LAPACK 3.12.1
LAPACK: Linear Algebra PACKage
Loading...
Searching...
No Matches
claed7.f
Go to the documentation of this file.
1*> \brief \b CLAED7 used by CSTEDC. Computes the updated eigensystem of a diagonal matrix after modification by a rank-one symmetric matrix. Used when the original matrix is dense.
2*
3* =========== DOCUMENTATION ===========
4*
5* Online html documentation available at
6* http://www.netlib.org/lapack/explore-html/
7*
8*> Download CLAED7 + dependencies
9*> <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/claed7.f">
10*> [TGZ]</a>
11*> <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/claed7.f">
12*> [ZIP]</a>
13*> <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/claed7.f">
14*> [TXT]</a>
15*
16* Definition:
17* ===========
18*
19* SUBROUTINE CLAED7( N, CUTPNT, QSIZ, TLVLS, CURLVL, CURPBM, D, Q,
20* LDQ, RHO, INDXQ, QSTORE, QPTR, PRMPTR, PERM,
21* GIVPTR, GIVCOL, GIVNUM, WORK, RWORK, IWORK,
22* INFO )
23*
24* .. Scalar Arguments ..
25* INTEGER CURLVL, CURPBM, CUTPNT, INFO, LDQ, N, QSIZ,
26* $ TLVLS
27* REAL RHO
28* ..
29* .. Array Arguments ..
30* INTEGER GIVCOL( 2, * ), GIVPTR( * ), INDXQ( * ),
31* $ IWORK( * ), PERM( * ), PRMPTR( * ), QPTR( * )
32* REAL D( * ), GIVNUM( 2, * ), QSTORE( * ), RWORK( * )
33* COMPLEX Q( LDQ, * ), WORK( * )
34* ..
35*
36*
37*> \par Purpose:
38* =============
39*>
40*> \verbatim
41*>
42*> CLAED7 computes the updated eigensystem of a diagonal
43*> matrix after modification by a rank-one symmetric matrix. This
44*> routine is used only for the eigenproblem which requires all
45*> eigenvalues and optionally eigenvectors of a dense or banded
46*> Hermitian matrix that has been reduced to tridiagonal form.
47*>
48*> T = Q(in) ( D(in) + RHO * Z*Z**H ) Q**H(in) = Q(out) * D(out) * Q**H(out)
49*>
50*> where Z = Q**Hu, u is a vector of length N with ones in the
51*> CUTPNT and CUTPNT + 1 th elements and zeros elsewhere.
52*>
53*> The eigenvectors of the original matrix are stored in Q, and the
54*> eigenvalues are in D. The algorithm consists of three stages:
55*>
56*> The first stage consists of deflating the size of the problem
57*> when there are multiple eigenvalues or if there is a zero in
58*> the Z vector. For each such occurrence the dimension of the
59*> secular equation problem is reduced by one. This stage is
60*> performed by the routine SLAED2.
61*>
62*> The second stage consists of calculating the updated
63*> eigenvalues. This is done by finding the roots of the secular
64*> equation via the routine SLAED4 (as called by SLAED3).
65*> This routine also calculates the eigenvectors of the current
66*> problem.
67*>
68*> The final stage consists of computing the updated eigenvectors
69*> directly using the updated eigenvalues. The eigenvectors for
70*> the current problem are multiplied with the eigenvectors from
71*> the overall problem.
72*> \endverbatim
73*
74* Arguments:
75* ==========
76*
77*> \param[in] N
78*> \verbatim
79*> N is INTEGER
80*> The dimension of the symmetric tridiagonal matrix. N >= 0.
81*> \endverbatim
82*>
83*> \param[in] CUTPNT
84*> \verbatim
85*> CUTPNT is INTEGER
86*> Contains the location of the last eigenvalue in the leading
87*> sub-matrix. min(1,N) <= CUTPNT <= N.
88*> \endverbatim
89*>
90*> \param[in] QSIZ
91*> \verbatim
92*> QSIZ is INTEGER
93*> The dimension of the unitary matrix used to reduce
94*> the full matrix to tridiagonal form. QSIZ >= N.
