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
      SUBROUTINE ZHEEVR( JOBZ, RANGE, UPLO, N, A, LDA, VL, VU, IL, IU,
     $                   ABSTOL, M, W, Z, LDZ, ISUPPZ, WORK, LWORK,
     $                   RWORK, LRWORK, IWORK, LIWORK, INFO )
*
*  -- LAPACK driver routine (version 3.3.1) --
*  -- LAPACK is a software package provided by Univ. of Tennessee,    --
*  -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
*  -- April 2011                                                      --
*
*     .. Scalar Arguments ..
      CHARACTER          JOBZ, RANGE, UPLO
      INTEGER            IL, INFO, IU, LDA, LDZ, LIWORK, LRWORK, LWORK,
     $                   M, N
      DOUBLE PRECISION   ABSTOL, VL, VU
*     ..
*     .. Array Arguments ..
      INTEGER            ISUPPZ( * ), IWORK( * )
      DOUBLE PRECISION   RWORK( * ), W( * )
      COMPLEX*16         A( LDA, * ), WORK( * ), Z( LDZ, * )
*     ..
*
*  Purpose
*  =======
*
*  ZHEEVR computes selected eigenvalues and, optionally, eigenvectors
*  of a complex Hermitian matrix A.  Eigenvalues and eigenvectors can
*  be selected by specifying either a range of values or a range of
*  indices for the desired eigenvalues.
*
*  ZHEEVR first reduces the matrix A to tridiagonal form T with a call
*  to ZHETRD.  Then, whenever possible, ZHEEVR calls ZSTEMR to compute
*  eigenspectrum using Relatively Robust Representations.  ZSTEMR
*  computes eigenvalues by the dqds algorithm, while orthogonal
*  eigenvectors are computed from various "good" L D L^T representations
*  (also known as Relatively Robust Representations). Gram-Schmidt
*  orthogonalization is avoided as far as possible. More specifically,
*  the various steps of the algorithm are as follows.
*
*  For each unreduced block (submatrix) of T,
*     (a) Compute T - sigma I  = L D L^T, so that L and D
*         define all the wanted eigenvalues to high relative accuracy.
*         This means that small relative changes in the entries of D and L
*         cause only small relative changes in the eigenvalues and
*         eigenvectors. The standard (unfactored) representation of the
*         tridiagonal matrix T does not have this property in general.
*     (b) Compute the eigenvalues to suitable accuracy.
*         If the eigenvectors are desired, the algorithm attains full
*         accuracy of the computed eigenvalues only right before
*         the corresponding vectors have to be computed, see steps c) and d).
*     (c) For each cluster of close eigenvalues, select a new
*         shift close to the cluster, find a new factorization, and refine
*         the shifted eigenvalues to suitable accuracy.
*     (d) For each eigenvalue with a large enough relative separation compute
*         the corresponding eigenvector by forming a rank revealing twisted
*         factorization. Go back to (c) for any clusters that remain.
*
*  The desired accuracy of the output can be specified by the input
*  parameter ABSTOL.
*
*  For more details, see DSTEMR's documentation and:
*  - Inderjit S. Dhillon and Beresford N. Parlett: "Multiple representations
*    to compute orthogonal eigenvectors of symmetric tridiagonal matrices,"
*    Linear Algebra and its Applications, 387(1), pp. 1-28, August 2004.
*  - Inderjit Dhillon and Beresford Parlett: "Orthogonal Eigenvectors and
*    Relative Gaps," SIAM Journal on Matrix Analysis and Applications, Vol. 25,
*    2004.  Also LAPACK Working Note 154.
*  - Inderjit Dhillon: "A new O(n^2) algorithm for the symmetric
*    tridiagonal eigenvalue/eigenvector problem",
*    Computer Science Division Technical Report No. UCB/CSD-97-971,
*    UC Berkeley, May 1997.
*
*
*  Note 1 : ZHEEVR calls ZSTEMR when the full spectrum is requested
*  on machines which conform to the ieee-754 floating point standard.
*  ZHEEVR calls DSTEBZ and ZSTEIN on non-ieee machines and
*  when partial spectrum requests are made.
*
*  Normal execution of ZSTEMR may create NaNs and infinities and
*  hence may abort due to a floating point exception in environments
*  which do not handle NaNs and infinities in the ieee standard default
*  manner.
*
*  Arguments
*  =========
*
*  JOBZ    (input) CHARACTER*1
*          = 'N':  Compute eigenvalues only;
*          = 'V':  Compute eigenvalues and eigenvectors.
