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
      SUBROUTINE SLAHQR( WANTT, WANTZ, N, ILO, IHI, H, LDH, WR, WI,
     $                   ILOZ, IHIZ, Z, LDZ, INFO )
*
*  -- LAPACK auxiliary routine (version 3.2) --
*     Univ. of Tennessee, Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..
*     November 2006
*
*     .. Scalar Arguments ..
      INTEGER            IHI, IHIZ, ILO, ILOZ, INFO, LDH, LDZ, N
      LOGICAL            WANTT, WANTZ
*     ..
*     .. Array Arguments ..
      REAL               H( LDH, * ), WI( * ), WR( * ), Z( LDZ, * )
*     ..
*
*     Purpose
*     =======
*
*     SLAHQR is an auxiliary routine called by SHSEQR to update the
*     eigenvalues and Schur decomposition already computed by SHSEQR, by
*     dealing with the Hessenberg submatrix in rows and columns ILO to
*     IHI.
*
*     Arguments
*     =========
*
*     WANTT   (input) LOGICAL
*          = .TRUE. : the full Schur form T is required;
*          = .FALSE.: only eigenvalues are required.
*
*     WANTZ   (input) LOGICAL
*          = .TRUE. : the matrix of Schur vectors Z is required;
*          = .FALSE.: Schur vectors are not required.
*
*     N       (input) INTEGER
*          The order of the matrix H.  N >= 0.
*
*     ILO     (input) INTEGER
*     IHI     (input) INTEGER
*          It is assumed that H is already upper quasi-triangular in
*          rows and columns IHI+1:N, and that H(ILO,ILO-1) = 0 (unless
*          ILO = 1). SLAHQR works primarily with the Hessenberg
*          submatrix in rows and columns ILO to IHI, but applies
*          transformations to all of H if WANTT is .TRUE..
*          1 <= ILO <= max(1,IHI); IHI <= N.
*
*     H       (input/output) REAL array, dimension (LDH,N)
*          On entry, the upper Hessenberg matrix H.
*          On exit, if INFO is zero and if WANTT is .TRUE., H is upper
*          quasi-triangular in rows and columns ILO:IHI, with any
*          2-by-2 diagonal blocks in standard form. If INFO is zero
*          and WANTT is .FALSE., the contents of H are unspecified on
*          exit.  The output state of H if INFO is nonzero is given
*          below under the description of INFO.
*
*     LDH     (input) INTEGER
*          The leading dimension of the array H. LDH >= max(1,N).
*
*     WR      (output) REAL array, dimension (N)
*     WI      (output) REAL array, dimension (N)
*          The real and imaginary parts, respectively, of the computed
*          eigenvalues ILO to IHI are stored in the corresponding
*          elements of WR and WI. If two eigenvalues are computed as a
*          complex conjugate pair, they are stored in consecutive
*          elements of WR and WI, say the i-th and (i+1)th, with
*          WI(i) > 0 and WI(i+1) < 0. If WANTT is .TRUE., the
*          eigenvalues are stored in the same order as on the diagonal
*          of the Schur form returned in H, with WR(i) = H(i,i), and, if
*          H(i:i+1,i:i+1) is a 2-by-2 diagonal block,
*          WI(i) = sqrt(H(i+1,i)*H(i,i+1)) and WI(i+1) = -WI(i).
*
*     ILOZ    (input) INTEGER
*     IHIZ    (input) INTEGER
*          Specify the rows of Z to which transformations must be
*          applied if WANTZ is .TRUE..
*          1 <= ILOZ <= ILO; IHI <= IHIZ <= N.
*
*     Z       (input/output) REAL array, dimension (LDZ,N)
*          If WANTZ is .TRUE., on entry Z must contain the current
*          matrix Z of transformations accumulated by SHSEQR, and on
*          exit Z has been updated; transformations are applied only to
*          the submatrix Z(ILOZ:IHIZ,ILO:IHI).
*          If WANTZ is .FALSE., Z is not referenced.
*
*     LDZ     (input) INTEGER
*          The leading dimension of the array Z. LDZ >= max(1,N).
