1: SUBROUTINE DTGSEN( IJOB, WANTQ, WANTZ, SELECT, N, A, LDA, B, LDB,
2: $ ALPHAR, ALPHAI, BETA, Q, LDQ, Z, LDZ, M, PL,
3: $ PR, DIF, WORK, LWORK, IWORK, LIWORK, INFO )
4: *
5: * -- LAPACK routine (version 3.2) --
6: * -- LAPACK is a software package provided by Univ. of Tennessee, --
7: * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
8: * January 2007
9: *
10: * Modified to call DLACN2 in place of DLACON, 5 Feb 03, SJH.
11: *
12: * .. Scalar Arguments ..
13: LOGICAL WANTQ, WANTZ
14: INTEGER IJOB, INFO, LDA, LDB, LDQ, LDZ, LIWORK, LWORK,
15: $ M, N
16: DOUBLE PRECISION PL, PR
17: * ..
18: * .. Array Arguments ..
19: LOGICAL SELECT( * )
20: INTEGER IWORK( * )
21: DOUBLE PRECISION A( LDA, * ), ALPHAI( * ), ALPHAR( * ),
22: $ B( LDB, * ), BETA( * ), DIF( * ), Q( LDQ, * ),
23: $ WORK( * ), Z( LDZ, * )
24: * ..
25: *
26: * Purpose
27: * =======
28: *
29: * DTGSEN reorders the generalized real Schur decomposition of a real
30: * matrix pair (A, B) (in terms of an orthonormal equivalence trans-
31: * formation Q' * (A, B) * Z), so that a selected cluster of eigenvalues
32: * appears in the leading diagonal blocks of the upper quasi-triangular
33: * matrix A and the upper triangular B. The leading columns of Q and
34: * Z form orthonormal bases of the corresponding left and right eigen-
35: * spaces (deflating subspaces). (A, B) must be in generalized real
36: * Schur canonical form (as returned by DGGES), i.e. A is block upper
37: * triangular with 1-by-1 and 2-by-2 diagonal blocks. B is upper
38: * triangular.
39: *
40: * DTGSEN also computes the generalized eigenvalues
41: *
42: * w(j) = (ALPHAR(j) + i*ALPHAI(j))/BETA(j)
43: *
44: * of the reordered matrix pair (A, B).
45: *
46: * Optionally, DTGSEN computes the estimates of reciprocal condition
47: * numbers for eigenvalues and eigenspaces. These are Difu[(A11,B11),
48: * (A22,B22)] and Difl[(A11,B11), (A22,B22)], i.e. the separation(s)
49: * between the matrix pairs (A11, B11) and (A22,B22) that correspond to
50: * the selected cluster and the eigenvalues outside the cluster, resp.,
51: * and norms of "projections" onto left and right eigenspaces w.r.t.
52: * the selected cluster in the (1,1)-block.
53: *
54: * Arguments
55: * =========
56: *
57: * IJOB (input) INTEGER
58: * Specifies whether condition numbers are required for the
59: * cluster of eigenvalues (PL and PR) or the deflating subspaces
60: * (Difu and Difl):
61: * =0: Only reorder w.r.t. SELECT. No extras.
62: * =1: Reciprocal of norms of "projections" onto left and right
63: * eigenspaces w.r.t. the selected cluster (PL and PR).
64: * =2: Upper bounds on Difu and Difl. F-norm-based estimate
65: * (DIF(1:2)).
66: * =3: Estimate of Difu and Difl. 1-norm-based estimate
67: * (DIF(1:2)).
68: * About 5 times as expensive as IJOB = 2.
69: * =4: Compute PL, PR and DIF (i.e. 0, 1 and 2 above): Economic
70: * version to get it all.
71: * =5: Compute PL, PR and DIF (i.e. 0, 1 and 3 above)
72: *
73: * WANTQ (input) LOGICAL
74: * .TRUE. : update the left transformation matrix Q;
75: * .FALSE.: do not update Q.
76: *
77: * WANTZ (input) LOGICAL
78: * .TRUE. : update the right transformation matrix Z;
79: * .FALSE.: do not update Z.
80: *
81: * SELECT (input) LOGICAL array, dimension (N)
82: * SELECT specifies the eigenvalues in the selected cluster.