95*> \endverbatim
96*>
97*> \param[in] TLVLS
98*> \verbatim
99*> TLVLS is INTEGER
100*> The total number of merging levels in the overall divide and
101*> conquer tree.
102*> \endverbatim
103*>
104*> \param[in] CURLVL
105*> \verbatim
106*> CURLVL is INTEGER
107*> The current level in the overall merge routine,
108*> 0 <= curlvl <= tlvls.
109*> \endverbatim
110*>
111*> \param[in] CURPBM
112*> \verbatim
113*> CURPBM is INTEGER
114*> The current problem in the current level in the overall
115*> merge routine (counting from upper left to lower right).
116*> \endverbatim
117*>
118*> \param[in,out] D
119*> \verbatim
120*> D is REAL array, dimension (N)
121*> On entry, the eigenvalues of the rank-1-perturbed matrix.
122*> On exit, the eigenvalues of the repaired matrix.
123*> \endverbatim
124*>
125*> \param[in,out] Q
126*> \verbatim
127*> Q is COMPLEX array, dimension (LDQ,N)
128*> On entry, the eigenvectors of the rank-1-perturbed matrix.
129*> On exit, the eigenvectors of the repaired tridiagonal matrix.
130*> \endverbatim
131*>
132*> \param[in] LDQ
133*> \verbatim
134*> LDQ is INTEGER
135*> The leading dimension of the array Q. LDQ >= max(1,N).
136*> \endverbatim
137*>
138*> \param[in] RHO
139*> \verbatim
140*> RHO is REAL
141*> Contains the subdiagonal element used to create the rank-1
142*> modification.
143*> \endverbatim
144*>
145*> \param[out] INDXQ
146*> \verbatim
147*> INDXQ is INTEGER array, dimension (N)
148*> This contains the permutation which will reintegrate the
149*> subproblem just solved back into sorted order,
150*> ie. D( INDXQ( I = 1, N ) ) will be in ascending order.
151*> \endverbatim
152*>
153*> \param[out] IWORK
154*> \verbatim
155*> IWORK is INTEGER array, dimension (4*N)
156*> \endverbatim
157*>
158*> \param[out] RWORK
159*> \verbatim
160*> RWORK is REAL array,
161*> dimension (3*N+2*QSIZ*N)
162*> \endverbatim
163*>
164*> \param[out] WORK
165*> \verbatim
166*> WORK is COMPLEX array, dimension (QSIZ*N)
167*> \endverbatim
168*>
169*> \param[in,out] QSTORE
170*> \verbatim
171*> QSTORE is REAL array, dimension (N**2+1)
172*> Stores eigenvectors of submatrices encountered during
173*> divide and conquer, packed together. QPTR points to
174*> beginning of the submatrices.
175*> \endverbatim
176*>
177*> \param[in,out] QPTR
178*> \verbatim
179*> QPTR is INTEGER array, dimension (N+2)
180*> List of indices pointing to beginning of submatrices stored
181*> in QSTORE. The submatrices are numbered starting at the
182*> bottom left of the divide and conquer tree, from left to
183*> right and bottom to top.
184*> \endverbatim
185*>
186*> \param[in] PRMPTR
187*> \verbatim
188*> PRMPTR is INTEGER array, dimension (N lg N)
189*> Contains a list of pointers which indicate where in PERM a
190*> level's permutation is stored. PRMPTR(i+1) - PRMPTR(i)
191*> indicates the size of the permutation and also the size of
192*> the full, non-deflated problem.
193*> \endverbatim
194*>
195*> \param[in] PERM
196*> \verbatim
197*> PERM is INTEGER array, dimension (N lg N)
198*> Contains the permutations (from deflation and sorting) to be
199*> applied to each eigenblock.
200*> \endverbatim
201*>
202*> \param[in] GIVPTR
203*> \verbatim
204*> GIVPTR is INTEGER array, dimension (N lg N)
205*> Contains a list of pointers which indicate where in GIVCOL a
206*> level's Givens rotations are stored. GIVPTR(i+1) - GIVPTR(i)
207*> indicates the number of Givens rotations.
