1
       2
       3
       4
       5
       6
       7
       8
       9
      10
      11
      12
      13
      14
      15
      16
      17
      18
      19
      20
      21
      22
      23
      24
      25
      26
      27
      28
      29
      30
      31
      32
      33
      34
      35
      36
      37
      38
      39
      40
      41
      42
      43
      44
      45
      46
      47
      48
      49
      50
      51
      52
      53
      54
      55
      56
      57
      58
      59
      60
      61
      62
      63
      64
      65
      66
      67
      68
      69
      70
      71
      72
      73
      74
      75
      76
      77
      78
      79
      80
      81
      82
      83
      84
      85
      86
      87
      88
      89
      90
      91
      92
      93
      94
      95
      96
      97
      98
      99
     100
     101
     102
     103
     104
     105
     106
     107
     108
     109
     110
     111
     112
     113
     114
     115
     116
     117
     118
     119
     120
     121
     122
     123
     124
     125
     126
     127
     128
     129
     130
     131
     132
     133
     134
     135
     136
     137
     138
     139
     140
     141
     142
     143
     144
     145
     146
     147
     148
     149
     150
     151
     152
     153
     154
     155
     156
     157
     158
     159
     160
     161
     162
     163
     164
     165
     166
     167
     168
     169
     170
     171
     172
     173
     174
     175
     176
     177
     178
     179
     180
     181
     182
     183
     184
     185
     186
     187
     188
     189
     190
     191
     192
     193
     194
     195
     196
     197
     198
     199
     200
     201
     202
     203
     204
     205
     206
     207
     208
     209
     210
     211
     212
     213
     214
     215
     216
     217
     218
     219
     220
     221
     222
     223
     224
     225
     226
     227
     228
     229
     230
     231
     232
     233
     234
     235
     236
     237
     238
     239
     240
     241
     242
     243
     244
     245
     246
     247
     248
     249
     250
     251
     252
     253
     254
     255
     256
     257
     258
     259
     260
     261
     262
     263
     264
     265
     266
     267
     268
     269
     270
     271
     272
     273
     274
     275
     276
     277
     278
     279
     280
     281
     282
     283
     284
     285
     286
     287
     288
     289
     290
     291
     292
     293
     294
     295
     296
     297
     298
     299
     300
     301
     302
     303
     304
     305
     306
     307
     308
     309
     310
     311
     312
     313
     314
     315
     316
     317
     318
     319
     320
     321
     322
     323
     324
     325
     326
     327
     328
     329
     330
     331
     332
     333
     334
     335
     336
     337
     338
     339
     340
     341
     342
     343
     344
     345
     346
     347
     348
     349
     350
     351
     352
     353
     354
     355
     356
     357
     358
     359
     360
     361
     362
     363
     364
     365
     366
     367
     368
     369
     370
     371
     372
     373
     374
     375
     376
     377
     378
     379
     380
     381
     382
     383
     384
     385
     386
     387
     388
     389
     390
     391
     392
     393
     394
     395
     396
     397
     398
     399
     400
     401
     402
     403
     404
     405
     406
     407
     408
     409
     410
     411
     412
     413
     414
     415
     416
     417
     418
     419
     420
     421
     422
     423
     424
     425
     426
     427
     428
     429
     430
     431
     432
     433
     434
     435
     436
     437
     438
     439
     440
     441
     442
     443
     444
     445
     446
     447
     448
     449
     450
     451
     452
     453
     454
     455
     456
     457
     458
     459
     460
     461
     462
     463
     464
     465
     466
     467
     468
     469
     470
     471
     472
     473
     474
     475
     476
     477
     478
     479
     480
     481
     482
     483
     484
     485
     486
     487
     488
     489
     490
     491
     492
     493
     494
     495
     496
     497