83: * To select a real eigenvalue w(j), SELECT(j) must be set to
84: * .TRUE.. To select a complex conjugate pair of eigenvalues
85: * w(j) and w(j+1), corresponding to a 2-by-2 diagonal block,
86: * either SELECT(j) or SELECT(j+1) or both must be set to
87: * .TRUE.; a complex conjugate pair of eigenvalues must be
88: * either both included in the cluster or both excluded.
89: *
90: * N (input) INTEGER
91: * The order of the matrices A and B. N >= 0.
92: *
93: * A (input/output) DOUBLE PRECISION array, dimension(LDA,N)
94: * On entry, the upper quasi-triangular matrix A, with (A, B) in
95: * generalized real Schur canonical form.
96: * On exit, A is overwritten by the reordered matrix A.
97: *
98: * LDA (input) INTEGER
99: * The leading dimension of the array A. LDA >= max(1,N).
100: *
101: * B (input/output) DOUBLE PRECISION array, dimension(LDB,N)
102: * On entry, the upper triangular matrix B, with (A, B) in
103: * generalized real Schur canonical form.
104: * On exit, B is overwritten by the reordered matrix B.
105: *
106: * LDB (input) INTEGER
107: * The leading dimension of the array B. LDB >= max(1,N).
108: *
109: * ALPHAR (output) DOUBLE PRECISION array, dimension (N)
110: * ALPHAI (output) DOUBLE PRECISION array, dimension (N)
111: * BETA (output) DOUBLE PRECISION array, dimension (N)
112: * On exit, (ALPHAR(j) + ALPHAI(j)*i)/BETA(j), j=1,...,N, will
113: * be the generalized eigenvalues. ALPHAR(j) + ALPHAI(j)*i
114: * and BETA(j),j=1,...,N are the diagonals of the complex Schur
115: * form (S,T) that would result if the 2-by-2 diagonal blocks of
116: * the real generalized Schur form of (A,B) were further reduced
117: * to triangular form using complex unitary transformations.
118: * If ALPHAI(j) is zero, then the j-th eigenvalue is real; if
119: * positive, then the j-th and (j+1)-st eigenvalues are a
120: * complex conjugate pair, with ALPHAI(j+1) negative.
121: *
122: * Q (input/output) DOUBLE PRECISION array, dimension (LDQ,N)
123: * On entry, if WANTQ = .TRUE., Q is an N-by-N matrix.
124: * On exit, Q has been postmultiplied by the left orthogonal
125: * transformation matrix which reorder (A, B); The leading M
126: * columns of Q form orthonormal bases for the specified pair of
127: * left eigenspaces (deflating subspaces).
128: * If WANTQ = .FALSE., Q is not referenced.
129: *
130: * LDQ (input) INTEGER
131: * The leading dimension of the array Q. LDQ >= 1;
132: * and if WANTQ = .TRUE., LDQ >= N.
133: *
134: * Z (input/output) DOUBLE PRECISION array, dimension (LDZ,N)
135: * On entry, if WANTZ = .TRUE., Z is an N-by-N matrix.
136: * On exit, Z has been postmultiplied by the left orthogonal
137: * transformation matrix which reorder (A, B); The leading M
138: * columns of Z form orthonormal bases for the specified pair of
139: * left eigenspaces (deflating subspaces).
140: * If WANTZ = .FALSE., Z is not referenced.
141: *
142: * LDZ (input) INTEGER
143: * The leading dimension of the array Z. LDZ >= 1;
144: * If WANTZ = .TRUE., LDZ >= N.
145: *
146: * M (output) INTEGER
147: * The dimension of the specified pair of left and right eigen-
148: * spaces (deflating subspaces). 0 <= M <= N.
149: *
150: * PL (output) DOUBLE PRECISION
151: * PR (output) DOUBLE PRECISION
152: * If IJOB = 1, 4 or 5, PL, PR are lower bounds on the
153: * reciprocal of the norm of "projections" onto left and right
154: * eigenspaces with respect to the selected cluster.
155: * 0 < PL, PR <= 1.
156: * If M = 0 or M = N, PL = PR = 1.
157: * If IJOB = 0, 2 or 3, PL and PR are not referenced.