*
*     INFO    (output) INTEGER
*           =   0: successful exit
*          .GT. 0: If INFO = i, SLAHQR failed to compute all the
*                  eigenvalues ILO to IHI in a total of 30 iterations
*                  per eigenvalue; elements i+1:ihi of WR and WI
*                  contain those eigenvalues which have been
*                  successfully computed.
*
*                  If INFO .GT. 0 and WANTT is .FALSE., then on exit,
*                  the remaining unconverged eigenvalues are the
*                  eigenvalues of the upper Hessenberg matrix rows
*                  and columns ILO thorugh INFO of the final, output
*                  value of H.
*
*                  If INFO .GT. 0 and WANTT is .TRUE., then on exit
*          (*)       (initial value of H)*U  = U*(final value of H)
*                  where U is an orthognal matrix.    The final
*                  value of H is upper Hessenberg and triangular in
*                  rows and columns INFO+1 through IHI.
*
*                  If INFO .GT. 0 and WANTZ is .TRUE., then on exit
*                      (final value of Z)  = (initial value of Z)*U
*                  where U is the orthogonal matrix in (*)
*                  (regardless of the value of WANTT.)
*
*     Further Details
*     ===============
*
*     02-96 Based on modifications by
*     David Day, Sandia National Laboratory, USA
*
*     12-04 Further modifications by
*     Ralph Byers, University of Kansas, USA
*     This is a modified version of SLAHQR from LAPACK version 3.0.
*     It is (1) more robust against overflow and underflow and
*     (2) adopts the more conservative Ahues & Tisseur stopping
*     criterion (LAWN 122, 1997).
*
*     =========================================================
*
*     .. Parameters ..
      INTEGER            ITMAX
      PARAMETER          ( ITMAX = 30 )
      REAL               ZERO, ONE, TWO
      PARAMETER          ( ZERO = 0.0e0, ONE = 1.0e0, TWO = 2.0e0 )
      REAL               DAT1, DAT2
      PARAMETER          ( DAT1 = 3.0e0 / 4.0e0, DAT2 = -0.4375e0 )
*     ..
*     .. Local Scalars ..
      REAL               AA, AB, BA, BB, CS, DET, H11, H12, H21, H21S,
     $                   H22, RT1I, RT1R, RT2I, RT2R, RTDISC, S, SAFMAX,
     $                   SAFMIN, SMLNUM, SN, SUM, T1, T2, T3, TR, TST,
     $                   ULP, V2, V3
      INTEGER            I, I1, I2, ITS, J, K, L, M, NH, NR, NZ
*     ..
*     .. Local Arrays ..
      REAL               V( 3 )
*     ..
*     .. External Functions ..
      REAL               SLAMCH
      EXTERNAL           SLAMCH
*     ..
*     .. External Subroutines ..
      EXTERNAL           SCOPY, SLABAD, SLANV2, SLARFG, SROT
*     ..
*     .. Intrinsic Functions ..
      INTRINSIC          ABSMAXMIN, REAL, SQRT
*     ..
*     .. Executable Statements ..
*
      INFO = 0
*
*     Quick return if possible
*
      IF( N.EQ.0 )
     $   RETURN
      IF( ILO.EQ.IHI ) THEN
         WR( ILO ) = H( ILO, ILO )
         WI( ILO ) = ZERO
         RETURN
      END IF
*
*     ==== clear out the trash ====
      DO 10 J = ILO, IHI - 3
         H( J+2, J ) = ZERO
         H( J+3, J ) = ZERO
   10 CONTINUE
      IF( ILO.LE.IHI-2 )
     $   H( IHI, IHI-2 ) = ZERO
*
      NH = IHI - ILO + 1
      NZ = IHIZ - ILOZ + 1
*
*     Set machine-dependent constants for the stopping criterion.
*
      SAFMIN = SLAMCH( 'SAFE MINIMUM' )
      SAFMAX = ONE / SAFMIN
      CALL SLABAD( SAFMIN, SAFMAX )
      ULP = SLAMCH( 'PRECISION' )
      SMLNUM = SAFMIN*REAL( NH ) / ULP )
*
*     I1 and I2 are the indices of the first row and last column of H
*     to which transformations must be applied. If eigenvalues only are
*     being computed, I1 and I2 are set inside the main loop.
