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
      SUBROUTINE ZLAR1V( N, B1, BN, LAMBDA, D, L, LD, LLD,
     $           PIVMIN, GAPTOL, Z, WANTNC, NEGCNT, ZTZ, MINGMA,
     $           R, ISUPPZ, NRMINV, RESID, RQCORR, WORK )
*
*  -- LAPACK auxiliary 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 ..
      LOGICAL            WANTNC
      INTEGER   B1, BN, N, NEGCNT, R
      DOUBLE PRECISION   GAPTOL, LAMBDA, MINGMA, NRMINV, PIVMIN, RESID,
     $                   RQCORR, ZTZ
*     ..
*     .. Array Arguments ..
      INTEGER            ISUPPZ( * )
      DOUBLE PRECISION   D( * ), L( * ), LD( * ), LLD( * ),
     $                  WORK( * )
      COMPLEX*16       Z( * )
*     ..
*
*  Purpose
*  =======
*
*  ZLAR1V computes the (scaled) r-th column of the inverse of
*  the sumbmatrix in rows B1 through BN of the tridiagonal matrix
*  L D L**T - sigma I. When sigma is close to an eigenvalue, the
*  computed vector is an accurate eigenvector. Usually, r corresponds
*  to the index where the eigenvector is largest in magnitude.
*  The following steps accomplish this computation :
*  (a) Stationary qd transform,  L D L**T - sigma I = L(+) D(+) L(+)**T,
*  (b) Progressive qd transform, L D L**T - sigma I = U(-) D(-) U(-)**T,
*  (c) Computation of the diagonal elements of the inverse of
*      L D L**T - sigma I by combining the above transforms, and choosing
*      r as the index where the diagonal of the inverse is (one of the)
*      largest in magnitude.
*  (d) Computation of the (scaled) r-th column of the inverse using the
*      twisted factorization obtained by combining the top part of the
*      the stationary and the bottom part of the progressive transform.
*
*  Arguments
*  =========
*
*  N        (input) INTEGER
*           The order of the matrix L D L**T.
*
*  B1       (input) INTEGER
*           First index of the submatrix of L D L**T.
*
*  BN       (input) INTEGER
*           Last index of the submatrix of L D L**T.
*
*  LAMBDA    (input) DOUBLE PRECISION
*           The shift. In order to compute an accurate eigenvector,
*           LAMBDA should be a good approximation to an eigenvalue
*           of L D L**T.
*
*  L        (input) DOUBLE PRECISION array, dimension (N-1)
*           The (n-1) subdiagonal elements of the unit bidiagonal matrix
*           L, in elements 1 to N-1.
*
*  D        (input) DOUBLE PRECISION array, dimension (N)
*           The n diagonal elements of the diagonal matrix D.
*
*  LD       (input) DOUBLE PRECISION array, dimension (N-1)
*           The n-1 elements L(i)*D(i).
*
*  LLD      (input) DOUBLE PRECISION array, dimension (N-1)
*           The n-1 elements L(i)*L(i)*D(i).
*
*  PIVMIN   (input) DOUBLE PRECISION