     498
     499
     500
     501
     502
     503
     504
     505
     506
     507
     508
     509
     510
     511
     512
     513
     514
     515
     516
     517
     518
     519
     520
     521
     522
     523
     524
     525
     526
     527
     528
     529
     530
     531
     532
     533
     534
     535
     536
     537
     538
     539
     540
     541
     542
     543
     544
     545
     546
     547
     548
     549
     550
     551
     552
     553
     554
     555
     556
     557
     558
     559
     560
     561
     562
     563
     564
     565
     566
     567
     568
     569
     570
     571
     572
     573
     574
     575
     576
     577
     578
     579
     580
     581
     582
     583
     584
     585
     586
     587
     588
     589
     590
     591
     592
     593
     594
     595
     596
     597
     598
     599
     600
     601
     602
     603
     604
     605
     606
     607
     608
     609
     610
     611
     612
     613
     614
     615
     616
     617
     618
     619
     620
     621
     622
     623
     624
     625
     626
     627
     628
     629
     630
     631
     632
     633
     634
     635
     636
     637
     638
     639
     640
     641
     642
     643
     644
     645
     646
     647
     648
     649
     650
     651
     652
     653
      SUBROUTINE CGGEVX( BALANC, JOBVL, JOBVR, SENSE, N, A, LDA, B, LDB,
     $                   ALPHA, BETA, VL, LDVL, VR, LDVR, ILO, IHI,
     $                   LSCALE, RSCALE, ABNRM, BBNRM, RCONDE, RCONDV,
     $                   WORK, LWORK, RWORK, IWORK, BWORK, INFO )
*
*  -- LAPACK driver routine (version 3.2) --
*  -- LAPACK is a software package provided by Univ. of Tennessee,    --
*  -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
*     November 2006
*
*     .. Scalar Arguments ..
      CHARACTER          BALANC, JOBVL, JOBVR, SENSE
      INTEGER            IHI, ILO, INFO, LDA, LDB, LDVL, LDVR, LWORK, N
      REAL               ABNRM, BBNRM
*     ..
*     .. Array Arguments ..
      LOGICAL            BWORK( * )
      INTEGER            IWORK( * )
      REAL               LSCALE( * ), RCONDE( * ), RCONDV( * ),
     $                   RSCALE( * ), RWORK( * )
      COMPLEX            A( LDA, * ), ALPHA( * ), B( LDB, * ),
     $                   BETA( * ), VL( LDVL, * ), VR( LDVR, * ),
     $                   WORK( * )
*     ..
*
*  Purpose
*  =======
*
*  CGGEVX computes for a pair of N-by-N complex nonsymmetric matrices
*  (A,B) the generalized eigenvalues, and optionally, the left and/or
*  right generalized eigenvectors.
*
*  Optionally, it also computes a balancing transformation to improve
*  the conditioning of the eigenvalues and eigenvectors (ILO, IHI,
*  LSCALE, RSCALE, ABNRM, and BBNRM), reciprocal condition numbers for
*  the eigenvalues (RCONDE), and reciprocal condition numbers for the
*  right eigenvectors (RCONDV).
*
*  A generalized eigenvalue for a pair of matrices (A,B) is a scalar
*  lambda or a ratio alpha/beta = lambda, such that A - lambda*B is
*  singular. It is usually represented as the pair (alpha,beta), as
*  there is a reasonable interpretation for beta=0, and even for both
*  being zero.
*
*  The right eigenvector v(j) corresponding to the eigenvalue lambda(j)
*  of (A,B) satisfies
*                   A * v(j) = lambda(j) * B * v(j) .
*  The left eigenvector u(j) corresponding to the eigenvalue lambda(j)
*  of (A,B) satisfies
*                   u(j)**H * A  = lambda(j) * u(j)**H * B.
*  where u(j)**H is the conjugate-transpose of u(j).
*
*
*  Arguments
*  =========
*
*  BALANC  (input) CHARACTER*1
*          Specifies the balance option to be performed:
*          = 'N':  do not diagonally scale or permute;
*          = 'P':  permute only;
*          = 'S':  scale only;
*          = 'B':  both permute and scale.
*          Computed reciprocal condition numbers will be for the
*          matrices after permuting and/or balancing. Permuting does
*          not change condition numbers (in exact arithmetic), but
*          balancing does.
*
*  JOBVL   (input) CHARACTER*1
*          = 'N':  do not compute the left generalized eigenvectors;
*          = 'V':  compute the left generalized eigenvectors.
*
*  JOBVR   (input) CHARACTER*1
*          = 'N':  do not compute the right generalized eigenvectors;
*          = 'V':  compute the right generalized eigenvectors.