158: *
159: * DIF (output) DOUBLE PRECISION array, dimension (2).
160: * If IJOB >= 2, DIF(1:2) store the estimates of Difu and Difl.
161: * If IJOB = 2 or 4, DIF(1:2) are F-norm-based upper bounds on
162: * Difu and Difl. If IJOB = 3 or 5, DIF(1:2) are 1-norm-based
163: * estimates of Difu and Difl.
164: * If M = 0 or N, DIF(1:2) = F-norm([A, B]).
165: * If IJOB = 0 or 1, DIF is not referenced.
166: *
167: * WORK (workspace/output) DOUBLE PRECISION array,
168: * dimension (MAX(1,LWORK))
169: * On exit, if INFO = 0, WORK(1) returns the optimal LWORK.
170: *
171: * LWORK (input) INTEGER
172: * The dimension of the array WORK. LWORK >= 4*N+16.
173: * If IJOB = 1, 2 or 4, LWORK >= MAX(4*N+16, 2*M*(N-M)).
174: * If IJOB = 3 or 5, LWORK >= MAX(4*N+16, 4*M*(N-M)).
175: *
176: * If LWORK = -1, then a workspace query is assumed; the routine
177: * only calculates the optimal size of the WORK array, returns
178: * this value as the first entry of the WORK array, and no error
179: * message related to LWORK is issued by XERBLA.
180: *
181: * IWORK (workspace/output) INTEGER array, dimension (MAX(1,LIWORK))
182: * IF IJOB = 0, IWORK is not referenced. Otherwise,
183: * on exit, if INFO = 0, IWORK(1) returns the optimal LIWORK.
184: *
185: * LIWORK (input) INTEGER
186: * The dimension of the array IWORK. LIWORK >= 1.
187: * If IJOB = 1, 2 or 4, LIWORK >= N+6.
188: * If IJOB = 3 or 5, LIWORK >= MAX(2*M*(N-M), N+6).
189: *
190: * If LIWORK = -1, then a workspace query is assumed; the
191: * routine only calculates the optimal size of the IWORK array,
192: * returns this value as the first entry of the IWORK array, and
193: * no error message related to LIWORK is issued by XERBLA.
194: *
195: * INFO (output) INTEGER
196: * =0: Successful exit.
197: * <0: If INFO = -i, the i-th argument had an illegal value.
198: * =1: Reordering of (A, B) failed because the transformed
199: * matrix pair (A, B) would be too far from generalized
200: * Schur form; the problem is very ill-conditioned.
201: * (A, B) may have been partially reordered.
202: * If requested, 0 is returned in DIF(*), PL and PR.
203: *
204: * Further Details
205: * ===============
206: *
207: * DTGSEN first collects the selected eigenvalues by computing
208: * orthogonal U and W that move them to the top left corner of (A, B).
209: * In other words, the selected eigenvalues are the eigenvalues of
210: * (A11, B11) in:
211: *
212: * U'*(A, B)*W = (A11 A12) (B11 B12) n1
213: * ( 0 A22),( 0 B22) n2
214: * n1 n2 n1 n2
215: *
216: * where N = n1+n2 and U' means the transpose of U. The first n1 columns
217: * of U and W span the specified pair of left and right eigenspaces
218: * (deflating subspaces) of (A, B).
219: *
220: * If (A, B) has been obtained from the generalized real Schur
221: * decomposition of a matrix pair (C, D) = Q*(A, B)*Z', then the
222: * reordered generalized real Schur form of (C, D) is given by
223: *
224: * (C, D) = (Q*U)*(U'*(A, B)*W)*(Z*W)',
225: *
226: * and the first n1 columns of Q*U and Z*W span the corresponding
227: * deflating subspaces of (C, D) (Q and Z store Q*U and Z*W, resp.).
228: *
229: * Note that if the selected eigenvalue is sufficiently ill-conditioned,
230: * then its value may differ significantly from its value before
231: * reordering.
232: *
233: * The reciprocal condition numbers of the left and right eigenspaces
234: * spanned by the first n1 columns of U and W (or Q*U and Z*W) may
235: * be returned in DIF(1:2), corresponding to Difu and Difl, resp.