*           The minimum pivot in the Sturm sequence.
*
*  GAPTOL   (input) DOUBLE PRECISION
*           Tolerance that indicates when eigenvector entries are negligible
*           w.r.t. their contribution to the residual.
*
*  Z        (input/output) COMPLEX*16       array, dimension (N)
*           On input, all entries of Z must be set to 0.
*           On output, Z contains the (scaled) r-th column of the
*           inverse. The scaling is such that Z(R) equals 1.
*
*  WANTNC   (input) LOGICAL
*           Specifies whether NEGCNT has to be computed.
*
*  NEGCNT   (output) INTEGER
*           If WANTNC is .TRUE. then NEGCNT = the number of pivots < pivmin
*           in the  matrix factorization L D L**T, and NEGCNT = -1 otherwise.
*
*  ZTZ      (output) DOUBLE PRECISION
*           The square of the 2-norm of Z.
*
*  MINGMA   (output) DOUBLE PRECISION
*           The reciprocal of the largest (in magnitude) diagonal
*           element of the inverse of L D L**T - sigma I.
*
*  R        (input/output) INTEGER
*           The twist index for the twisted factorization used to
*           compute Z.
*           On input, 0 <= R <= N. If R is input as 0, R is set to
*           the index where (L D L**T - sigma I)^{-1} is largest
*           in magnitude. If 1 <= R <= N, R is unchanged.
*           On output, R contains the twist index used to compute Z.
*           Ideally, R designates the position of the maximum entry in the
*           eigenvector.
*
*  ISUPPZ   (output) INTEGER array, dimension (2)
*           The support of the vector in Z, i.e., the vector Z is
*           nonzero only in elements ISUPPZ(1) through ISUPPZ( 2 ).
*
*  NRMINV   (output) DOUBLE PRECISION
*           NRMINV = 1/SQRT( ZTZ )
*
*  RESID    (output) DOUBLE PRECISION
*           The residual of the FP vector.
*           RESID = ABS( MINGMA )/SQRT( ZTZ )
*
*  RQCORR   (output) DOUBLE PRECISION
*           The Rayleigh Quotient correction to LAMBDA.
*           RQCORR = MINGMA*TMP
*
*  WORK     (workspace) DOUBLE PRECISION array, dimension (4*N)
*
*  Further Details
*  ===============
*
*  Based on contributions by
*     Beresford Parlett, University of California, Berkeley, USA
*     Jim Demmel, University of California, Berkeley, USA
*     Inderjit Dhillon, University of Texas, Austin, USA
*     Osni Marques, LBNL/NERSC, USA
*     Christof Voemel, University of California, Berkeley, USA
*
*  =====================================================================
*
*     .. Parameters ..
      DOUBLE PRECISION   ZERO, ONE
      PARAMETER          ( ZERO = 0.0D0, ONE = 1.0D0 )
      COMPLEX*16         CONE
      PARAMETER          ( CONE = ( 1.0D00.0D0 ) )