208*> \endverbatim
209*>
210*> \param[in] GIVCOL
211*> \verbatim
212*> GIVCOL is INTEGER array, dimension (2, N lg N)
213*> Each pair of numbers indicates a pair of columns to take place
214*> in a Givens rotation.
215*> \endverbatim
216*>
217*> \param[in] GIVNUM
218*> \verbatim
219*> GIVNUM is REAL array, dimension (2, N lg N)
220*> Each number indicates the S value to be used in the
221*> corresponding Givens rotation.
222*> \endverbatim
223*>
224*> \param[out] INFO
225*> \verbatim
226*> INFO is INTEGER
227*> = 0: successful exit.
228*> < 0: if INFO = -i, the i-th argument had an illegal value.
229*> > 0: if INFO = 1, an eigenvalue did not converge
230*> \endverbatim
231*
232* Authors:
233* ========
234*
235*> \author Univ. of Tennessee
236*> \author Univ. of California Berkeley
237*> \author Univ. of Colorado Denver
238*> \author NAG Ltd.
239*
240*> \ingroup laed7
241*
242* =====================================================================
243 SUBROUTINE claed7( N, CUTPNT, QSIZ, TLVLS, CURLVL, CURPBM, D,
244 $ Q,
245 $ LDQ, RHO, INDXQ, QSTORE, QPTR, PRMPTR, PERM,
246 $ GIVPTR, GIVCOL, GIVNUM, WORK, RWORK, IWORK,
247 $ INFO )
248*
249* -- LAPACK computational routine --
250* -- LAPACK is a software package provided by Univ. of Tennessee, --
251* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
252*
253* .. Scalar Arguments ..
254 INTEGER CURLVL, CURPBM, CUTPNT, INFO, LDQ, N, QSIZ,
255 $ TLVLS
256 REAL RHO
257* ..
258* .. Array Arguments ..
259 INTEGER GIVCOL( 2, * ), GIVPTR( * ), INDXQ( * ),
260 $ iwork( * ), perm( * ), prmptr( * ), qptr( * )
261 REAL D( * ), GIVNUM( 2, * ), QSTORE( * ), RWORK( * )
262 COMPLEX Q( LDQ, * ), WORK( * )
263* ..
264*
265* =====================================================================
266*
267* .. Local Scalars ..
268 INTEGER COLTYP, CURR, I, IDLMDA, INDX,
269 $ INDXC, INDXP, IQ, IW, IZ, K, N1, N2, PTR
270* ..
271* .. External Subroutines ..
272 EXTERNAL clacrm, claed8, slaed9, slaeda, slamrg,
273 $ xerbla
274* ..
275* .. Intrinsic Functions ..
276 INTRINSIC max, min
277* ..
278* .. Executable Statements ..
279*
280* Test the input parameters.
281*
282 info = 0
283*
284* IF( ICOMPQ.LT.0 .OR. ICOMPQ.GT.1 ) THEN
285* INFO = -1
286* ELSE IF( N.LT.0 ) THEN
287 IF( n.LT.0 ) THEN
288 info = -1
289 ELSE IF( min( 1, n ).GT.cutpnt .OR. n.LT.cutpnt ) THEN
290 info = -2
291 ELSE IF( qsiz.LT.n ) THEN
292 info = -3
293 ELSE IF( ldq.LT.max( 1, n ) ) THEN
294 info = -9
295 END IF
296 IF( info.NE.0 ) THEN
297 CALL xerbla( 'CLAED7', -info )
298 RETURN
299 END IF
300*
301* Quick return if possible
302*
303 IF( n.EQ.0 )
304 $ RETURN
305*
306* The following values are for bookkeeping purposes only. They are
307* integer pointers which indicate the portion of the workspace
308* used by a particular array in SLAED2 and SLAED3.
309*
310 iz = 1
311 idlmda = iz + n
312 iw = idlmda + n
313 iq = iw + n
314*
315 indx = 1
316 indxc = indx + n
317 coltyp = indxc + n
318 indxp = coltyp + n
319*
320* Form the z-vector which consists of the last row of Q_1 and the
321* first row of Q_2.