236: *
237: * The Difu and Difl are defined as:
238: *
239: * Difu[(A11, B11), (A22, B22)] = sigma-min( Zu )
240: * and
241: * Difl[(A11, B11), (A22, B22)] = Difu[(A22, B22), (A11, B11)],
242: *
243: * where sigma-min(Zu) is the smallest singular value of the
244: * (2*n1*n2)-by-(2*n1*n2) matrix
245: *
246: * Zu = [ kron(In2, A11) -kron(A22', In1) ]
247: * [ kron(In2, B11) -kron(B22', In1) ].
248: *
249: * Here, Inx is the identity matrix of size nx and A22' is the
250: * transpose of A22. kron(X, Y) is the Kronecker product between
251: * the matrices X and Y.
252: *
253: * When DIF(2) is small, small changes in (A, B) can cause large changes
254: * in the deflating subspace. An approximate (asymptotic) bound on the
255: * maximum angular error in the computed deflating subspaces is
256: *
257: * EPS * norm((A, B)) / DIF(2),
258: *
259: * where EPS is the machine precision.
260: *
261: * The reciprocal norm of the projectors on the left and right
262: * eigenspaces associated with (A11, B11) may be returned in PL and PR.
263: * They are computed as follows. First we compute L and R so that
264: * P*(A, B)*Q is block diagonal, where
265: *
266: * P = ( I -L ) n1 Q = ( I R ) n1
267: * ( 0 I ) n2 and ( 0 I ) n2
268: * n1 n2 n1 n2
269: *
270: * and (L, R) is the solution to the generalized Sylvester equation
271: *
272: * A11*R - L*A22 = -A12
273: * B11*R - L*B22 = -B12
274: *
275: * Then PL = (F-norm(L)**2+1)**(-1/2) and PR = (F-norm(R)**2+1)**(-1/2).
276: * An approximate (asymptotic) bound on the average absolute error of
277: * the selected eigenvalues is
278: *
279: * EPS * norm((A, B)) / PL.
280: *
281: * There are also global error bounds which valid for perturbations up
282: * to a certain restriction: A lower bound (x) on the smallest
283: * F-norm(E,F) for which an eigenvalue of (A11, B11) may move and
284: * coalesce with an eigenvalue of (A22, B22) under perturbation (E,F),
285: * (i.e. (A + E, B + F), is
286: *
287: * x = min(Difu,Difl)/((1/(PL*PL)+1/(PR*PR))**(1/2)+2*max(1/PL,1/PR)).
288: *
289: * An approximate bound on x can be computed from DIF(1:2), PL and PR.
290: *
291: * If y = ( F-norm(E,F) / x) <= 1, the angles between the perturbed
292: * (L', R') and unperturbed (L, R) left and right deflating subspaces
293: * associated with the selected cluster in the (1,1)-blocks can be
294: * bounded as
295: *
296: * max-angle(L, L') <= arctan( y * PL / (1 - y * (1 - PL * PL)**(1/2))
297: * max-angle(R, R') <= arctan( y * PR / (1 - y * (1 - PR * PR)**(1/2))
298: *
299: * See LAPACK User's Guide section 4.11 or the following references
300: * for more information.
301: *
302: * Note that if the default method for computing the Frobenius-norm-
303: * based estimate DIF is not wanted (see DLATDF), then the parameter
304: * IDIFJB (see below) should be changed from 3 to 4 (routine DLATDF
305: * (IJOB = 2 will be used)). See DTGSYL for more details.
306: *
307: * Based on contributions by
308: * Bo Kagstrom and Peter Poromaa, Department of Computing Science,
309: * Umea University, S-901 87 Umea, Sweden.
310: *
311: * References
312: * ==========
313: *
314: * [1] B. Kagstrom; A Direct Method for Reordering Eigenvalues in the
315: * Generalized Real Schur Form of a Regular Matrix Pair (A, B), in
316: * M.S. Moonen et al (eds), Linear Algebra for Large Scale and
317: * Real-Time Applications, Kluwer Academic Publ. 1993, pp 195-218.
