1 *> \brief <b> DGGES3 computes the eigenvalues, the Schur form, and, optionally, the matrix of Schur vectors for GE matrices (blocked algorithm)</b>
3 * =========== DOCUMENTATION ===========
5 * Online html documentation available at
6 * http://www.netlib.org/lapack/explore-html/
9 *> Download DGGES3 + dependencies
10 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/dgges3.f">
12 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/dgges3.f">
14 *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/dgges3.f">
21 * SUBROUTINE DGGES3( JOBVSL, JOBVSR, SORT, SELCTG, N, A, LDA, B, LDB,
22 * SDIM, ALPHAR, ALPHAI, BETA, VSL, LDVSL, VSR,
23 * LDVSR, WORK, LWORK, BWORK, INFO )
25 * .. Scalar Arguments ..
26 * CHARACTER JOBVSL, JOBVSR, SORT
27 * INTEGER INFO, LDA, LDB, LDVSL, LDVSR, LWORK, N, SDIM
29 * .. Array Arguments ..
31 * DOUBLE PRECISION A( LDA, * ), ALPHAI( * ), ALPHAR( * ),
32 * $ B( LDB, * ), BETA( * ), VSL( LDVSL, * ),
33 * $ VSR( LDVSR, * ), WORK( * )
35 * .. Function Arguments ..
46 *> DGGES3 computes for a pair of N-by-N real nonsymmetric matrices (A,B),
47 *> the generalized eigenvalues, the generalized real Schur form (S,T),
48 *> optionally, the left and/or right matrices of Schur vectors (VSL and
49 *> VSR). This gives the generalized Schur factorization
51 *> (A,B) = ( (VSL)*S*(VSR)**T, (VSL)*T*(VSR)**T )
53 *> Optionally, it also orders the eigenvalues so that a selected cluster
54 *> of eigenvalues appears in the leading diagonal blocks of the upper
55 *> quasi-triangular matrix S and the upper triangular matrix T.The
56 *> leading columns of VSL and VSR then form an orthonormal basis for the
57 *> corresponding left and right eigenspaces (deflating subspaces).
59 *> (If only the generalized eigenvalues are needed, use the driver
60 *> DGGEV instead, which is faster.)
62 *> A generalized eigenvalue for a pair of matrices (A,B) is a scalar w
63 *> or a ratio alpha/beta = w, such that A - w*B is singular. It is
64 *> usually represented as the pair (alpha,beta), as there is a
65 *> reasonable interpretation for beta=0 or both being zero.
67 *> A pair of matrices (S,T) is in generalized real Schur form if T is
68 *> upper triangular with non-negative diagonal and S is block upper
69 *> triangular with 1-by-1 and 2-by-2 blocks. 1-by-1 blocks correspond
70 *> to real generalized eigenvalues, while 2-by-2 blocks of S will be
71 *> "standardized" by making the corresponding elements of T have the
76 *> and the pair of corresponding 2-by-2 blocks in S and T will have a
77 *> complex conjugate pair of generalized eigenvalues.
86 *> JOBVSL is CHARACTER*1
87 *> = 'N': do not compute the left Schur vectors;
88 *> = 'V': compute the left Schur vectors.
93 *> JOBVSR is CHARACTER*1
94 *> = 'N': do not compute the right Schur vectors;
95 *> = 'V': compute the right Schur vectors.
100 *> SORT is CHARACTER*1
101 *> Specifies whether or not to order the eigenvalues on the
102 *> diagonal of the generalized Schur form.
103 *> = 'N': Eigenvalues are not ordered;
104 *> = 'S': Eigenvalues are ordered (see SELCTG);
109 *> SELCTG is a LOGICAL FUNCTION of three DOUBLE PRECISION arguments
110 *> SELCTG must be declared EXTERNAL in the calling subroutine.
111 *> If SORT = 'N', SELCTG is not referenced.
112 *> If SORT = 'S', SELCTG is used to select eigenvalues to sort
113 *> to the top left of the Schur form.
114 *> An eigenvalue (ALPHAR(j)+ALPHAI(j))/BETA(j) is selected if
115 *> SELCTG(ALPHAR(j),ALPHAI(j),BETA(j)) is true; i.e. if either
116 *> one of a complex conjugate pair of eigenvalues is selected,
117 *> then both complex eigenvalues are selected.
119 *> Note that in the ill-conditioned case, a selected complex
120 *> eigenvalue may no longer satisfy SELCTG(ALPHAR(j),ALPHAI(j),
121 *> BETA(j)) = .TRUE. after ordering. INFO is to be set to N+2
128 *> The order of the matrices A, B, VSL, and VSR. N >= 0.