*
*  RANGE   (input) CHARACTER*1
*          = 'A': all eigenvalues will be found.
*          = 'V': all eigenvalues in the half-open interval (VL,VU]
*                 will be found.
*          = 'I': the IL-th through IU-th eigenvalues will be found.
*          For RANGE = 'V' or 'I' and IU - IL < N - 1, DSTEBZ and
*          ZSTEIN are called
*
*  UPLO    (input) CHARACTER*1
*          = 'U':  Upper triangle of A is stored;
*          = 'L':  Lower triangle of A is stored.
*
*  N       (input) INTEGER
*          The order of the matrix A.  N >= 0.
*
*  A       (input/output) COMPLEX*16 array, dimension (LDA, N)
*          On entry, the Hermitian matrix A.  If UPLO = 'U', the
*          leading N-by-N upper triangular part of A contains the
*          upper triangular part of the matrix A.  If UPLO = 'L',
*          the leading N-by-N lower triangular part of A contains
*          the lower triangular part of the matrix A.
*          On exit, the lower triangle (if UPLO='L') or the upper
*          triangle (if UPLO='U') of A, including the diagonal, is
*          destroyed.
*
*  LDA     (input) INTEGER
*          The leading dimension of the array A.  LDA >= max(1,N).
*
*  VL      (input) DOUBLE PRECISION
*  VU      (input) DOUBLE PRECISION
*          If RANGE='V', the lower and upper bounds of the interval to
*          be searched for eigenvalues. VL < VU.
*          Not referenced if RANGE = 'A' or 'I'.
*
*  IL      (input) INTEGER
*  IU      (input) 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'.
*
*  ABSTOL  (input) DOUBLE PRECISION
*          The absolute error tolerance for the eigenvalues.
*          An approximate eigenvalue is accepted as converged
*          when it is determined to lie in an interval [a,b]
*          of width less than or equal to
*
*                  ABSTOL + EPS *   max( |a|,|b| ) ,
*
*          where EPS is the machine precision.  If ABSTOL is less than
*          or equal to zero, then  EPS*|T|  will be used in its place,
*          where |T| is the 1-norm of the tridiagonal matrix obtained
*          by reducing A to tridiagonal form.
*
*          See "Computing Small Singular Values of Bidiagonal Matrices
*          with Guaranteed High Relative Accuracy," by Demmel and
*          Kahan, LAPACK Working Note #3.
*
*          If high relative accuracy is important, set ABSTOL to
*          DLAMCH( 'Safe minimum' ).  Doing so will guarantee that
*          eigenvalues are computed to high relative accuracy when
*          possible in future releases.  The current code does not
*          make any guarantees about high relative accuracy, but
*          furutre releases will. See J. Barlow and J. Demmel,
*          "Computing Accurate Eigensystems of Scaled Diagonally
*          Dominant Matrices", LAPACK Working Note #7, for a discussion
*          of which matrices define their eigenvalues to high relative
*          accuracy.
*
*  M       (output) INTEGER
*          The total number of eigenvalues found.  0 <= M <= N.
*          If RANGE = 'A', M = N, and if RANGE = 'I', M = IU-IL+1.
*
*  W       (output) DOUBLE PRECISION array, dimension (N)
*          The first M elements contain the selected eigenvalues in
*          ascending order.
*
*  Z       (output) COMPLEX*16 array, dimension (LDZ, max(1,M))
*          If JOBZ = 'V', then if INFO = 0, the first M columns of Z
*          contain the orthonormal eigenvectors of the matrix A
*          corresponding to the selected eigenvalues, with the i-th
*          column of Z holding the eigenvector associated with W(i).
*          If JOBZ = 'N', then Z is not referenced.
*          Note: the user must ensure that at least max(1,M) columns are
*          supplied in the array Z; if RANGE = 'V', the exact value of M
*          is not known in advance and an upper bound must be used.
*
*  LDZ     (input) INTEGER
*          The leading dimension of the array Z.  LDZ >= 1, and if
*          JOBZ = 'V', LDZ >= max(1,N).
*
*  ISUPPZ  (output) INTEGER array, dimension ( 2*max(1,M) )
*          The support of the eigenvectors in Z, i.e., the indices
*          indicating the nonzero elements in Z. The i-th eigenvector
*          is nonzero only in elements ISUPPZ( 2*i-1 ) through
*          ISUPPZ( 2*i ).