*
*  SENSE   (input) CHARACTER*1
*          Determines which reciprocal condition numbers are computed.
*          = 'N': none are computed;
*          = 'E': computed for eigenvalues only;
*          = 'V': computed for eigenvectors only;
*          = 'B': computed for eigenvalues and eigenvectors.
*
*  N       (input) INTEGER
*          The order of the matrices A, B, VL, and VR.  N >= 0.
*
*  A       (input/output) COMPLEX array, dimension (LDA, N)
*          On entry, the matrix A in the pair (A,B).
*          On exit, A has been overwritten. If JOBVL='V' or JOBVR='V'
*          or both, then A contains the first part of the complex Schur
*          form of the "balanced" versions of the input A and B.
*
*  LDA     (input) INTEGER
*          The leading dimension of A.  LDA >= max(1,N).
*
*  B       (input/output) COMPLEX array, dimension (LDB, N)
*          On entry, the matrix B in the pair (A,B).
*          On exit, B has been overwritten. If JOBVL='V' or JOBVR='V'
*          or both, then B contains the second part of the complex
*          Schur form of the "balanced" versions of the input A and B.
*
*  LDB     (input) INTEGER
*          The leading dimension of B.  LDB >= max(1,N).
*
*  ALPHA   (output) COMPLEX array, dimension (N)
*  BETA    (output) COMPLEX array, dimension (N)
*          On exit, ALPHA(j)/BETA(j), j=1,...,N, will be the generalized
*          eigenvalues.
*
*          Note: the quotient ALPHA(j)/BETA(j) ) may easily over- or
*          underflow, and BETA(j) may even be zero.  Thus, the user
*          should avoid naively computing the ratio ALPHA/BETA.
*          However, ALPHA will be always less than and usually
*          comparable with norm(A) in magnitude, and BETA always less
*          than and usually comparable with norm(B).
*
*  VL      (output) COMPLEX array, dimension (LDVL,N)
*          If JOBVL = 'V', the left generalized eigenvectors u(j) are
*          stored one after another in the columns of VL, in the same
*          order as their eigenvalues.
*          Each eigenvector will be scaled so the largest component
*          will have abs(real part) + abs(imag. part) = 1.
*          Not referenced if JOBVL = 'N'.
*
*  LDVL    (input) INTEGER
*          The leading dimension of the matrix VL. LDVL >= 1, and
*          if JOBVL = 'V', LDVL >= N.
*
*  VR      (output) COMPLEX array, dimension (LDVR,N)
*          If JOBVR = 'V', the right generalized eigenvectors v(j) are
*          stored one after another in the columns of VR, in the same
*          order as their eigenvalues.
*          Each eigenvector will be scaled so the largest component
*          will have abs(real part) + abs(imag. part) = 1.
*          Not referenced if JOBVR = 'N'.
*
*  LDVR    (input) INTEGER
*          The leading dimension of the matrix VR. LDVR >= 1, and
*          if JOBVR = 'V', LDVR >= N.
*
*  ILO     (output) INTEGER
*  IHI     (output) INTEGER
*          ILO and IHI are integer values such that on exit
*          A(i,j) = 0 and B(i,j) = 0 if i > j and
*          j = 1,...,ILO-1 or i = IHI+1,...,N.
*          If BALANC = 'N' or 'S', ILO = 1 and IHI = N.
*
*  LSCALE  (output) REAL array, dimension (N)
*          Details of the permutations and scaling factors applied
*          to the left side of A and B.  If PL(j) is the index of the
*          row interchanged with row j, and DL(j) is the scaling
*          factor applied to row j, then
*            LSCALE(j) = PL(j)  for j = 1,...,ILO-1
*                      = DL(j)  for j = ILO,...,IHI
*                      = PL(j)  for j = IHI+1,...,N.
*          The order in which the interchanges are made is N to IHI+1,
*          then 1 to ILO-1.