133 *> A is DOUBLE PRECISION array, dimension (LDA, N)
134 *> On entry, the first of the pair of matrices.
135 *> On exit, A has been overwritten by its generalized Schur
142 *> The leading dimension of A. LDA >= max(1,N).
147 *> B is DOUBLE PRECISION array, dimension (LDB, N)
148 *> On entry, the second of the pair of matrices.
149 *> On exit, B has been overwritten by its generalized Schur
156 *> The leading dimension of B. LDB >= max(1,N).
162 *> If SORT = 'N', SDIM = 0.
163 *> If SORT = 'S', SDIM = number of eigenvalues (after sorting)
164 *> for which SELCTG is true. (Complex conjugate pairs for which
165 *> SELCTG is true for either eigenvalue count as 2.)
168 *> \param[out] ALPHAR
170 *> ALPHAR is DOUBLE PRECISION array, dimension (N)
173 *> \param[out] ALPHAI
175 *> ALPHAI is DOUBLE PRECISION array, dimension (N)
180 *> BETA is DOUBLE PRECISION array, dimension (N)
181 *> On exit, (ALPHAR(j) + ALPHAI(j)*i)/BETA(j), j=1,...,N, will
182 *> be the generalized eigenvalues. ALPHAR(j) + ALPHAI(j)*i,
183 *> and BETA(j),j=1,...,N are the diagonals of the complex Schur
184 *> form (S,T) that would result if the 2-by-2 diagonal blocks of
185 *> the real Schur form of (A,B) were further reduced to
186 *> triangular form using 2-by-2 complex unitary transformations.
187 *> If ALPHAI(j) is zero, then the j-th eigenvalue is real; if
188 *> positive, then the j-th and (j+1)-st eigenvalues are a
189 *> complex conjugate pair, with ALPHAI(j+1) negative.
191 *> Note: the quotients ALPHAR(j)/BETA(j) and ALPHAI(j)/BETA(j)
192 *> may easily over- or underflow, and BETA(j) may even be zero.
193 *> Thus, the user should avoid naively computing the ratio.
194 *> However, ALPHAR and ALPHAI will be always less than and
195 *> usually comparable with norm(A) in magnitude, and BETA always
196 *> less than and usually comparable with norm(B).
201 *> VSL is DOUBLE PRECISION array, dimension (LDVSL,N)
202 *> If JOBVSL = 'V', VSL will contain the left Schur vectors.
203 *> Not referenced if JOBVSL = 'N'.
209 *> The leading dimension of the matrix VSL. LDVSL >=1, and
210 *> if JOBVSL = 'V', LDVSL >= N.
215 *> VSR is DOUBLE PRECISION array, dimension (LDVSR,N)
216 *> If JOBVSR = 'V', VSR will contain the right Schur vectors.
217 *> Not referenced if JOBVSR = 'N'.
223 *> The leading dimension of the matrix VSR. LDVSR >= 1, and
224 *> if JOBVSR = 'V', LDVSR >= N.
229 *> WORK is DOUBLE PRECISION array, dimension (MAX(1,LWORK))
230 *> On exit, if INFO = 0, WORK(1) returns the optimal LWORK.
236 *> The dimension of the array WORK.
238 *> If LWORK = -1, then a workspace query is assumed; the routine
239 *> only calculates the optimal size of the WORK array, returns
240 *> this value as the first entry of the WORK array, and no error
241 *> message related to LWORK is issued by XERBLA.
246 *> BWORK is LOGICAL array, dimension (N)
247 *> Not referenced if SORT = 'N'.
253 *> = 0: successful exit
254 *> < 0: if INFO = -i, the i-th argument had an illegal value.
256 *> The QZ iteration failed. (A,B) are not in Schur
257 *> form, but ALPHAR(j), ALPHAI(j), and BETA(j) should
258 *> be correct for j=INFO+1,...,N.
259 *> > N: =N+1: other than QZ iteration failed in DHGEQZ.
260 *> =N+2: after reordering, roundoff changed values of
261 *> some complex eigenvalues so that leading
262 *> eigenvalues in the Generalized Schur form no
263 *> longer satisfy SELCTG=.TRUE. This could also
264 *> be caused due to scaling.
265 *> =N+3: reordering failed in DTGSEN.