318: *
319: * [2] B. Kagstrom and P. Poromaa; Computing Eigenspaces with Specified
320: * Eigenvalues of a Regular Matrix Pair (A, B) and Condition
321: * Estimation: Theory, Algorithms and Software,
322: * Report UMINF - 94.04, Department of Computing Science, Umea
323: * University, S-901 87 Umea, Sweden, 1994. Also as LAPACK Working
324: * Note 87. To appear in Numerical Algorithms, 1996.
325: *
326: * [3] B. Kagstrom and P. Poromaa, LAPACK-Style Algorithms and Software
327: * for Solving the Generalized Sylvester Equation and Estimating the
328: * Separation between Regular Matrix Pairs, Report UMINF - 93.23,
329: * Department of Computing Science, Umea University, S-901 87 Umea,
330: * Sweden, December 1993, Revised April 1994, Also as LAPACK Working
331: * Note 75. To appear in ACM Trans. on Math. Software, Vol 22, No 1,
332: * 1996.
333: *
334: * =====================================================================
335: *
336: * .. Parameters ..
337: INTEGER IDIFJB
338: PARAMETER ( IDIFJB = 3 )
339: DOUBLE PRECISION ZERO, ONE
340: PARAMETER ( ZERO = 0.0D+0, ONE = 1.0D+0 )
341: * ..
342: * .. Local Scalars ..
343: LOGICAL LQUERY, PAIR, SWAP, WANTD, WANTD1, WANTD2,
344: $ WANTP
345: INTEGER I, IERR, IJB, K, KASE, KK, KS, LIWMIN, LWMIN,
346: $ MN2, N1, N2
347: DOUBLE PRECISION DSCALE, DSUM, EPS, RDSCAL, SMLNUM
348: * ..
349: * .. Local Arrays ..
350: INTEGER ISAVE( 3 )
351: * ..
352: * .. External Subroutines ..
353: EXTERNAL DLACN2, DLACPY, DLAG2, DLASSQ, DTGEXC, DTGSYL,
354: $ XERBLA
355: * ..
356: * .. External Functions ..
357: DOUBLE PRECISION DLAMCH
358: EXTERNAL DLAMCH
359: * ..
360: * .. Intrinsic Functions ..
361: INTRINSIC MAX, SIGN, SQRT
362: * ..
363: * .. Executable Statements ..
364: *
365: * Decode and test the input parameters
366: *
367: INFO = 0
368: LQUERY = ( LWORK.EQ.-1 .OR. LIWORK.EQ.-1 )
369: *
370: IF( IJOB.LT.0 .OR. IJOB.GT.5 ) THEN
371: INFO = -1
372: ELSE IF( N.LT.0 ) THEN
373: INFO = -5
374: ELSE IF( LDA.LT.MAX( 1, N ) ) THEN
375: INFO = -7
376: ELSE IF( LDB.LT.MAX( 1, N ) ) THEN
377: INFO = -9
378: ELSE IF( LDQ.LT.1 .OR. ( WANTQ .AND. LDQ.LT.N ) ) THEN
379: INFO = -14
380: ELSE IF( LDZ.LT.1 .OR. ( WANTZ .AND. LDZ.LT.N ) ) THEN
381: INFO = -16
382: END IF
383: *
384: IF( INFO.NE.0 ) THEN
385: CALL XERBLA( 'DTGSEN', -INFO )
386: RETURN
387: END IF
388: *
389: * Get machine constants
390: *
391: EPS = DLAMCH( 'P' )
392: SMLNUM = DLAMCH( 'S' ) / EPS
393: IERR = 0
394: *
395: WANTP = IJOB.EQ.1 .OR. IJOB.GE.4
396: WANTD1 = IJOB.EQ.2 .OR. IJOB.EQ.4
397: WANTD2 = IJOB.EQ.3 .OR. IJOB.EQ.5
398: WANTD = WANTD1 .OR. WANTD2
399: *
400: * Set M to the dimension of the specified pair of deflating
401: * subspaces.
402: *
403: M = 0
404: PAIR = .FALSE.
405: DO 10 K = 1, N
406: IF( PAIR ) THEN
407: PAIR = .FALSE.
408: ELSE
409: IF( K.LT.N ) THEN
410: IF( A( K+1, K ).EQ.ZERO ) THEN
411: IF( SELECT( K ) )
412: $ M = M + 1
413: ELSE
414: PAIR = .TRUE.