*
*  RSCALE  (output) REAL array, dimension (N)
*          Details of the permutations and scaling factors applied
*          to the right side of A and B.  If PR(j) is the index of the
*          column interchanged with column j, and DR(j) is the scaling
*          factor applied to column j, then
*            RSCALE(j) = PR(j)  for j = 1,...,ILO-1
*                      = DR(j)  for j = ILO,...,IHI
*                      = PR(j)  for j = IHI+1,...,N
*          The order in which the interchanges are made is N to IHI+1,
*          then 1 to ILO-1.
*
*  ABNRM   (output) REAL
*          The one-norm of the balanced matrix A.
*
*  BBNRM   (output) REAL
*          The one-norm of the balanced matrix B.
*
*  RCONDE  (output) REAL array, dimension (N)
*          If SENSE = 'E' or 'B', the reciprocal condition numbers of
*          the eigenvalues, stored in consecutive elements of the array.
*          If SENSE = 'N' or 'V', RCONDE is not referenced.
*
*  RCONDV  (output) REAL array, dimension (N)
*          If SENSE = 'V' or 'B', the estimated reciprocal condition
*          numbers of the eigenvectors, stored in consecutive elements
*          of the array. If the eigenvalues cannot be reordered to
*          compute RCONDV(j), RCONDV(j) is set to 0; this can only occur
*          when the true value would be very small anyway. 
*          If SENSE = 'N' or 'E', RCONDV is not referenced.
*
*  WORK    (workspace/output) COMPLEX array, dimension (MAX(1,LWORK))
*          On exit, if INFO = 0, WORK(1) returns the optimal LWORK.
*
*  LWORK   (input) INTEGER
*          The dimension of the array WORK. LWORK >= max(1,2*N).
*          If SENSE = 'E', LWORK >= max(1,4*N).
*          If SENSE = 'V' or 'B', LWORK >= max(1,2*N*N+2*N).
*
*          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.
*
*  RWORK   (workspace) REAL array, dimension (lrwork)
*          lrwork must be at least max(1,6*N) if BALANC = 'S' or 'B',
*          and at least max(1,2*N) otherwise.
*          Real workspace.
*
*  IWORK   (workspace) INTEGER array, dimension (N+2)
*          If SENSE = 'E', IWORK is not referenced.
*
*  BWORK   (workspace) LOGICAL array, dimension (N)
*          If SENSE = 'N', BWORK is not referenced.
*
*  INFO    (output) INTEGER
*          = 0:  successful exit
*          < 0:  if INFO = -i, the i-th argument had an illegal value.
*          = 1,...,N:
*                The QZ iteration failed.  No eigenvectors have been
*                calculated, but ALPHA(j) and BETA(j) should be correct
*                for j=INFO+1,...,N.
*          > N:  =N+1: other than QZ iteration failed in CHGEQZ.
*                =N+2: error return from CTGEVC.
*
*  Further Details
*  ===============
*
*  Balancing a matrix pair (A,B) includes, first, permuting rows and
*  columns to isolate eigenvalues, second, applying diagonal similarity
*  transformation to the rows and columns to make the rows and columns
*  as close in norm as possible. The computed reciprocal condition
*  numbers correspond to the balanced matrix. Permuting rows and columns
*  will not change the condition numbers (in exact arithmetic) but
*  diagonal scaling will.  For further explanation of balancing, see
*  section 4.11.1.2 of LAPACK Users' Guide.
*
*  An approximate error bound on the chordal distance between the i-th
*  computed generalized eigenvalue w and the corresponding exact
*  eigenvalue lambda is
*
*       chord(w, lambda) <= EPS * norm(ABNRM, BBNRM) / RCONDE(I)
*
*  An approximate error bound for the angle between the i-th computed
*  eigenvector VL(i) or VR(i) is given by
*
*       EPS * norm(ABNRM, BBNRM) / DIF(i).
*
*  For further explanation of the reciprocal condition numbers RCONDE
*  and RCONDV, see section 4.11 of LAPACK User's Guide.
*
*     .. Parameters ..
      REAL               ZERO, ONE
      PARAMETER          ( ZERO = 0.0E+0, ONE = 1.0E+0 )
      COMPLEX            CZERO, CONE
      PARAMETER          ( CZERO = ( 0.0E+00.0E+0 ),
     $                   CONE = ( 1.0E+00.0E+0 ) )
*     ..
*     .. Local Scalars ..