*          Implemented only for RANGE = 'A' or 'I' and IU - IL = N - 1
*
*  WORK    (workspace/output) COMPLEX*16 array, dimension (MAX(1,LWORK))
*          On exit, if INFO = 0, WORK(1) returns the optimal LWORK.
*
*  LWORK   (input) INTEGER
*          The length of the array WORK.  LWORK >= max(1,2*N).
*          For optimal efficiency, LWORK >= (NB+1)*N,
*          where NB is the max of the blocksize for ZHETRD and for
*          ZUNMTR as returned by ILAENV.
*
*          If LWORK = -1, then a workspace query is assumed; the routine
*          only calculates the optimal sizes of the WORK, RWORK and
*          IWORK arrays, returns these values as the first entries of
*          the WORK, RWORK and IWORK arrays, and no error message
*          related to LWORK or LRWORK or LIWORK is issued by XERBLA.
*
*  RWORK   (workspace/output) DOUBLE PRECISION array, dimension (MAX(1,LRWORK))
*          On exit, if INFO = 0, RWORK(1) returns the optimal
*          (and minimal) LRWORK.
*
*  LRWORK   (input) INTEGER
*          The length of the array RWORK.  LRWORK >= max(1,24*N).
*
*          If LRWORK = -1, then a workspace query is assumed; the
*          routine only calculates the optimal sizes of the WORK, RWORK
*          and IWORK arrays, returns these values as the first entries
*          of the WORK, RWORK and IWORK arrays, and no error message
*          related to LWORK or LRWORK or LIWORK is issued by XERBLA.
*
*  IWORK   (workspace/output) INTEGER array, dimension (MAX(1,LIWORK))
*          On exit, if INFO = 0, IWORK(1) returns the optimal
*          (and minimal) LIWORK.
*
*  LIWORK   (input) INTEGER
*          The dimension of the array IWORK.  LIWORK >= max(1,10*N).
*
*          If LIWORK = -1, then a workspace query is assumed; the
*          routine only calculates the optimal sizes of the WORK, RWORK
*          and IWORK arrays, returns these values as the first entries
*          of the WORK, RWORK and IWORK arrays, and no error message
*          related to LWORK or LRWORK or LIWORK is issued by XERBLA.
*
*  INFO    (output) INTEGER
*          = 0:  successful exit
*          < 0:  if INFO = -i, the i-th argument had an illegal value
*          > 0:  Internal error
*
*  Further Details
*  ===============
*
*  Based on contributions by
*     Inderjit Dhillon, IBM Almaden, USA
*     Osni Marques, LBNL/NERSC, USA
*     Ken Stanley, Computer Science Division, University of
*       California at Berkeley, USA
*     Jason Riedy, Computer Science Division, University of
*       California at Berkeley, USA
*
*  =====================================================================
*
*     .. Parameters ..
      DOUBLE PRECISION   ZERO, ONE, TWO
      PARAMETER          ( ZERO = 0.0D+0, ONE = 1.0D+0, TWO = 2.0D+0 )
*     ..
*     .. Local Scalars ..
      LOGICAL            ALLEIG, INDEIG, LOWER, LQUERY, TEST, VALEIG,
     $                   WANTZ, TRYRAC
      CHARACTER          ORDER
      INTEGER            I, IEEEOK, IINFO, IMAX, INDIBL, INDIFL, INDISP,
     $                   INDIWO, INDRD, INDRDD, INDRE, INDREE, INDRWK,
     $                   INDTAU, INDWK, INDWKN, ISCALE, ITMP1, J, JJ,
     $                   LIWMIN, LLWORK, LLRWORK, LLWRKN, LRWMIN,
     $                   LWKOPT, LWMIN, NB, NSPLIT
      DOUBLE PRECISION   ABSTLL, ANRM, BIGNUM, EPS, RMAX, RMIN, SAFMIN,
     $                   SIGMA, SMLNUM, TMP1, VLL, VUU
*     ..
*     .. External Functions ..
      LOGICAL            LSAME
      INTEGER            ILAENV
      DOUBLE PRECISION   DLAMCH, ZLANSY
      EXTERNAL           LSAME, ILAENV, DLAMCH, ZLANSY
*     ..
*     .. External Subroutines ..
      EXTERNAL           DCOPY, DSCAL, DSTEBZ, DSTERF, XERBLA, ZDSCAL,
     $                   ZHETRD, ZSTEMR, ZSTEIN, ZSWAP, ZUNMTR
*     ..