*     ..
*     .. Local Scalars ..
      LOGICAL            SAWNAN1, SAWNAN2
      INTEGER            I, INDLPL, INDP, INDS, INDUMN, NEG1, NEG2, R1,
     $                   R2
      DOUBLE PRECISION   DMINUS, DPLUS, EPS, S, TMP
*     ..
*     .. External Functions ..
      LOGICAL DISNAN
      DOUBLE PRECISION   DLAMCH
      EXTERNAL           DISNAN, DLAMCH
*     ..
*     .. Intrinsic Functions ..
      INTRINSIC          ABSDBLE
*     ..
*     .. Executable Statements ..
*
      EPS = DLAMCH( 'Precision' )


      IF( R.EQ.0 ) THEN
         R1 = B1
         R2 = BN
      ELSE
         R1 = R
         R2 = R
      END IF

*     Storage for LPLUS
      INDLPL = 0
*     Storage for UMINUS
      INDUMN = N
      INDS = 2*+ 1
      INDP = 3*+ 1

      IF( B1.EQ.1 ) THEN
         WORK( INDS ) = ZERO
      ELSE
         WORK( INDS+B1-1 ) = LLD( B1-1 )
      END IF

*
*     Compute the stationary transform (using the differential form)
*     until the index R2.
*
      SAWNAN1 = .FALSE.
      NEG1 = 0
      S = WORK( INDS+B1-1 ) - LAMBDA
      DO 50 I = B1, R1 - 1
         DPLUS = D( I ) + S
         WORK( INDLPL+I ) = LD( I ) / DPLUS
         IF(DPLUS.LT.ZERO) NEG1 = NEG1 + 1
         WORK( INDS+I ) = S*WORK( INDLPL+I )*L( I )
         S = WORK( INDS+I ) - LAMBDA
 50   CONTINUE
      SAWNAN1 = DISNAN( S )
      IF( SAWNAN1 ) GOTO 60
      DO 51 I = R1, R2 - 1
         DPLUS = D( I ) + S
         WORK( INDLPL+I ) = LD( I ) / DPLUS
         WORK( INDS+I ) = S*WORK( INDLPL+I )*L( I )
         S = WORK( INDS+I ) - LAMBDA
 51   CONTINUE
      SAWNAN1 = DISNAN( S )
*
 60   CONTINUE
      IF( SAWNAN1 ) THEN
*        Runs a slower version of the above loop if a NaN is detected
         NEG1 = 0
         S = WORK( INDS+B1-1 ) - LAMBDA
         DO 70 I = B1, R1 - 1
            DPLUS = D( I ) + S
            IF(ABS(DPLUS).LT.PIVMIN) DPLUS = -PIVMIN
            WORK( INDLPL+I ) = LD( I ) / DPLUS
            IF(DPLUS.LT.ZERO) NEG1 = NEG1 + 1
            WORK( INDS+I ) = S*WORK( INDLPL+I )*L( I )
            IF( WORK( INDLPL+I ).EQ.ZERO )
     $                      WORK( INDS+I ) = LLD( I )
            S = WORK( INDS+I ) - LAMBDA
 70      CONTINUE
         DO 71 I = R1, R2 - 1
            DPLUS = D( I ) + S
            IF(ABS(DPLUS).LT.PIVMIN) DPLUS = -PIVMIN
            WORK( INDLPL+I ) = LD( I ) / DPLUS
            WORK( INDS+I ) = S*WORK( INDLPL+I )*L( I )
            IF( WORK( INDLPL+I ).EQ.ZERO )
     $                      WORK( INDS+I ) = LLD( I )
            S = WORK( INDS+I ) - LAMBDA
 71      CONTINUE
      END IF
*
*     Compute the progressive transform (using the differential form)
*     until the index R1
*
      SAWNAN2 = .FALSE.
      NEG2 = 0
      WORK( INDP+BN-1 ) = D( BN ) - LAMBDA
      DO 80 I = BN - 1, R1, -1
         DMINUS = LLD( I ) + WORK( INDP+I )
         TMP = D( I ) / DMINUS
         IF(DMINUS.LT.ZERO) NEG2 = NEG2 + 1
         WORK( INDUMN+I ) = L( I )*TMP
         WORK( INDP+I-1 ) = WORK( INDP+I )*TMP - LAMBDA
 80   CONTINUE
      TMP = WORK( INDP+R1-1 )
      SAWNAN2 = DISNAN( TMP )