271 *> \author Univ. of Tennessee
272 *> \author Univ. of California Berkeley
273 *> \author Univ. of Colorado Denver
276 *> \date January 2015
278 *> \ingroup doubleGEeigen
280 * =====================================================================
281 SUBROUTINE DGGES3( JOBVSL, JOBVSR, SORT, SELCTG, N, A, LDA, B,
282 $ LDB, SDIM, ALPHAR, ALPHAI, BETA, VSL, LDVSL,
283 $ VSR, LDVSR, WORK, LWORK, BWORK, INFO )
285 * -- LAPACK driver routine (version 3.6.0) --
286 * -- LAPACK is a software package provided by Univ. of Tennessee, --
287 * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
290 * .. Scalar Arguments ..
291 CHARACTER JOBVSL, JOBVSR, SORT
292 INTEGER INFO, LDA, LDB, LDVSL, LDVSR, LWORK, N, SDIM
294 * .. Array Arguments ..
296 DOUBLE PRECISION A( LDA, * ), ALPHAI( * ), ALPHAR( * ),
297 $ B( LDB, * ), BETA( * ), VSL( LDVSL, * ),
298 $ VSR( LDVSR, * ), WORK( * )
300 * .. Function Arguments ..
305 * =====================================================================
308 DOUBLE PRECISION ZERO, ONE
309 PARAMETER ( ZERO = 0.0D+0, ONE = 1.0D+0 )
311 * .. Local Scalars ..
312 LOGICAL CURSL, ILASCL, ILBSCL, ILVSL, ILVSR, LASTSL,
313 $ LQUERY, LST2SL, WANTST
314 INTEGER I, ICOLS, IERR, IHI, IJOBVL, IJOBVR, ILEFT,
315 $ ILO, IP, IRIGHT, IROWS, ITAU, IWRK, LWKOPT
316 DOUBLE PRECISION ANRM, ANRMTO, BIGNUM, BNRM, BNRMTO, EPS, PVSL,
317 $ PVSR, SAFMAX, SAFMIN, SMLNUM
321 DOUBLE PRECISION DIF( 2 )
323 * .. External Subroutines ..
324 EXTERNAL DGEQRF, DGGBAK, DGGBAL, DGGHD3, DHGEQZ, DLABAD,
325 $ DLACPY, DLASCL, DLASET, DORGQR, DORMQR, DTGSEN,
328 * .. External Functions ..
330 DOUBLE PRECISION DLAMCH, DLANGE
331 EXTERNAL LSAME, DLAMCH, DLANGE
333 * .. Intrinsic Functions ..
334 INTRINSIC ABS, MAX, SQRT
336 * .. Executable Statements ..
338 * Decode the input arguments
340 IF( LSAME( JOBVSL, 'N' ) ) THEN
343 ELSE IF( LSAME( JOBVSL, 'V' ) ) THEN
351 IF( LSAME( JOBVSR, 'N' ) ) THEN
354 ELSE IF( LSAME( JOBVSR, 'V' ) ) THEN
362 WANTST = LSAME( SORT, 'S' )
364 * Test the input arguments
367 LQUERY = ( LWORK.EQ.-1 )
368 IF( IJOBVL.LE.0 ) THEN
370 ELSE IF( IJOBVR.LE.0 ) THEN
372 ELSE IF( ( .NOT.WANTST ) .AND. ( .NOT.LSAME( SORT, 'N' ) ) ) THEN
374 ELSE IF( N.LT.0 ) THEN
376 ELSE IF( LDA.LT.MAX( 1, N ) ) THEN
378 ELSE IF( LDB.LT.MAX( 1, N ) ) THEN
380 ELSE IF( LDVSL.LT.1 .OR. ( ILVSL .AND. LDVSL.LT.N ) ) THEN
382 ELSE IF( LDVSR.LT.1 .OR. ( ILVSR .AND. LDVSR.LT.N ) ) THEN
384 ELSE IF( LWORK.LT.6*N+16 .AND. .NOT.LQUERY ) THEN
391 CALL DGEQRF( N, N, B, LDB, WORK, WORK, -1, IERR )
392 LWKOPT = MAX( 6*N+16, 3*N+INT( WORK ( 1 ) ) )
393 CALL DORMQR( 'L', 'T', N, N, N, B, LDB, WORK, A, LDA, WORK,
395 LWKOPT = MAX( LWKOPT, 3*N+INT( WORK ( 1 ) ) )