*     .. Intrinsic Functions ..
      INTRINSIC          DBLEMAXMINSQRT
*     ..
*     .. Executable Statements ..
*
*     Test the input parameters.
*
      IEEEOK = ILAENV( 10'ZHEEVR''N'1234 )
*
      LOWER = LSAME( UPLO, 'L' )
      WANTZ = LSAME( JOBZ, 'V' )
      ALLEIG = LSAME( RANGE'A' )
      VALEIG = LSAME( RANGE'V' )
      INDEIG = LSAME( RANGE'I' )
*
      LQUERY = ( ( LWORK.EQ.-1 ) .OR. ( LRWORK.EQ.-1 ) .OR.
     $         ( LIWORK.EQ.-1 ) )
*
      LRWMIN = MAX124*N )
      LIWMIN = MAX110*N )
      LWMIN = MAX12*N )
*
      INFO = 0
      IF.NOT.( WANTZ .OR. LSAME( JOBZ, 'N' ) ) ) THEN
         INFO = -1
      ELSE IF.NOT.( ALLEIG .OR. VALEIG .OR. INDEIG ) ) THEN
         INFO = -2
      ELSE IF.NOT.( LOWER .OR. LSAME( UPLO, 'U' ) ) ) THEN
         INFO = -3
      ELSE IF( N.LT.0 ) THEN
         INFO = -4
      ELSE IF( LDA.LT.MAX1, N ) ) THEN
         INFO = -6
      ELSE
         IF( VALEIG ) THEN
            IF( N.GT.0 .AND. VU.LE.VL )
     $         INFO = -8
         ELSE IF( INDEIG ) THEN
            IF( IL.LT.1 .OR. IL.GT.MAX1, N ) ) THEN
               INFO = -9
            ELSE IF( IU.LT.MIN( N, IL ) .OR. IU.GT.N ) THEN
               INFO = -10
            END IF
         END IF
      END IF
      IF( INFO.EQ.0 ) THEN
         IF( LDZ.LT.1 .OR. ( WANTZ .AND. LDZ.LT.N ) ) THEN
            INFO = -15
         END IF
      END IF
*
      IF( INFO.EQ.0 ) THEN
         NB = ILAENV( 1'ZHETRD', UPLO, N, -1-1-1 )
         NB = MAX( NB, ILAENV( 1'ZUNMTR', UPLO, N, -1-1-1 ) )
         LWKOPT = MAX( ( NB+1 )*N, LWMIN )
         WORK( 1 ) = LWKOPT
         RWORK( 1 ) = LRWMIN
         IWORK( 1 ) = LIWMIN
*
         IF( LWORK.LT.LWMIN .AND. .NOT.LQUERY ) THEN
            INFO = -18
         ELSE IF( LRWORK.LT.LRWMIN .AND. .NOT.LQUERY ) THEN
            INFO = -20
         ELSE IF( LIWORK.LT.LIWMIN .AND. .NOT.LQUERY ) THEN
            INFO = -22
         END IF
      END IF
*
      IF( INFO.NE.0 ) THEN
         CALL XERBLA( 'ZHEEVR'-INFO )
         RETURN
      ELSE IF( LQUERY ) THEN
         RETURN
      END IF
*
*     Quick return if possible
*
      M = 0
      IF( N.EQ.0 ) THEN
         WORK( 1 ) = 1
         RETURN
      END IF
*
      IF( N.EQ.1 ) THEN
         WORK( 1 ) = 2
         IF( ALLEIG .OR. INDEIG ) THEN
            M = 1
            W( 1 ) = DBLE( A( 11 ) )
         ELSE
            IF( VL.LT.DBLE( A( 11 ) ) .AND. VU.GE.DBLE( A( 11 ) ) )
     $           THEN
               M = 1
               W( 1 ) = DBLE( A( 11 ) )
            END IF
         END IF
         IF( WANTZ ) THEN
            Z( 11 ) = ONE
            ISUPPZ( 1 ) = 1
            ISUPPZ( 2 ) = 1
         END IF
         RETURN
      END IF
*
*     Get machine constants.
*
      SAFMIN = DLAMCH( 'Safe minimum' )
      EPS = DLAMCH( 'Precision' )
      SMLNUM = SAFMIN / EPS
      BIGNUM = ONE / SMLNUM
      RMIN = SQRT( SMLNUM )
      RMAX = MINSQRT( BIGNUM ), ONE / SQRTSQRT( SAFMIN ) ) )
*
*     Scale matrix to allowable range, if necessary.
