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
      SUBROUTINE SLAED3( K, N, N1, D, Q, LDQ, RHO, DLAMDA, Q2, INDX,
     $                   CTOT, W, S, INFO )
*
*  -- LAPACK 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 ..
      INTEGER            INFO, K, LDQ, N, N1
      REAL               RHO
*     ..
*     .. Array Arguments ..
      INTEGER            CTOT( * ), INDX( * )
      REAL               D( * ), DLAMDA( * ), Q( LDQ, * ), Q2( * ),
     $                   S( * ), W( * )
*     ..
*
*  Purpose
*  =======
*
*  SLAED3 finds the roots of the secular equation, as defined by the
*  values in D, W, and RHO, between 1 and K.  It makes the
*  appropriate calls to SLAED4 and then updates the eigenvectors by
*  multiplying the matrix of eigenvectors of the pair of eigensystems
*  being combined by the matrix of eigenvectors of the K-by-K system
*  which is solved here.
*
*  This code makes very mild assumptions about floating point
*  arithmetic. It will work on machines with a guard digit in
*  add/subtract, or on those binary machines without guard digits
*  which subtract like the Cray X-MP, Cray Y-MP, Cray C-90, or Cray-2.
*  It could conceivably fail on hexadecimal or decimal machines
*  without guard digits, but we know of none.
*
*  Arguments
*  =========
*
*  K       (input) INTEGER
*          The number of terms in the rational function to be solved by
*          SLAED4.  K >= 0.
*
*  N       (input) INTEGER
*          The number of rows and columns in the Q matrix.
*          N >= K (deflation may result in N>K).
*
*  N1      (input) INTEGER
*          The location of the last eigenvalue in the leading submatrix.
*          min(1,N) <= N1 <= N/2.
*
*  D       (output) REAL array, dimension (N)
*          D(I) contains the updated eigenvalues for
*          1 <= I <= K.
*
*  Q       (output) REAL array, dimension (LDQ,N)
*          Initially the first K columns are used as workspace.
*          On output the columns 1 to K contain
*          the updated eigenvectors.
*
*  LDQ     (input) INTEGER
*          The leading dimension of the array Q.  LDQ >= max(1,N).
*
*  RHO     (input) REAL
*          The value of the parameter in the rank one update equation.
*          RHO >= 0 required.
*
*  DLAMDA  (input/output) REAL array, dimension (K)
*          The first K elements of this array contain the old roots
*          of the deflated updating problem.  These are the poles
*          of the secular equation. May be changed on output by
*          having lowest order bit set to zero on Cray X-MP, Cray Y-MP,
*          Cray-2, or Cray C-90, as described above.
*
*  Q2      (input) REAL array, dimension (LDQ2, N)
*          The first K columns of this matrix contain the non-deflated
*          eigenvectors for the split problem.
*
*  INDX    (input) INTEGER array, dimension (N)
*          The permutation used to arrange the columns of the deflated
*          Q matrix into three groups (see SLAED2).
*          The rows of the eigenvectors found by SLAED4 must be likewise
*          permuted before the matrix multiply can take place.
*
*  CTOT    (input) INTEGER array, dimension (4)
*          A count of the total number of the various types of columns
*          in Q, as described in INDX.  The fourth column type is any
*          column which has been deflated.
*
*  W       (input/output) REAL array, dimension (K)
*          The first K elements of this array contain the components
*          of the deflation-adjusted updating vector. Destroyed on
*          output.
*
*  S       (workspace) REAL array, dimension (N1 + 1)*K
*          Will contain the eigenvectors of the repaired matrix which
*          will be multiplied by the previously accumulated eigenvectors
*          to update the system.
*
*  LDS     (input) INTEGER
*          The leading dimension of S.  LDS >= max(1,K).
*
*  INFO    (output) INTEGER
*          = 0:  successful exit.
*          < 0:  if INFO = -i, the i-th argument had an illegal value.
*          > 0:  if INFO = 1, an eigenvalue did not converge
*
*  Further Details
*  ===============
*
*  Based on contributions by
*     Jeff Rutter, Computer Science Division, University of California
*     at Berkeley, USA
*  Modified by Francoise Tisseur, University of Tennessee.
*
*  =====================================================================
*
*     .. Parameters ..
      REAL               ONE, ZERO
      PARAMETER          ( ONE = 1.0E0, ZERO = 0.0E0 )
*     ..
*     .. Local Scalars ..
      INTEGER            I, II, IQ2, J, N12, N2, N23
      REAL               TEMP
*     ..
*     .. External Functions ..
      REAL               SLAMC3, SNRM2
      EXTERNAL           SLAMC3, SNRM2
*     ..
*     .. External Subroutines ..
      EXTERNAL           SCOPY, SGEMM, SLACPY, SLAED4, SLASET, XERBLA
*     ..
*     .. Intrinsic Functions ..
      INTRINSIC          MAXSIGNSQRT
*     ..
*     .. Executable Statements ..
*
*     Test the input parameters.