      IF( SAWNAN2 ) THEN
*        Runs a slower version of the above loop if a NaN is detected
         NEG2 = 0
         DO 100 I = BN-1, R1, -1
            DMINUS = LLD( I ) + WORK( INDP+I )
            IF(ABS(DMINUS).LT.PIVMIN) DMINUS = -PIVMIN
            TMP = D( I ) / DMINUS
            IF(DMINUS.LT.ZERO) NEG2 = NEG2 + 1
            WORK( INDUMN+I ) = L( I )*TMP
            WORK( INDP+I-1 ) = WORK( INDP+I )*TMP - LAMBDA
            IF( TMP.EQ.ZERO )
     $          WORK( INDP+I-1 ) = D( I ) - LAMBDA
 100     CONTINUE
      END IF
*
*     Find the index (from R1 to R2) of the largest (in magnitude)
*     diagonal element of the inverse
*
      MINGMA = WORK( INDS+R1-1 ) + WORK( INDP+R1-1 )
      IF( MINGMA.LT.ZERO ) NEG1 = NEG1 + 1
      IF( WANTNC ) THEN
         NEGCNT = NEG1 + NEG2
      ELSE
         NEGCNT = -1
      ENDIF
      IFABS(MINGMA).EQ.ZERO )
     $   MINGMA = EPS*WORK( INDS+R1-1 )
      R = R1
      DO 110 I = R1, R2 - 1
         TMP = WORK( INDS+I ) + WORK( INDP+I )
         IF( TMP.EQ.ZERO )
     $      TMP = EPS*WORK( INDS+I )
         IFABS( TMP ).LE.ABS( MINGMA ) ) THEN
            MINGMA = TMP
            R = I + 1
         END IF
 110  CONTINUE
*
*     Compute the FP vector: solve N^T v = e_r
*
      ISUPPZ( 1 ) = B1
      ISUPPZ( 2 ) = BN
      Z( R ) = CONE
      ZTZ = ONE
*
*     Compute the FP vector upwards from R
*
      IF.NOT.SAWNAN1 .AND. .NOT.SAWNAN2 ) THEN
         DO 210 I = R-1, B1, -1
            Z( I ) = -( WORK( INDLPL+I )*Z( I+1 ) )
            IF( (ABS(Z(I))+ABS(Z(I+1)))* ABS(LD(I)).LT.GAPTOL )
     $           THEN
               Z( I ) = ZERO
               ISUPPZ( 1 ) = I + 1
               GOTO 220
            ENDIF
            ZTZ = ZTZ + DBLE( Z( I )*Z( I ) )
 210     CONTINUE
 220     CONTINUE
      ELSE
*        Run slower loop if NaN occurred.
         DO 230 I = R - 1, B1, -1
            IF( Z( I+1 ).EQ.ZERO ) THEN
               Z( I ) = -( LD( I+1 ) / LD( I ) )*Z( I+2 )
            ELSE
               Z( I ) = -( WORK( INDLPL+I )*Z( I+1 ) )
            END IF
            IF( (ABS(Z(I))+ABS(Z(I+1)))* ABS(LD(I)).LT.GAPTOL )
     $           THEN
               Z( I ) = ZERO
               ISUPPZ( 1 ) = I + 1
               GO TO 240
            END IF
            ZTZ = ZTZ + DBLE( Z( I )*Z( I ) )
 230     CONTINUE
 240     CONTINUE
      ENDIF

*     Compute the FP vector downwards from R in blocks of size BLKSIZ
      IF.NOT.SAWNAN1 .AND. .NOT.SAWNAN2 ) THEN
         DO 250 I = R, BN-1
            Z( I+1 ) = -( WORK( INDUMN+I )*Z( I ) )
            IF( (ABS(Z(I))+ABS(Z(I+1)))* ABS(LD(I)).LT.GAPTOL )
     $         THEN
               Z( I+1 ) = ZERO
               ISUPPZ( 2 ) = I
               GO TO 260
            END IF
            ZTZ = ZTZ + DBLE( Z( I+1 )*Z( I+1 ) )
 250     CONTINUE
 260     CONTINUE
      ELSE
*        Run slower loop if NaN occurred.
         DO 270 I = R, BN - 1
            IF( Z( I ).EQ.ZERO ) THEN
               Z( I+1 ) = -( LD( I-1 ) / LD( I ) )*Z( I-1 )
            ELSE
               Z( I+1 ) = -( WORK( INDUMN+I )*Z( I ) )
            END IF
            IF( (ABS(Z(I))+ABS(Z(I+1)))* ABS(LD(I)).LT.GAPTOL )
     $           THEN
               Z( I+1 ) = ZERO
               ISUPPZ( 2 ) = I
               GO TO 280
            END IF
            ZTZ = ZTZ + DBLE( Z( I+1 )*Z( I+1 ) )
 270     CONTINUE
 280     CONTINUE
      END IF
*
*     Compute quantities for convergence test
*
      TMP = ONE / ZTZ
      NRMINV = SQRT( TMP )
      RESID = ABS( MINGMA )*NRMINV
      RQCORR = MINGMA*TMP
*
*
      RETURN
*
*     End of ZLAR1V
*
      END