397 CALL DORGQR( N, N, N, VSL, LDVSL, WORK, WORK, -1, IERR )
398 LWKOPT = MAX( LWKOPT, 3*N+INT( WORK ( 1 ) ) )
400 CALL DGGHD3( JOBVSL, JOBVSR, N, 1, N, A, LDA, B, LDB, VSL,
401 $ LDVSL, VSR, LDVSR, WORK, -1, IERR )
402 LWKOPT = MAX( LWKOPT, 3*N+INT( WORK ( 1 ) ) )
403 CALL DHGEQZ( 'S', JOBVSL, JOBVSR, N, 1, N, A, LDA, B, LDB,
404 $ ALPHAR, ALPHAI, BETA, VSL, LDVSL, VSR, LDVSR,
406 LWKOPT = MAX( LWKOPT, 2*N+INT( WORK ( 1 ) ) )
408 CALL DTGSEN( 0, ILVSL, ILVSR, BWORK, N, A, LDA, B, LDB,
409 $ ALPHAR, ALPHAI, BETA, VSL, LDVSL, VSR, LDVSR,
410 $ SDIM, PVSL, PVSR, DIF, WORK, -1, IDUM, 1,
412 LWKOPT = MAX( LWKOPT, 2*N+INT( WORK ( 1 ) ) )
418 CALL XERBLA( 'DGGES3 ', -INFO )
420 ELSE IF( LQUERY ) THEN
424 * Quick return if possible
431 * Get machine constants
434 SAFMIN = DLAMCH( 'S' )
435 SAFMAX = ONE / SAFMIN
436 CALL DLABAD( SAFMIN, SAFMAX )
437 SMLNUM = SQRT( SAFMIN ) / EPS
438 BIGNUM = ONE / SMLNUM
440 * Scale A if max element outside range [SMLNUM,BIGNUM]
442 ANRM = DLANGE( 'M', N, N, A, LDA, WORK )
444 IF( ANRM.GT.ZERO .AND. ANRM.LT.SMLNUM ) THEN
447 ELSE IF( ANRM.GT.BIGNUM ) THEN
452 $ CALL DLASCL( 'G', 0, 0, ANRM, ANRMTO, N, N, A, LDA, IERR )
454 * Scale B if max element outside range [SMLNUM,BIGNUM]
456 BNRM = DLANGE( 'M', N, N, B, LDB, WORK )
458 IF( BNRM.GT.ZERO .AND. BNRM.LT.SMLNUM ) THEN
461 ELSE IF( BNRM.GT.BIGNUM ) THEN
466 $ CALL DLASCL( 'G', 0, 0, BNRM, BNRMTO, N, N, B, LDB, IERR )
468 * Permute the matrix to make it more nearly triangular
473 CALL DGGBAL( 'P', N, A, LDA, B, LDB, ILO, IHI, WORK( ILEFT ),
474 $ WORK( IRIGHT ), WORK( IWRK ), IERR )
476 * Reduce B to triangular form (QR decomposition of B)
478 IROWS = IHI + 1 - ILO
482 CALL DGEQRF( IROWS, ICOLS, B( ILO, ILO ), LDB, WORK( ITAU ),
483 $ WORK( IWRK ), LWORK+1-IWRK, IERR )
485 * Apply the orthogonal transformation to matrix A
487 CALL DORMQR( 'L', 'T', IROWS, ICOLS, IROWS, B( ILO, ILO ), LDB,
488 $ WORK( ITAU ), A( ILO, ILO ), LDA, WORK( IWRK ),
489 $ LWORK+1-IWRK, IERR )
494 CALL DLASET( 'Full', N, N, ZERO, ONE, VSL, LDVSL )
495 IF( IROWS.GT.1 ) THEN
496 CALL DLACPY( 'L', IROWS-1, IROWS-1, B( ILO+1, ILO ), LDB,
497 $ VSL( ILO+1, ILO ), LDVSL )
499 CALL DORGQR( IROWS, IROWS, IROWS, VSL( ILO, ILO ), LDVSL,
500 $ WORK( ITAU ), WORK( IWRK ), LWORK+1-IWRK, IERR )
506 $ CALL DLASET( 'Full', N, N, ZERO, ONE, VSR, LDVSR )
508 * Reduce to generalized Hessenberg form
510 CALL DGGHD3( JOBVSL, JOBVSR, N, ILO, IHI, A, LDA, B, LDB, VSL,
511 $ LDVSL, VSR, LDVSR, WORK( IWRK ), LWORK+1-IWRK,
514 * Perform QZ algorithm, computing Schur vectors if desired
517 CALL DHGEQZ( 'S', JOBVSL, JOBVSR, N, ILO, IHI, A, LDA, B, LDB,
518 $ ALPHAR, ALPHAI, BETA, VSL, LDVSL, VSR, LDVSR,