      LOGICAL            ILASCL, ILBSCL, ILV, ILVL, ILVR, LQUERY, NOSCL,
     $                   WANTSB, WANTSE, WANTSN, WANTSV
      CHARACTER          CHTEMP
      INTEGER            I, ICOLS, IERR, IJOBVL, IJOBVR, IN, IROWS,
     $                   ITAU, IWRK, IWRK1, J, JC, JR, M, MAXWRK, MINWRK
      REAL               ANRM, ANRMTO, BIGNUM, BNRM, BNRMTO, EPS,
     $                   SMLNUM, TEMP
      COMPLEX            X
*     ..
*     .. Local Arrays ..
      LOGICAL            LDUMMA( 1 )
*     ..
*     .. External Subroutines ..
      EXTERNAL           CGEQRF, CGGBAK, CGGBAL, CGGHRD, CHGEQZ, CLACPY,
     $                   CLASCL, CLASET, CTGEVC, CTGSNA, CUNGQR, CUNMQR,
     $                   SLABAD, SLASCL, XERBLA
*     ..
*     .. External Functions ..
      LOGICAL            LSAME
      INTEGER            ILAENV
      REAL               CLANGE, SLAMCH
      EXTERNAL           LSAME, ILAENV, CLANGE, SLAMCH
*     ..
*     .. Intrinsic Functions ..
      INTRINSIC          ABSAIMAGMAX, REAL, SQRT
*     ..
*     .. Statement Functions ..
      REAL               ABS1
*     ..
*     .. Statement Function definitions ..
      ABS1( X ) = ABSREAL( X ) ) + ABSAIMAG( X ) )
*     ..
*     .. Executable Statements ..
*
*     Decode the input arguments
*
      IF( LSAME( JOBVL, 'N' ) ) THEN
         IJOBVL = 1
         ILVL = .FALSE.
      ELSE IF( LSAME( JOBVL, 'V' ) ) THEN
         IJOBVL = 2
         ILVL = .TRUE.
      ELSE
         IJOBVL = -1
         ILVL = .FALSE.
      END IF
*
      IF( LSAME( JOBVR, 'N' ) ) THEN
         IJOBVR = 1
         ILVR = .FALSE.
      ELSE IF( LSAME( JOBVR, 'V' ) ) THEN
         IJOBVR = 2
         ILVR = .TRUE.
      ELSE
         IJOBVR = -1
         ILVR = .FALSE.
      END IF
      ILV = ILVL .OR. ILVR
*
      NOSCL  = LSAME( BALANC, 'N' ) .OR. LSAME( BALANC, 'P' )
      WANTSN = LSAME( SENSE, 'N' )
      WANTSE = LSAME( SENSE, 'E' )
      WANTSV = LSAME( SENSE, 'V' )
      WANTSB = LSAME( SENSE, 'B' )
*
*     Test the input arguments
*
      INFO = 0
      LQUERY = ( LWORK.EQ.-1 )
      IF.NOT.( NOSCL .OR. LSAME( BALANC,'S' ) .OR.
     $    LSAME( BALANC, 'B' ) ) ) THEN
         INFO = -1
      ELSE IF( IJOBVL.LE.0 ) THEN
         INFO = -2
      ELSE IF( IJOBVR.LE.0 ) THEN
         INFO = -3
      ELSE IF.NOT.( WANTSN .OR. WANTSE .OR. WANTSB .OR. WANTSV ) )
     $          THEN
         INFO = -4
      ELSE IF( N.LT.0 ) THEN
         INFO = -5
      ELSE IF( LDA.LT.MAX1, N ) ) THEN
         INFO = -7
      ELSE IF( LDB.LT.MAX1, N ) ) THEN
         INFO = -9
      ELSE IF( LDVL.LT.1 .OR. ( ILVL .AND. LDVL.LT.N ) ) THEN
         INFO = -13
      ELSE IF( LDVR.LT.1 .OR. ( ILVR .AND. LDVR.LT.N ) ) THEN
         INFO = -15
      END IF
*
*     Compute workspace
*      (Note: Comments in the code beginning "Workspace:" describe the
*       minimal amount of workspace needed at that point in the code,
*       as well as the preferred amount for good performance.
