSL and VSR
  564: *     (Workspace: none needed)
  565: *
  566:       IF( ILVSL )
  567:      $   CALL DGGBAK( 'P', 'L', N, ILO, IHI, WORK( ILEFT ),
  568:      $                WORK( IRIGHT ), N, VSL, LDVSL, IERR )
  569: *
  570:       IF( ILVSR )
  571:      $   CALL DGGBAK( 'P', 'R', N, ILO, IHI, WORK( ILEFT ),
  572:      $                WORK( IRIGHT ), N, VSR, LDVSR, IERR )
  573: *
  574: *     Check if unscaling would cause over/underflow, if so, rescale
  575: *     (ALPHAR(I),ALPHAI(I),BETA(I)) so BETA(I) is on the order of
  576: *     B(I,I) and ALPHAR(I) and ALPHAI(I) are on the order of A(I,I)
  577: *
  578:       IF( ILASCL ) THEN
  579:          DO 20 I = 1, N
  580:             IF( ALPHAI( I ).NE.ZERO ) THEN
  581:                IF( ( ALPHAR( I ) / SAFMAX ).GT.( ANRMTO / ANRM ) .OR.
  582:      $             ( SAFMIN / ALPHAR( I ) ).GT.( ANRM / ANRMTO ) ) THEN
  583:                   WORK( 1 ) = ABS( A( I, I ) / ALPHAR( I ) )
  584:                   BETA( I ) = BETA( I )*WORK( 1 )
  585:                   ALPHAR( I ) = ALPHAR( I )*WORK( 1 )
  586:                   ALPHAI( I ) = ALPHAI( I )*WORK( 1 )
  587:                ELSE IF( ( ALPHAI( I ) / SAFMAX ).GT.
  588:      $                  ( ANRMTO / ANRM ) .OR.
  589:      $                  ( SAFMIN / ALPHAI( I ) ).GT.( ANRM / ANRMTO ) )
  590:      $                   THEN
  591:                   WORK( 1 ) = ABS( A( I, I+1 ) / ALPHAI( I ) )
  592:                   BETA( I ) = BETA( I )*WORK( 1 )
  593:                   ALPHAR( I ) = ALPHAR( I )*WORK( 1 )
  594:                   ALPHAI( I ) = ALPHAI( I )*WORK( 1 )
  595:                END IF
  596:             END IF
  597:    20    CONTINUE
  598:       END IF
  599: *
  600:       IF( ILBSCL ) THEN
  601:          DO 30 I = 1, N
  602:             IF( ALPHAI( I ).NE.ZERO ) THEN
  603:                IF( ( BETA( I ) / SAFMAX ).GT.( BNRMTO / BNRM ) .OR.
  604:      $             ( SAFMIN / BETA( I ) ).GT.( BNRM / BNRMTO ) ) THEN
  605:                   WORK( 1 ) = ABS( B( I, I ) / BETA( I ) )
  606:                   BETA( I ) = BETA( I )*WORK( 1 )
  607:                   ALPHAR( I ) = ALPHAR( I )*WORK( 1 )
  608:                   ALPHAI( I ) = ALPHAI( I )*WORK( 1 )
  609:                END IF
  610:             END IF
  611:    30    CONTINUE
  612:       END IF
  613: *
  614: *     Undo scaling
  615: *
  616:       IF( ILASCL ) THEN
  617:          CALL DLASCL( 'H', 0, 0, ANRMTO, ANRM, N, N, A, LDA, IERR )
  618:          CALL DLASCL( 'G', 0, 0, ANRMTO, ANRM, N, 1, ALPHAR, N, IERR )
  619:          CALL DLASCL( 'G', 0, 0, ANRMTO, ANRM, N, 1, ALPHAI, N, IERR )
  620:       END IF
  621: *
  622:       IF( ILBSCL ) THEN
  623:          CALL DLASCL( 'U', 0, 0, BNRMTO, BNRM, N, N, B, LDB, IERR )
  624:          CALL DLASCL( 'G', 0, 0, BNRMTO, BNRM, N, 1, BETA, N, IERR )
  625:       END IF
  626: *
  627:       IF( WANTST ) THEN
  628: *
  629: *        Check if reordering is correct
  630: *
  631:          LASTSL = .TRUE.
  632:          LST2SL = .TRUE.
  633:          SDIM = 0
  634:          IP = 0
  635:          DO 50 I = 1, N
  636:             CURSL = SELCTG( ALPHAR( I ), ALPHAI( I ), BETA( I ) )
  637:             IF( ALPHAI( I ).EQ.ZERO ) THEN
  638:                IF( CURSL )
  639:      $            SDIM = SDIM + 1
  640:                IP = 0
  641:                IF( CURSL .AND. .NOT.LASTSL )
  642:      $            INFO = N + 2
  643:             ELSE
  644:                IF( IP.EQ.1 ) THEN
  645: *
  646: *                 Last eigenvalue of conjugate pair
  647: *
  648:                   CURSL = CURSL .OR. LASTSL
  649:                   LASTSL = CURSL
  650:                   IF( CURSL )
  651:      $               SDIM = SDIM + 2
  652:                   IP = -1
  653:                   IF( CURSL .AND. .NOT.LST2SL )
  654:      $               INFO = N + 2
  655:                ELSE
  656: *
  657: *                 First eigenvalue of conjugate pair
  658: *
  659:                   IP = 1
  660:                END IF
  661:             END IF
  662:             LST2SL = LASTSL
  663:             LASTSL = CURSL
  664:    50    CONTINUE
  665: *
  666:       END IF
  667: *
  668:    60 CONTINUE
  669: *
  670:       WORK( 1 ) = MAXWRK
  671:       IWORK( 1 ) = LIWMIN
  672: *
  673:       RETURN
  674: *
  675: *     End of DGGESX
  676: *
  677:       END

CVSweb interface <joel.bertrand@systella.fr>