*
      ISCALE = 0
      ABSTLL = ABSTOL
      IF (VALEIG) THEN
         VLL = VL
         VUU = VU
      END IF
      ANRM = ZLANSY( 'M', UPLO, N, A, LDA, RWORK )
      IF( ANRM.GT.ZERO .AND. ANRM.LT.RMIN ) THEN
         ISCALE = 1
         SIGMA = RMIN / ANRM
      ELSE IF( ANRM.GT.RMAX ) THEN
         ISCALE = 1
         SIGMA = RMAX / ANRM
      END IF
      IF( ISCALE.EQ.1 ) THEN
         IF( LOWER ) THEN
            DO 10 J = 1, N
               CALL ZDSCAL( N-J+1, SIGMA, A( J, J ), 1 )
   10       CONTINUE
         ELSE
            DO 20 J = 1, N
               CALL ZDSCAL( J, SIGMA, A( 1, J ), 1 )
   20       CONTINUE
         END IF
         IF( ABSTOL.GT.0 )
     $      ABSTLL = ABSTOL*SIGMA
         IF( VALEIG ) THEN
            VLL = VL*SIGMA
            VUU = VU*SIGMA
         END IF
      END IF

*     Initialize indices into workspaces.  Note: The IWORK indices are
*     used only if DSTERF or ZSTEMR fail.

*     WORK(INDTAU:INDTAU+N-1) stores the complex scalar factors of the
*     elementary reflectors used in ZHETRD.
      INDTAU = 1
*     INDWK is the starting offset of the remaining complex workspace,
*     and LLWORK is the remaining complex workspace size.
      INDWK = INDTAU + N
      LLWORK = LWORK - INDWK + 1

*     RWORK(INDRD:INDRD+N-1) stores the real tridiagonal's diagonal
*     entries.
      INDRD = 1
*     RWORK(INDRE:INDRE+N-1) stores the off-diagonal entries of the
*     tridiagonal matrix from ZHETRD.
      INDRE = INDRD + N
*     RWORK(INDRDD:INDRDD+N-1) is a copy of the diagonal entries over
*     -written by ZSTEMR (the DSTERF path copies the diagonal to W).
      INDRDD = INDRE + N
*     RWORK(INDREE:INDREE+N-1) is a copy of the off-diagonal entries over
*     -written while computing the eigenvalues in DSTERF and ZSTEMR.
      INDREE = INDRDD + N
*     INDRWK is the starting offset of the left-over real workspace, and
*     LLRWORK is the remaining workspace size.
      INDRWK = INDREE + N
      LLRWORK = LRWORK - INDRWK + 1

*     IWORK(INDIBL:INDIBL+M-1) corresponds to IBLOCK in DSTEBZ and
*     stores the block indices of each of the M<=N eigenvalues.
      INDIBL = 1
*     IWORK(INDISP:INDISP+NSPLIT-1) corresponds to ISPLIT in DSTEBZ and
*     stores the starting and finishing indices of each block.
      INDISP = INDIBL + N
*     IWORK(INDIFL:INDIFL+N-1) stores the indices of eigenvectors
*     that corresponding to eigenvectors that fail to converge in
*     DSTEIN.  This information is discarded; if any fail, the driver
*     returns INFO > 0.
      INDIFL = INDISP + N
*     INDIWO is the offset of the remaining integer workspace.
      INDIWO = INDISP + N

*
*     Call ZHETRD to reduce Hermitian matrix to tridiagonal form.
*
      CALL ZHETRD( UPLO, N, A, LDA, RWORK( INDRD ), RWORK( INDRE ),
     $             WORK( INDTAU ), WORK( INDWK ), LLWORK, IINFO )
*
*     If all eigenvalues are desired
*     then call DSTERF or ZSTEMR and ZUNMTR.
*
      TEST = .FALSE.
      IF( INDEIG ) THEN
         IF( IL.EQ.1 .AND. IU.EQ.N ) THEN
            TEST = .TRUE.
         END IF
      END IF
      IF( ( ALLEIG.OR.TEST ) .AND. ( IEEEOK.EQ.1 ) ) THEN
         IF.NOT.WANTZ ) THEN
            CALL DCOPY( N, RWORK( INDRD ), 1, W, 1 )
            CALL DCOPY( N-1, RWORK( INDRE ), 1, RWORK( INDREE ), 1 )
            CALL DSTERF( N, W, RWORK( INDREE ), INFO )
         ELSE
            CALL DCOPY( N-1, RWORK( INDRE ), 1, RWORK( INDREE ), 1 )
            CALL DCOPY( N, RWORK( INDRD ), 1, RWORK( INDRDD ), 1 )
*
            IF (ABSTOL .LE. TWO*N*EPS) THEN
               TRYRAC = .TRUE.