*       NB refers to the optimal block size for the immediately
*       following subroutine, as returned by ILAENV. The workspace is
*       computed assuming ILO = 1 and IHI = N, the worst case.)
*
      IF( INFO.EQ.0 ) THEN
         IF( N.EQ.0 ) THEN
            MINWRK = 1
            MAXWRK = 1
         ELSE
            MINWRK = 2*N
            IF( WANTSE ) THEN
               MINWRK = 4*N
            ELSE IF( WANTSV .OR. WANTSB ) THEN
               MINWRK = 2*N*( N + 1)
            END IF
            MAXWRK = MINWRK
            MAXWRK = MAX( MAXWRK,
     $                    N + N*ILAENV( 1'CGEQRF'' ', N, 1, N, 0 ) )
            MAXWRK = MAX( MAXWRK,
     $                    N + N*ILAENV( 1'CUNMQR'' ', N, 1, N, 0 ) )
            IF( ILVL ) THEN
               MAXWRK = MAX( MAXWRK, N +
     $                       N*ILAENV( 1'CUNGQR'' ', N, 1, N, 0 ) )
            END IF
         END IF
         WORK( 1 ) = MAXWRK
*
         IF( LWORK.LT.MINWRK .AND. .NOT.LQUERY ) THEN
            INFO = -25
         END IF
      END IF
*
      IF( INFO.NE.0 ) THEN
         CALL XERBLA( 'CGGEVX'-INFO )
         RETURN
      ELSE IF( LQUERY ) THEN
         RETURN
      END IF
*
*     Quick return if possible
*
      IF( N.EQ.0 )
     $   RETURN
*
*     Get machine constants
*
      EPS = SLAMCH( 'P' )
      SMLNUM = SLAMCH( 'S' )
      BIGNUM = ONE / SMLNUM
      CALL SLABAD( SMLNUM, BIGNUM )
      SMLNUM = SQRT( SMLNUM ) / EPS
      BIGNUM = ONE / SMLNUM
*
*     Scale A if max element outside range [SMLNUM,BIGNUM]
*
      ANRM = CLANGE( 'M', N, N, A, LDA, RWORK )
      ILASCL = .FALSE.
      IF( ANRM.GT.ZERO .AND. ANRM.LT.SMLNUM ) THEN
         ANRMTO = SMLNUM
         ILASCL = .TRUE.
      ELSE IF( ANRM.GT.BIGNUM ) THEN
         ANRMTO = BIGNUM
         ILASCL = .TRUE.
      END IF
      IF( ILASCL )
     $   CALL CLASCL( 'G'00, ANRM, ANRMTO, N, N, A, LDA, IERR )
*
*     Scale B if max element outside range [SMLNUM,BIGNUM]
*
      BNRM = CLANGE( 'M', N, N, B, LDB, RWORK )
      ILBSCL = .FALSE.
      IF( BNRM.GT.ZERO .AND. BNRM.LT.SMLNUM ) THEN
         BNRMTO = SMLNUM
         ILBSCL = .TRUE.
      ELSE IF( BNRM.GT.BIGNUM ) THEN
         BNRMTO = BIGNUM
         ILBSCL = .TRUE.