*
      IF( WANTT ) THEN
         I1 = 1
         I2 = N
      END IF
*
*     The main loop begins here. I is the loop index and decreases from
*     IHI to ILO in steps of 1 or 2. Each iteration of the loop works
*     with the active submatrix in rows and columns L to I.
*     Eigenvalues I+1 to IHI have already converged. Either L = ILO or
*     H(L,L-1) is negligible so that the matrix splits.
*
      I = IHI
   20 CONTINUE
      L = ILO
      IF( I.LT.ILO )
     $   GO TO 160
*
*     Perform QR iterations on rows and columns ILO to I until a
*     submatrix of order 1 or 2 splits off at the bottom because a
*     subdiagonal element has become negligible.
*
      DO 140 ITS = 0, ITMAX
*
*        Look for a single small subdiagonal element.
*
         DO 30 K = I, L + 1-1
            IFABS( H( K, K-1 ) ).LE.SMLNUM )
     $         GO TO 40
            TST = ABS( H( K-1, K-1 ) ) + ABS( H( K, K ) )
            IF( TST.EQ.ZERO ) THEN
               IF( K-2.GE.ILO )
     $            TST = TST + ABS( H( K-1, K-2 ) )
               IF( K+1.LE.IHI )
     $            TST = TST + ABS( H( K+1, K ) )
            END IF
*           ==== The following is a conservative small subdiagonal
*           .    deflation  criterion due to Ahues & Tisseur (LAWN 122,
*           .    1997). It has better mathematical foundation and
*           .    improves accuracy in some cases.  ====
            IFABS( H( K, K-1 ) ).LE.ULP*TST ) THEN
               AB = MAXABS( H( K, K-1 ) ), ABS( H( K-1, K ) ) )
               BA = MINABS( H( K, K-1 ) ), ABS( H( K-1, K ) ) )
               AA = MAXABS( H( K, K ) ),
     $              ABS( H( K-1, K-1 )-H( K, K ) ) )
               BB = MINABS( H( K, K ) ),
     $              ABS( H( K-1, K-1 )-H( K, K ) ) )
               S = AA + AB
               IF( BA*( AB / S ).LE.MAX( SMLNUM,
     $             ULP*( BB*( AA / S ) ) ) )GO TO 40
            END IF
   30    CONTINUE
   40    CONTINUE
         L = K
         IF( L.GT.ILO ) THEN
*
*           H(L,L-1) is negligible
*
            H( L, L-1 ) = ZERO
         END IF
*
*        Exit from loop if a submatrix of order 1 or 2 has split off.
*
         IF( L.GE.I-1 )
     $      GO TO 150
*
*        Now the active submatrix is in rows and columns L to I. If
*        eigenvalues only are being computed, only the active submatrix
*        need be transformed.
*
         IF.NOT.WANTT ) THEN
            I1 = L
            I2 = I
         END IF
*
         IF( ITS.EQ.10 ) THEN
*
*           Exceptional shift.
*
            S = ABS( H( L+1, L ) ) + ABS( H( L+2, L+1 ) )
            H11 = DAT1*+ H( L, L )
            H12 = DAT2*S
            H21 = S
            H22 = H11
         ELSE IF( ITS.EQ.20 ) THEN
*
*           Exceptional shift.