            ELSE
               TRYRAC = .FALSE.
            END IF
            CALL ZSTEMR( JOBZ, 'A', N, RWORK( INDRDD ),
     $                   RWORK( INDREE ), VL, VU, IL, IU, M, W,
     $                   Z, LDZ, N, ISUPPZ, TRYRAC,
     $                   RWORK( INDRWK ), LLRWORK,
     $                   IWORK, LIWORK, INFO )
*
*           Apply unitary matrix used in reduction to tridiagonal
*           form to eigenvectors returned by ZSTEIN.
*
            IF( WANTZ .AND. INFO.EQ.0 ) THEN
               INDWKN = INDWK
               LLWRKN = LWORK - INDWKN + 1
               CALL ZUNMTR( 'L', UPLO, 'N', N, M, A, LDA,
     $                      WORK( INDTAU ), Z, LDZ, WORK( INDWKN ),
     $                      LLWRKN, IINFO )
            END IF
         END IF
*
*
         IF( INFO.EQ.0 ) THEN
            M = N
            GO TO 30
         END IF
         INFO = 0
      END IF
*
*     Otherwise, call DSTEBZ and, if eigenvectors are desired, ZSTEIN.
*     Also call DSTEBZ and ZSTEIN if ZSTEMR fails.
*
      IF( WANTZ ) THEN
         ORDER = 'B'
      ELSE
         ORDER = 'E'
      END IF

      CALL DSTEBZ( RANGE, ORDER, N, VLL, VUU, IL, IU, ABSTLL,
     $             RWORK( INDRD ), RWORK( INDRE ), M, NSPLIT, W,
     $             IWORK( INDIBL ), IWORK( INDISP ), RWORK( INDRWK ),
     $             IWORK( INDIWO ), INFO )
*
      IF( WANTZ ) THEN
         CALL ZSTEIN( N, RWORK( INDRD ), RWORK( INDRE ), M, W,
     $                IWORK( INDIBL ), IWORK( INDISP ), Z, LDZ,
     $                RWORK( INDRWK ), IWORK( INDIWO ), IWORK( INDIFL ),
     $                INFO )
*
*        Apply unitary matrix used in reduction to tridiagonal
*        form to eigenvectors returned by ZSTEIN.
*
         INDWKN = INDWK
         LLWRKN = LWORK - INDWKN + 1
         CALL ZUNMTR( 'L', UPLO, 'N', N, M, A, LDA, WORK( INDTAU ), Z,
     $                LDZ, WORK( INDWKN ), LLWRKN, IINFO )
      END IF
*
*     If matrix was scaled, then rescale eigenvalues appropriately.
*
   30 CONTINUE
      IF( ISCALE.EQ.1 ) THEN
         IF( INFO.EQ.0 ) THEN
            IMAX = M
         ELSE
            IMAX = INFO - 1
         END IF
         CALL DSCAL( IMAX, ONE / SIGMA, W, 1 )
      END IF
*
*     If eigenvalues are not in order, then sort them, along with
*     eigenvectors.
*
      IF( WANTZ ) THEN
         DO 50 J = 1, M - 1
            I = 0
            TMP1 = W( J )
            DO 40 JJ = J + 1, M
               IF( W( JJ ).LT.TMP1 ) THEN
                  I = JJ
                  TMP1 = W( JJ )
               END IF
   40       CONTINUE
*
            IF( I.NE.0 ) THEN
               ITMP1 = IWORK( INDIBL+I-1 )
               W( I ) = W( J )
               IWORK( INDIBL+I-1 ) = IWORK( INDIBL+J-1 )
               W( J ) = TMP1
               IWORK( INDIBL+J-1 ) = ITMP1
               CALL ZSWAP( N, Z( 1, I ), 1, Z( 1, J ), 1 )
            END IF
   50    CONTINUE
      END IF
*
*     Set WORK(1) to optimal workspace size.
*
      WORK( 1 ) = LWKOPT
      RWORK( 1 ) = LRWMIN
      IWORK( 1 ) = LIWMIN
*
      RETURN
*
*     End of ZHEEVR
*
      END