*
      INFO = 0
*
      IF( K.LT.0 ) THEN
         INFO = -1
      ELSE IF( N.LT.K ) THEN
         INFO = -2
      ELSE IF( LDQ.LT.MAX1, N ) ) THEN
         INFO = -6
      END IF
      IF( INFO.NE.0 ) THEN
         CALL XERBLA( 'SLAED3'-INFO )
         RETURN
      END IF
*
*     Quick return if possible
*
      IF( K.EQ.0 )
     $   RETURN
*
*     Modify values DLAMDA(i) to make sure all DLAMDA(i)-DLAMDA(j) can
*     be computed with high relative accuracy (barring over/underflow).
*     This is a problem on machines without a guard digit in
*     add/subtract (Cray XMP, Cray YMP, Cray C 90 and Cray 2).
*     The following code replaces DLAMDA(I) by 2*DLAMDA(I)-DLAMDA(I),
*     which on any of these machines zeros out the bottommost
*     bit of DLAMDA(I) if it is 1; this makes the subsequent
*     subtractions DLAMDA(I)-DLAMDA(J) unproblematic when cancellation
*     occurs. On binary machines with a guard digit (almost all
*     machines) it does not change DLAMDA(I) at all. On hexadecimal
*     and decimal machines with a guard digit, it slightly
*     changes the bottommost bits of DLAMDA(I). It does not account
*     for hexadecimal or decimal machines without guard digits
*     (we know of none). We use a subroutine call to compute
*     2*DLAMBDA(I) to prevent optimizing compilers from eliminating
*     this code.
*
      DO 10 I = 1, K
         DLAMDA( I ) = SLAMC3( DLAMDA( I ), DLAMDA( I ) ) - DLAMDA( I )
   10 CONTINUE
*
      DO 20 J = 1, K
         CALL SLAED4( K, J, DLAMDA, W, Q( 1, J ), RHO, D( J ), INFO )
*
*        If the zero finder fails, the computation is terminated.
*
         IF( INFO.NE.0 )
     $      GO TO 120
   20 CONTINUE
*
      IF( K.EQ.1 )
     $   GO TO 110
      IF( K.EQ.2 ) THEN
         DO 30 J = 1, K
            W( 1 ) = Q( 1, J )
            W( 2 ) = Q( 2, J )
            II = INDX( 1 )
            Q( 1, J ) = W( II )
            II = INDX( 2 )
            Q( 2, J ) = W( II )
   30    CONTINUE
         GO TO 110
      END IF
*
*     Compute updated W.
*
      CALL SCOPY( K, W, 1, S, 1 )
*
*     Initialize W(I) = Q(I,I)
*
      CALL SCOPY( K, Q, LDQ+1, W, 1 )
      DO 60 J = 1, K
         DO 40 I = 1, J - 1
            W( I ) = W( I )*( Q( I, J ) / ( DLAMDA( I )-DLAMDA( J ) ) )
   40    CONTINUE
         DO 50 I = J + 1, K
            W( I ) = W( I )*( Q( I, J ) / ( DLAMDA( I )-DLAMDA( J ) ) )
   50    CONTINUE
   60 CONTINUE
      DO 70 I = 1, K
         W( I ) = SIGNSQRT-W( I ) ), S( I ) )
   70 CONTINUE
*
*     Compute eigenvectors of the modified rank-1 modification.
*
      DO 100 J = 1, K
         DO 80 I = 1, K
            S( I ) = W( I ) / Q( I, J )
   80    CONTINUE
         TEMP = SNRM2( K, S, 1 )
         DO 90 I = 1, K
            II = INDX( I )
            Q( I, J ) = S( II ) / TEMP
   90    CONTINUE
  100 CONTINUE
*
*     Compute the updated eigenvectors.
*
  110 CONTINUE
*
      N2 = N - N1
      N12 = CTOT( 1 ) + CTOT( 2 )
      N23 = CTOT( 2 ) + CTOT( 3 )
*
      CALL SLACPY( 'A', N23, K, Q( CTOT( 1 )+11 ), LDQ, S, N23 )
      IQ2 = N1*N12 + 1
      IF( N23.NE.0 ) THEN
         CALL SGEMM( 'N''N', N2, K, N23, ONE, Q2( IQ2 ), N2, S, N23,
     $               ZERO, Q( N1+11 ), LDQ )
      ELSE
         CALL SLASET( 'A', N2, K, ZERO, ZERO, Q( N1+11 ), LDQ )
      END IF
*
      CALL SLACPY( 'A', N12, K, Q, LDQ, S, N12 )
      IF( N12.NE.0 ) THEN
         CALL SGEMM( 'N''N', N1, K, N12, ONE, Q2, N1, S, N12, ZERO, Q,
     $               LDQ )
      ELSE
         CALL SLASET( 'A', N1, K, ZERO, ZERO, Q( 11 ), LDQ )
      END IF
*
*
  120 CONTINUE
      RETURN
*
*     End of SLAED3
*
      END