519 $ WORK( IWRK ), LWORK+1-IWRK, IERR )
521 IF( IERR.GT.0 .AND. IERR.LE.N ) THEN
523 ELSE IF( IERR.GT.N .AND. IERR.LE.2*N ) THEN
531 * Sort eigenvalues ALPHA/BETA if desired
536 * Undo scaling on eigenvalues before SELCTGing
539 CALL DLASCL( 'G', 0, 0, ANRMTO, ANRM, N, 1, ALPHAR, N,
541 CALL DLASCL( 'G', 0, 0, ANRMTO, ANRM, N, 1, ALPHAI, N,
545 $ CALL DLASCL( 'G', 0, 0, BNRMTO, BNRM, N, 1, BETA, N, IERR )
550 BWORK( I ) = SELCTG( ALPHAR( I ), ALPHAI( I ), BETA( I ) )
553 CALL DTGSEN( 0, ILVSL, ILVSR, BWORK, N, A, LDA, B, LDB, ALPHAR,
554 $ ALPHAI, BETA, VSL, LDVSL, VSR, LDVSR, SDIM, PVSL,
555 $ PVSR, DIF, WORK( IWRK ), LWORK-IWRK+1, IDUM, 1,
562 * Apply back-permutation to VSL and VSR
565 $ CALL DGGBAK( 'P', 'L', N, ILO, IHI, WORK( ILEFT ),
566 $ WORK( IRIGHT ), N, VSL, LDVSL, IERR )
569 $ CALL DGGBAK( 'P', 'R', N, ILO, IHI, WORK( ILEFT ),
570 $ WORK( IRIGHT ), N, VSR, LDVSR, IERR )
572 * Check if unscaling would cause over/underflow, if so, rescale
573 * (ALPHAR(I),ALPHAI(I),BETA(I)) so BETA(I) is on the order of
574 * B(I,I) and ALPHAR(I) and ALPHAI(I) are on the order of A(I,I)
578 IF( ALPHAI( I ).NE.ZERO ) THEN
579 IF( ( ALPHAR( I ) / SAFMAX ).GT.( ANRMTO / ANRM ) .OR.
580 $ ( SAFMIN / ALPHAR( I ) ).GT.( ANRM / ANRMTO ) ) THEN
581 WORK( 1 ) = ABS( A( I, I ) / ALPHAR( I ) )
582 BETA( I ) = BETA( I )*WORK( 1 )
583 ALPHAR( I ) = ALPHAR( I )*WORK( 1 )
584 ALPHAI( I ) = ALPHAI( I )*WORK( 1 )
585 ELSE IF( ( ALPHAI( I ) / SAFMAX ).GT.
586 $ ( ANRMTO / ANRM ) .OR.
587 $ ( SAFMIN / ALPHAI( I ) ).GT.( ANRM / ANRMTO ) )
589 WORK( 1 ) = ABS( A( I, I+1 ) / ALPHAI( I ) )
590 BETA( I ) = BETA( I )*WORK( 1 )
591 ALPHAR( I ) = ALPHAR( I )*WORK( 1 )
592 ALPHAI( I ) = ALPHAI( I )*WORK( 1 )
600 IF( ALPHAI( I ).NE.ZERO ) THEN
601 IF( ( BETA( I ) / SAFMAX ).GT.( BNRMTO / BNRM ) .OR.
602 $ ( SAFMIN / BETA( I ) ).GT.( BNRM / BNRMTO ) ) THEN
603 WORK( 1 ) = ABS( B( I, I ) / BETA( I ) )
604 BETA( I ) = BETA( I )*WORK( 1 )
605 ALPHAR( I ) = ALPHAR( I )*WORK( 1 )
606 ALPHAI( I ) = ALPHAI( I )*WORK( 1 )
615 CALL DLASCL( 'H', 0, 0, ANRMTO, ANRM, N, N, A, LDA, IERR )
616 CALL DLASCL( 'G', 0, 0, ANRMTO, ANRM, N, 1, ALPHAR, N, IERR )
617 CALL DLASCL( 'G', 0, 0, ANRMTO, ANRM, N, 1, ALPHAI, N, IERR )
621 CALL DLASCL( 'U', 0, 0, BNRMTO, BNRM, N, N, B, LDB, IERR )
622 CALL DLASCL( 'G', 0, 0, BNRMTO, BNRM, N, 1, BETA, N, IERR )
627 * Check if reordering is correct
634 CURSL = SELCTG( ALPHAR( I ), ALPHAI( I ), BETA( I ) )
635 IF( ALPHAI( I ).EQ.ZERO ) THEN
639 IF( CURSL .AND. .NOT.LASTSL )
644 * Last eigenvalue of conjugate pair
646 CURSL = CURSL .OR. LASTSL
651 IF( CURSL .AND. .NOT.LST2SL )
655 * First eigenvalue of conjugate pair