415: IF( SELECT( K ) .OR. SELECT( K+1 ) )
416: $ M = M + 2
417: END IF
418: ELSE
419: IF( SELECT( N ) )
420: $ M = M + 1
421: END IF
422: END IF
423: 10 CONTINUE
424: *
425: IF( IJOB.EQ.1 .OR. IJOB.EQ.2 .OR. IJOB.EQ.4 ) THEN
426: LWMIN = MAX( 1, 4*N+16, 2*M*( N-M ) )
427: LIWMIN = MAX( 1, N+6 )
428: ELSE IF( IJOB.EQ.3 .OR. IJOB.EQ.5 ) THEN
429: LWMIN = MAX( 1, 4*N+16, 4*M*( N-M ) )
430: LIWMIN = MAX( 1, 2*M*( N-M ), N+6 )
431: ELSE
432: LWMIN = MAX( 1, 4*N+16 )
433: LIWMIN = 1
434: END IF
435: *
436: WORK( 1 ) = LWMIN
437: IWORK( 1 ) = LIWMIN
438: *
439: IF( LWORK.LT.LWMIN .AND. .NOT.LQUERY ) THEN
440: INFO = -22
441: ELSE IF( LIWORK.LT.LIWMIN .AND. .NOT.LQUERY ) THEN
442: INFO = -24
443: END IF
444: *
445: IF( INFO.NE.0 ) THEN
446: CALL XERBLA( 'DTGSEN', -INFO )
447: RETURN
448: ELSE IF( LQUERY ) THEN
449: RETURN
450: END IF
451: *
452: * Quick return if possible.
453: *
454: IF( M.EQ.N .OR. M.EQ.0 ) THEN
455: IF( WANTP ) THEN
456: PL = ONE
457: PR = ONE
458: END IF
459: IF( WANTD ) THEN
460: DSCALE = ZERO
461: DSUM = ONE
462: DO 20 I = 1, N
463: CALL DLASSQ( N, A( 1, I ), 1, DSCALE, DSUM )
464: CALL DLASSQ( N, B( 1, I ), 1, DSCALE, DSUM )
465: 20 CONTINUE
466: DIF( 1 ) = DSCALE*SQRT( DSUM )
467: DIF( 2 ) = DIF( 1 )
468: END IF
469: GO TO 60
470: END IF
471: *
472: * Collect the selected blocks at the top-left corner of (A, B).
473: *
474: KS = 0
475: PAIR = .FALSE.
476: DO 30 K = 1, N
477: IF( PAIR ) THEN
478: PAIR = .FALSE.
479: ELSE
480: *
481: SWAP = SELECT( K )
482: IF( K.LT.N ) THEN
483: IF( A( K+1, K ).NE.ZERO ) THEN
484: PAIR = .TRUE.
485: SWAP = SWAP .OR. SELECT( K+1 )
486: END IF
487: END IF
488: *
489: IF( SWAP ) THEN
490: KS = KS + 1
491: *
492: * Swap the K-th block to position KS.
493: * Perform the reordering of diagonal blocks in (A, B)
494: * by orthogonal transformation matrices and update
495: * Q and Z accordingly (if requested):
496: *
497: KK = K
498: IF( K.NE.KS )
499: $ CALL DTGEXC( WANTQ, WANTZ, N, A, LDA, B, LDB, Q, LDQ,
500: $ Z, LDZ, KK, KS, WORK, LWORK, IERR )
501: *
502: IF( IERR.GT.0 ) THEN
503: *
504: * Swap is rejected: exit.
505: *
506: INFO = 1
507: IF( WANTP ) THEN
508: PL = ZERO
509: PR = ZERO
510: END IF
511: IF( WANTD ) THEN
512: DIF( 1 ) = ZERO
513: DIF( 2 ) = ZERO
514: END IF
515: GO TO 60
516: END IF
517: *
518: IF( PAIR )
519: $ KS = KS + 1
520: END IF
521: END IF
522: 30 CONTINUE
523: IF( WANTP ) THEN
524: *
525: * Solve generalized Sylvester equation for R and L
526: * and compute PL and PR.