      END IF
      IF( ILBSCL )
     $   CALL CLASCL( 'G'00, BNRM, BNRMTO, N, N, B, LDB, IERR )
*
*     Permute and/or balance the matrix pair (A,B)
*     (Real Workspace: need 6*N if BALANC = 'S' or 'B', 1 otherwise)
*
      CALL CGGBAL( BALANC, N, A, LDA, B, LDB, ILO, IHI, LSCALE, RSCALE,
     $             RWORK, IERR )
*
*     Compute ABNRM and BBNRM
*
      ABNRM = CLANGE( '1', N, N, A, LDA, RWORK( 1 ) )
      IF( ILASCL ) THEN
         RWORK( 1 ) = ABNRM
         CALL SLASCL( 'G'00, ANRMTO, ANRM, 11, RWORK( 1 ), 1,
     $                IERR )
         ABNRM = RWORK( 1 )
      END IF
*
      BBNRM = CLANGE( '1', N, N, B, LDB, RWORK( 1 ) )
      IF( ILBSCL ) THEN
         RWORK( 1 ) = BBNRM
         CALL SLASCL( 'G'00, BNRMTO, BNRM, 11, RWORK( 1 ), 1,
     $                IERR )
         BBNRM = RWORK( 1 )
      END IF
*
*     Reduce B to triangular form (QR decomposition of B)
*     (Complex Workspace: need N, prefer N*NB )
*
      IROWS = IHI + 1 - ILO
      IF( ILV .OR. .NOT.WANTSN ) THEN
         ICOLS = N + 1 - ILO
      ELSE
         ICOLS = IROWS
      END IF
      ITAU = 1
      IWRK = ITAU + IROWS
      CALL CGEQRF( IROWS, ICOLS, B( ILO, ILO ), LDB, WORK( ITAU ),
     $             WORK( IWRK ), LWORK+1-IWRK, IERR )
*
*     Apply the unitary transformation to A
*     (Complex Workspace: need N, prefer N*NB)
*
      CALL CUNMQR( 'L''C', IROWS, ICOLS, IROWS, B( ILO, ILO ), LDB,
     $             WORK( ITAU ), A( ILO, ILO ), LDA, WORK( IWRK ),
     $             LWORK+1-IWRK, IERR )
*
*     Initialize VL and/or VR
*     (Workspace: need N, prefer N*NB)
*
      IF( ILVL ) THEN
         CALL CLASET( 'Full', N, N, CZERO, CONE, VL, LDVL )
         IF( IROWS.GT.1 ) THEN
            CALL CLACPY( 'L', IROWS-1, IROWS-1, B( ILO+1, ILO ), LDB,
     $                   VL( ILO+1, ILO ), LDVL )
         END IF
         CALL CUNGQR( IROWS, IROWS, IROWS, VL( ILO, ILO ), LDVL,
     $                WORK( ITAU ), WORK( IWRK ), LWORK+1-IWRK, IERR )
      END IF
*
      IF( ILVR )
     $   CALL CLASET( 'Full', N, N, CZERO, CONE, VR, LDVR )
*
*     Reduce to generalized Hessenberg form
*     (Workspace: none needed)
*
      IF( ILV .OR. .NOT.WANTSN ) THEN
*
*        Eigenvectors requested -- work on whole matrix.
*
         CALL CGGHRD( JOBVL, JOBVR, N, ILO, IHI, A, LDA, B, LDB, VL,
     $                LDVL, VR, LDVR, IERR )
      ELSE
         CALL CGGHRD( 'N''N', IROWS, 1, IROWS, A( ILO, ILO ), LDA,
     $                B( ILO, ILO ), LDB, VL, LDVL, VR, LDVR, IERR )
      END IF
*
*     Perform QZ algorithm (Compute eigenvalues, and optionally, the
*     Schur forms and Schur vectors)
*     (Complex Workspace: need N)
*     (Real Workspace: need N)
*
      IWRK = ITAU
      IF( ILV .OR. .NOT.WANTSN ) THEN
         CHTEMP = 'S'
      ELSE
         CHTEMP = 'E'
      END IF
*
      CALL CHGEQZ( CHTEMP, JOBVL, JOBVR, N, ILO, IHI, A, LDA, B, LDB,
     $             ALPHA, BETA, VL, LDVL, VR, LDVR, WORK( IWRK ),
     $             LWORK+1-IWRK, RWORK, IERR )
      IF( IERR.NE.0 ) THEN
         IF( IERR.GT.0 .AND. IERR.LE.N ) THEN
            INFO = IERR
         ELSE IF( IERR.GT..AND. IERR.LE.2*N ) THEN
            INFO = IERR - N
         ELSE
            INFO = N + 1
         END IF
         GO TO 90
      END IF
*
*     Compute Eigenvectors and estimate condition numbers if desired
*     CTGEVC: (Complex Workspace: need 2*N )
*             (Real Workspace:    need 2*N )
*     CTGSNA: (Complex Workspace: need 2*N*N if SENSE='V' or 'B')
*             (Integer Workspace: need N+2 )
*
      IF( ILV .OR. .NOT.WANTSN ) THEN
         IF( ILV ) THEN
            IF( ILVL ) THEN
               IF( ILVR ) THEN
                  CHTEMP = 'B'
               ELSE
                  CHTEMP = 'L'
               END IF
            ELSE
               CHTEMP = 'R'
            END IF
*
            CALL CTGEVC( CHTEMP, 'B', LDUMMA, N, A, LDA, B, LDB, VL,
     $                   LDVL, VR, LDVR, N, IN, WORK( IWRK ), RWORK,
     $                   IERR )
            IF( IERR.NE.0 ) THEN
               INFO = N + 2
               GO TO 90
            END IF
         END IF
*
         IF.NOT.WANTSN ) THEN
*
*           compute eigenvectors (STGEVC) and estimate condition
*           numbers (STGSNA). Note that the definition of the condition
*           number is not invariant under transformation (u,v) to
*           (Q*u, Z*v), where (u,v) are eigenvectors of the generalized
*           Schur form (S,T), Q and Z are orthogonal matrices. In order
*           to avoid using extra 2*N*N workspace, we have to
*           re-calculate eigenvectors and estimate the condition numbers
*           one at a time.