*
            S = ABS( H( I, I-1 ) ) + ABS( H( I-1, I-2 ) )
            H11 = DAT1*+ H( I, I )
            H12 = DAT2*S
            H21 = S
            H22 = H11
         ELSE
*
*           Prepare to use Francis' double shift
*           (i.e. 2nd degree generalized Rayleigh quotient)
*
            H11 = H( I-1, I-1 )
            H21 = H( I, I-1 )
            H12 = H( I-1, I )
            H22 = H( I, I )
         END IF
         S = ABS( H11 ) + ABS( H12 ) + ABS( H21 ) + ABS( H22 )
         IF( S.EQ.ZERO ) THEN
            RT1R = ZERO
            RT1I = ZERO
            RT2R = ZERO
            RT2I = ZERO
         ELSE
            H11 = H11 / S
            H21 = H21 / S
            H12 = H12 / S
            H22 = H22 / S
            TR = ( H11+H22 ) / TWO
            DET = ( H11-TR )*( H22-TR ) - H12*H21
            RTDISC = SQRTABS( DET ) )
            IF( DET.GE.ZERO ) THEN
*
*              ==== complex conjugate shifts ====
*
               RT1R = TR*S
               RT2R = RT1R
               RT1I = RTDISC*S
               RT2I = -RT1I
            ELSE
*
*              ==== real shifts (use only one of them)  ====
*
               RT1R = TR + RTDISC
               RT2R = TR - RTDISC
               IFABS( RT1R-H22 ).LE.ABS( RT2R-H22 ) ) THEN
                  RT1R = RT1R*S
                  RT2R = RT1R
               ELSE
                  RT2R = RT2R*S
                  RT1R = RT2R
               END IF
               RT1I = ZERO
               RT2I = ZERO
            END IF
         END IF
*
*        Look for two consecutive small subdiagonal elements.
*
         DO 50 M = I - 2, L, -1
*           Determine the effect of starting the double-shift QR
*           iteration at row M, and see if this would make H(M,M-1)
*           negligible.  (The following uses scaling to avoid
*           overflows and most underflows.)
*
            H21S = H( M+1, M )
            S = ABS( H( M, M )-RT2R ) + ABS( RT2I ) + ABS( H21S )
            H21S = H( M+1, M ) / S
            V( 1 ) = H21S*H( M, M+1 ) + ( H( M, M )-RT1R )*
     $               ( ( H( M, M )-RT2R ) / S ) - RT1I*( RT2I / S )
            V( 2 ) = H21S*( H( M, M )+H( M+1, M+1 )-RT1R-RT2R )
            V( 3 ) = H21S*H( M+2, M+1 )
            S = ABS( V( 1 ) ) + ABS( V( 2 ) ) + ABS( V( 3 ) )
            V( 1 ) = V( 1 ) / S
            V( 2 ) = V( 2 ) / S
            V( 3 ) = V( 3 ) / S
            IF( M.EQ.L )
     $         GO TO 60
            IFABS( H( M, M-1 ) )*ABS( V( 2 ) )+ABS( V( 3 ) ) ).LE.
     $          ULP*ABS( V( 1 ) )*ABS( H( M-1, M-1 ) )+ABS( H( M,
     $          M ) )+ABS( H( M+1, M+1 ) ) ) )GO TO 60
   50    CONTINUE
   60    CONTINUE
*
*        Double-shift QR step
*
         DO 130 K = M, I - 1
*
*           The first iteration of this loop determines a reflection G
*           from the vector V and applies it from left and right to H,
*           thus creating a nonzero bulge below the subdiagonal.
*
*           Each subsequent iteration determines a reflection G to
*           restore the Hessenberg form in the (K-1)th column, and thus
*           chases the bulge one step toward the bottom of the active
*           submatrix. NR is the order of G.
*
            NR = MIN3, I-K+1 )
            IF( K.GT.M )
     $         CALL SCOPY( NR, H( K, K-1 ), 1, V, 1 )
            CALL SLARFG( NR, V( 1 ), V( 2 ), 1, T1 )
            IF( K.GT.M ) THEN
               H( K, K-1 ) = V( 1 )
               H( K+1, K-1 ) = ZERO
               IF( K.LT.I-1 )
     $            H( K+2, K-1 ) = ZERO
            ELSE IF( M.GT.L ) THEN
*               ==== Use the following instead of
*               .    H( K, K-1 ) = -H( K, K-1 ) to
*               .    avoid a bug when v(2) and v(3)
*               .    underflow. ====
               H( K, K-1 ) = H( K, K-1 )*( ONE-T1 )
            END IF
            V2 = V( 2 )
            T2 = T1*V2
            IF( NR.EQ.3 ) THEN
               V3 = V( 3 )
               T3 = T1*V3
*
*              Apply G from the left to transform the rows of the matrix
*              in columns K to I2.