527: *
528: N1 = M
529: N2 = N - M
530: I = N1 + 1
531: IJB = 0
532: CALL DLACPY( 'Full', N1, N2, A( 1, I ), LDA, WORK, N1 )
533: CALL DLACPY( 'Full', N1, N2, B( 1, I ), LDB, WORK( N1*N2+1 ),
534: $ N1 )
535: CALL DTGSYL( 'N', IJB, N1, N2, A, LDA, A( I, I ), LDA, WORK,
536: $ N1, B, LDB, B( I, I ), LDB, WORK( N1*N2+1 ), N1,
537: $ DSCALE, DIF( 1 ), WORK( N1*N2*2+1 ),
538: $ LWORK-2*N1*N2, IWORK, IERR )
539: *
540: * Estimate the reciprocal of norms of "projections" onto left
541: * and right eigenspaces.
542: *
543: RDSCAL = ZERO
544: DSUM = ONE
545: CALL DLASSQ( N1*N2, WORK, 1, RDSCAL, DSUM )
546: PL = RDSCAL*SQRT( DSUM )
547: IF( PL.EQ.ZERO ) THEN
548: PL = ONE
549: ELSE
550: PL = DSCALE / ( SQRT( DSCALE*DSCALE / PL+PL )*SQRT( PL ) )
551: END IF
552: RDSCAL = ZERO
553: DSUM = ONE
554: CALL DLASSQ( N1*N2, WORK( N1*N2+1 ), 1, RDSCAL, DSUM )
555: PR = RDSCAL*SQRT( DSUM )
556: IF( PR.EQ.ZERO ) THEN
557: PR = ONE
558: ELSE
559: PR = DSCALE / ( SQRT( DSCALE*DSCALE / PR+PR )*SQRT( PR ) )
560: END IF
561: END IF
562: *
563: IF( WANTD ) THEN
564: *
565: * Compute estimates of Difu and Difl.
566: *
567: IF( WANTD1 ) THEN
568: N1 = M
569: N2 = N - M
570: I = N1 + 1
571: IJB = IDIFJB
572: *
573: * Frobenius norm-based Difu-estimate.
574: *
575: CALL DTGSYL( 'N', IJB, N1, N2, A, LDA, A( I, I ), LDA, WORK,
576: $ N1, B, LDB, B( I, I ), LDB, WORK( N1*N2+1 ),
577: $ N1, DSCALE, DIF( 1 ), WORK( 2*N1*N2+1 ),
578: $ LWORK-2*N1*N2, IWORK, IERR )
579: *
580: * Frobenius norm-based Difl-estimate.
581: *
582: CALL DTGSYL( 'N', IJB, N2, N1, A( I, I ), LDA, A, LDA, WORK,
583: $ N2, B( I, I ), LDB, B, LDB, WORK( N1*N2+1 ),
584: $ N2, DSCALE, DIF( 2 ), WORK( 2*N1*N2+1 ),
585: $ LWORK-2*N1*N2, IWORK, IERR )
586: ELSE
587: *
588: *
589: * Compute 1-norm-based estimates of Difu and Difl using
590: * reversed communication with DLACN2. In each step a
591: * generalized Sylvester equation or a transposed variant
592: * is solved.
593: *
594: KASE = 0
595: N1 = M
596: N2 = N - M
597: I = N1 + 1
598: IJB = 0
599: MN2 = 2*N1*N2
600: *
601: * 1-norm-based estimate of Difu.
602: *
603: 40 CONTINUE
604: CALL DLACN2( MN2, WORK( MN2+1 ), WORK, IWORK, DIF( 1 ),
605: $ KASE, ISAVE )
606: IF( KASE.NE.0 ) THEN
607: IF( KASE.EQ.1 ) THEN
608: *
609: * Solve generalized Sylvester equation.
610: *
611: CALL DTGSYL( 'N', IJB, N1, N2, A, LDA, A( I, I ), LDA,
612: $ WORK, N1, B, LDB, B( I, I ), LDB,
613: $ WORK( N1*N2+1 ), N1, DSCALE, DIF( 1 ),
614: $ WORK( 2*N1*N2+1 ), LWORK-2*N1*N2, IWORK,
615: $ IERR )
616: ELSE
617: *
618: * Solve the transposed variant.