322*
323 ptr = 1 + 2**tlvls
324 DO 10 i = 1, curlvl - 1
325 ptr = ptr + 2**( tlvls-i )
326 10 CONTINUE
327 curr = ptr + curpbm
328 CALL slaeda( n, tlvls, curlvl, curpbm, prmptr, perm, givptr,
329 $ givcol, givnum, qstore, qptr, rwork( iz ),
330 $ rwork( iz+n ), info )
331*
332* When solving the final problem, we no longer need the stored data,
333* so we will overwrite the data from this level onto the previously
334* used storage space.
335*
336 IF( curlvl.EQ.tlvls ) THEN
337 qptr( curr ) = 1
338 prmptr( curr ) = 1
339 givptr( curr ) = 1
340 END IF
341*
342* Sort and Deflate eigenvalues.
343*
344 CALL claed8( k, n, qsiz, q, ldq, d, rho, cutpnt, rwork( iz ),
345 $ rwork( idlmda ), work, qsiz, rwork( iw ),
346 $ iwork( indxp ), iwork( indx ), indxq,
347 $ perm( prmptr( curr ) ), givptr( curr+1 ),
348 $ givcol( 1, givptr( curr ) ),
349 $ givnum( 1, givptr( curr ) ), info )
350 prmptr( curr+1 ) = prmptr( curr ) + n
351 givptr( curr+1 ) = givptr( curr+1 ) + givptr( curr )
352*
353* Solve Secular Equation.
354*
355 IF( k.NE.0 ) THEN
356 CALL slaed9( k, 1, k, n, d, rwork( iq ), k, rho,
357 $ rwork( idlmda ), rwork( iw ),
358 $ qstore( qptr( curr ) ), k, info )
359 CALL clacrm( qsiz, k, work, qsiz, qstore( qptr( curr ) ), k,
360 $ q,
361 $ ldq, rwork( iq ) )
362 qptr( curr+1 ) = qptr( curr ) + k**2
363 IF( info.NE.0 ) THEN
364 RETURN
365 END IF
366*
367* Prepare the INDXQ sorting permutation.
368*
369 n1 = k
370 n2 = n - k
371 CALL slamrg( n1, n2, d, 1, -1, indxq )
372 ELSE
373 qptr( curr+1 ) = qptr( curr )
374 DO 20 i = 1, n
375 indxq( i ) = i
376 20 CONTINUE
377 END IF
378*
379 RETURN
380*
381* End of CLAED7
382*
383 END
subroutine xerbla(srname, info)
Definition cblat2.f:3285
subroutine clacrm(m, n, a, lda, b, ldb, c, ldc, rwork)
CLACRM multiplies a complex matrix by a square real matrix.
Definition clacrm.f:112
subroutine claed7(n, cutpnt, qsiz, tlvls, curlvl, curpbm, d, q, ldq, rho, indxq, qstore, qptr, prmptr, perm, givptr, givcol, givnum, work, rwork, iwork, info)
CLAED7 used by CSTEDC. Computes the updated eigensystem of a diagonal matrix after modification by a ...
Definition claed7.f:248
subroutine claed8(k, n, qsiz, q, ldq, d, rho, cutpnt, z, dlambda, q2, ldq2, w, indxp, indx, indxq, perm, givptr, givcol, givnum, info)
CLAED8 used by CSTEDC. Merges eigenvalues and deflates secular equation. Used when the original matri...
Definition claed8.f:227
subroutine slaed9(k, kstart, kstop, n, d, q, ldq, rho, dlambda, w, s, lds, info)
SLAED9 used by SSTEDC. Finds the roots of the secular equation and updates the eigenvectors....
Definition slaed9.f:155
subroutine slaeda(n, tlvls, curlvl, curpbm, prmptr, perm, givptr, givcol, givnum, q, qptr, z, ztemp, info)
SLAEDA used by SSTEDC. Computes the Z vector determining the rank-one modification of the diagonal ma...
Definition slaeda.f:165
subroutine slamrg(n1, n2, a, strd1, strd2, index)
SLAMRG creates a permutation list to merge the entries of two independently sorted sets into a single...
Definition slamrg.f:97