*
               DO 70 J = K, I2
                  SUM = H( K, J ) + V2*H( K+1, J ) + V3*H( K+2, J )
                  H( K, J ) = H( K, J ) - SUM*T1
                  H( K+1, J ) = H( K+1, J ) - SUM*T2
                  H( K+2, J ) = H( K+2, J ) - SUM*T3
   70          CONTINUE
*
*              Apply G from the right to transform the columns of the
*              matrix in rows I1 to min(K+3,I).
*
               DO 80 J = I1, MIN( K+3, I )
                  SUM = H( J, K ) + V2*H( J, K+1 ) + V3*H( J, K+2 )
                  H( J, K ) = H( J, K ) - SUM*T1
                  H( J, K+1 ) = H( J, K+1 ) - SUM*T2
                  H( J, K+2 ) = H( J, K+2 ) - SUM*T3
   80          CONTINUE
*
               IF( WANTZ ) THEN
*
*                 Accumulate transformations in the matrix Z
*
                  DO 90 J = ILOZ, IHIZ
                     SUM = Z( J, K ) + V2*Z( J, K+1 ) + V3*Z( J, K+2 )
                     Z( J, K ) = Z( J, K ) - SUM*T1
                     Z( J, K+1 ) = Z( J, K+1 ) - SUM*T2
                     Z( J, K+2 ) = Z( J, K+2 ) - SUM*T3
   90             CONTINUE
               END IF
            ELSE IF( NR.EQ.2 ) THEN
*
*              Apply G from the left to transform the rows of the matrix
*              in columns K to I2.
*
               DO 100 J = K, I2
                  SUM = H( K, J ) + V2*H( K+1, J )
                  H( K, J ) = H( K, J ) - SUM*T1
                  H( K+1, J ) = H( K+1, J ) - SUM*T2
  100          CONTINUE
*
*              Apply G from the right to transform the columns of the
*              matrix in rows I1 to min(K+3,I).
*
               DO 110 J = I1, I
                  SUM = H( J, K ) + V2*H( J, K+1 )
                  H( J, K ) = H( J, K ) - SUM*T1
                  H( J, K+1 ) = H( J, K+1 ) - SUM*T2
  110          CONTINUE
*
               IF( WANTZ ) THEN
*
*                 Accumulate transformations in the matrix Z
*
                  DO 120 J = ILOZ, IHIZ
                     SUM = Z( J, K ) + V2*Z( J, K+1 )
                     Z( J, K ) = Z( J, K ) - SUM*T1
                     Z( J, K+1 ) = Z( J, K+1 ) - SUM*T2
  120             CONTINUE
               END IF
            END IF
  130    CONTINUE
*
  140 CONTINUE
*
*     Failure to converge in remaining number of iterations
*
      INFO = I
      RETURN
*
  150 CONTINUE
*
      IF( L.EQ.I ) THEN
*
*        H(I,I-1) is negligible: one eigenvalue has converged.
*
         WR( I ) = H( I, I )
         WI( I ) = ZERO
      ELSE IF( L.EQ.I-1 ) THEN
*
*        H(I-1,I-2) is negligible: a pair of eigenvalues have converged.
*
*        Transform the 2-by-2 submatrix to standard Schur form,
*        and compute and store the eigenvalues.
*
         CALL SLANV2( H( I-1, I-1 ), H( I-1, I ), H( I, I-1 ),
     $                H( I, I ), WR( I-1 ), WI( I-1 ), WR( I ), WI( I ),
     $                CS, SN )
*
         IF( WANTT ) THEN
*
*           Apply the transformation to the rest of H.
*
            IF( I2.GT.I )
     $         CALL SROT( I2-I, H( I-1, I+1 ), LDH, H( I, I+1 ), LDH,
     $                    CS, SN )
            CALL SROT( I-I1-1, H( I1, I-1 ), 1, H( I1, I ), 1, CS, SN )
         END IF
         IF( WANTZ ) THEN
*
*           Apply the transformation to Z.
*
            CALL SROT( NZ, Z( ILOZ, I-1 ), 1, Z( ILOZ, I ), 1, CS, SN )
         END IF
      END IF
*
*     return to start of the main loop with new value of I.
*
      I = L - 1
      GO TO 20
*
  160 CONTINUE
      RETURN
*
*     End of SLAHQR
*
      END