619: *
620: CALL DTGSYL( 'T', IJB, N1, N2, A, LDA, A( I, I ), LDA,
621: $ WORK, N1, B, LDB, B( I, I ), LDB,
622: $ WORK( N1*N2+1 ), N1, DSCALE, DIF( 1 ),
623: $ WORK( 2*N1*N2+1 ), LWORK-2*N1*N2, IWORK,
624: $ IERR )
625: END IF
626: GO TO 40
627: END IF
628: DIF( 1 ) = DSCALE / DIF( 1 )
629: *
630: * 1-norm-based estimate of Difl.
631: *
632: 50 CONTINUE
633: CALL DLACN2( MN2, WORK( MN2+1 ), WORK, IWORK, DIF( 2 ),
634: $ KASE, ISAVE )
635: IF( KASE.NE.0 ) THEN
636: IF( KASE.EQ.1 ) THEN
637: *
638: * Solve generalized Sylvester equation.
639: *
640: CALL DTGSYL( 'N', IJB, N2, N1, A( I, I ), LDA, A, LDA,
641: $ WORK, N2, B( I, I ), LDB, B, LDB,
642: $ WORK( N1*N2+1 ), N2, DSCALE, DIF( 2 ),
643: $ WORK( 2*N1*N2+1 ), LWORK-2*N1*N2, IWORK,
644: $ IERR )
645: ELSE
646: *
647: * Solve the transposed variant.
648: *
649: CALL DTGSYL( 'T', IJB, N2, N1, A( I, I ), LDA, A, LDA,
650: $ WORK, N2, B( I, I ), LDB, B, LDB,
651: $ WORK( N1*N2+1 ), N2, DSCALE, DIF( 2 ),
652: $ WORK( 2*N1*N2+1 ), LWORK-2*N1*N2, IWORK,
653: $ IERR )
654: END IF
655: GO TO 50
656: END IF
657: DIF( 2 ) = DSCALE / DIF( 2 )
658: *
659: END IF
660: END IF
661: *
662: 60 CONTINUE
663: *
664: * Compute generalized eigenvalues of reordered pair (A, B) and
665: * normalize the generalized Schur form.
666: *
667: PAIR = .FALSE.
668: DO 80 K = 1, N
669: IF( PAIR ) THEN
670: PAIR = .FALSE.
671: ELSE
672: *
673: IF( K.LT.N ) THEN
674: IF( A( K+1, K ).NE.ZERO ) THEN
675: PAIR = .TRUE.
676: END IF
677: END IF
678: *
679: IF( PAIR ) THEN
680: *
681: * Compute the eigenvalue(s) at position K.
682: *
683: WORK( 1 ) = A( K, K )
684: WORK( 2 ) = A( K+1, K )
685: WORK( 3 ) = A( K, K+1 )
686: WORK( 4 ) = A( K+1, K+1 )
687: WORK( 5 ) = B( K, K )
688: WORK( 6 ) = B( K+1, K )
689: WORK( 7 ) = B( K, K+1 )
690: WORK( 8 ) = B( K+1, K+1 )
691: CALL DLAG2( WORK, 2, WORK( 5 ), 2, SMLNUM*EPS, BETA( K ),
692: $ BETA( K+1 ), ALPHAR( K ), ALPHAR( K+1 ),
693: $ ALPHAI( K ) )
694: ALPHAI( K+1 ) = -ALPHAI( K )
695: *
696: ELSE
697: *
698: IF( SIGN( ONE, B( K, K ) ).LT.ZERO ) THEN
699: *
700: * If B(K,K) is negative, make it positive
701: *
702: DO 70 I = 1, N
703: A( K, I ) = -A( K, I )
704: B( K, I ) = -B( K, I )
705: IF( WANTQ ) Q( I, K ) = -Q( I, K )
706: 70 CONTINUE
707: END IF
708: *
709: ALPHAR( K ) = A( K, K )
710: ALPHAI( K ) = ZERO
711: BETA( K ) = B( K, K )
712: *
713: END IF
714: END IF
715: 80 CONTINUE
716: *
717: WORK( 1 ) = LWMIN
718: IWORK( 1 ) = LIWMIN
719: *
720: RETURN
721: *
722: * End of DTGSEN
723: *
724: END
CVSweb interface <joel.bertrand@systella.fr>