*
            DO 20 I = 1, N
*
               DO 10 J = 1, N
                  BWORK( J ) = .FALSE.
   10          CONTINUE
               BWORK( I ) = .TRUE.
*
               IWRK = N + 1
               IWRK1 = IWRK + N
*
               IF( WANTSE .OR. WANTSB ) THEN
                  CALL CTGEVC( 'B''S', BWORK, N, A, LDA, B, LDB,
     $                         WORK( 1 ), N, WORK( IWRK ), N, 1, M,
     $                         WORK( IWRK1 ), RWORK, IERR )
                  IF( IERR.NE.0 ) THEN
                     INFO = N + 2
                     GO TO 90
                  END IF
               END IF
*
               CALL CTGSNA( SENSE, 'S', BWORK, N, A, LDA, B, LDB,
     $                      WORK( 1 ), N, WORK( IWRK ), N, RCONDE( I ),
     $                      RCONDV( I ), 1, M, WORK( IWRK1 ),
     $                      LWORK-IWRK1+1, IWORK, IERR )
*
   20       CONTINUE
         END IF
      END IF
*
*     Undo balancing on VL and VR and normalization
*     (Workspace: none needed)
*
      IF( ILVL ) THEN
         CALL CGGBAK( BALANC, 'L', N, ILO, IHI, LSCALE, RSCALE, N, VL,
     $                LDVL, IERR )
*
         DO 50 JC = 1, N
            TEMP = ZERO
            DO 30 JR = 1, N
               TEMP = MAX( TEMP, ABS1( VL( JR, JC ) ) )
   30       CONTINUE
            IF( TEMP.LT.SMLNUM )
     $         GO TO 50
            TEMP = ONE / TEMP
            DO 40 JR = 1, N
               VL( JR, JC ) = VL( JR, JC )*TEMP
   40       CONTINUE
   50    CONTINUE
      END IF
*
      IF( ILVR ) THEN
         CALL CGGBAK( BALANC, 'R', N, ILO, IHI, LSCALE, RSCALE, N, VR,
     $                LDVR, IERR )
         DO 80 JC = 1, N
            TEMP = ZERO
            DO 60 JR = 1, N
               TEMP = MAX( TEMP, ABS1( VR( JR, JC ) ) )
   60       CONTINUE
            IF( TEMP.LT.SMLNUM )
     $         GO TO 80
            TEMP = ONE / TEMP
            DO 70 JR = 1, N
               VR( JR, JC ) = VR( JR, JC )*TEMP
   70       CONTINUE
   80    CONTINUE
      END IF
*
*     Undo scaling if necessary
*
      IF( ILASCL )
     $   CALL CLASCL( 'G'00, ANRMTO, ANRM, N, 1, ALPHA, N, IERR )
*
      IF( ILBSCL )
     $   CALL CLASCL( 'G'00, BNRMTO, BNRM, N, 1, BETA, N, IERR )
*
   90 CONTINUE
      WORK( 1 ) = MAXWRK
*
      RETURN
*
*     End of CGGEVX
*
      END