LAPACK 3.12.0
LAPACK: Linear Algebra PACKage
Loading...
Searching...
No Matches
slaqr2.f
Go to the documentation of this file.
1*> \brief \b SLAQR2 performs the orthogonal similarity transformation of a Hessenberg matrix to detect and deflate fully converged eigenvalues from a trailing principal submatrix (aggressive early deflation).
2*
3* =========== DOCUMENTATION ===========
4*
5* Online html documentation available at
6* http://www.netlib.org/lapack/explore-html/
7*
8*> \htmlonly
9*> Download SLAQR2 + dependencies
10*> <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/slaqr2.f">
11*> [TGZ]</a>
12*> <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/slaqr2.f">
13*> [ZIP]</a>
14*> <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/slaqr2.f">
15*> [TXT]</a>
16*> \endhtmlonly
17*
18* Definition:
19* ===========
20*
21* SUBROUTINE SLAQR2( WANTT, WANTZ, N, KTOP, KBOT, NW, H, LDH, ILOZ,
22* IHIZ, Z, LDZ, NS, ND, SR, SI, V, LDV, NH, T,
23* LDT, NV, WV, LDWV, WORK, LWORK )
24*
25* .. Scalar Arguments ..
26* INTEGER IHIZ, ILOZ, KBOT, KTOP, LDH, LDT, LDV, LDWV,
27* $ LDZ, LWORK, N, ND, NH, NS, NV, NW
28* LOGICAL WANTT, WANTZ
29* ..
30* .. Array Arguments ..
31* REAL H( LDH, * ), SI( * ), SR( * ), T( LDT, * ),
32* $ V( LDV, * ), WORK( * ), WV( LDWV, * ),
33* $ Z( LDZ, * )
34* ..
35*
36*
37*> \par Purpose:
38* =============
39*>
40*> \verbatim
41*>
42*> SLAQR2 is identical to SLAQR3 except that it avoids
43*> recursion by calling SLAHQR instead of SLAQR4.
44*>
45*> Aggressive early deflation:
46*>
47*> This subroutine accepts as input an upper Hessenberg matrix
48*> H and performs an orthogonal similarity transformation
49*> designed to detect and deflate fully converged eigenvalues from
50*> a trailing principal submatrix. On output H has been over-
51*> written by a new Hessenberg matrix that is a perturbation of
52*> an orthogonal similarity transformation of H. It is to be
53*> hoped that the final version of H has many zero subdiagonal
54*> entries.
55*> \endverbatim
56*
57* Arguments:
58* ==========
59*
60*> \param[in] WANTT
61*> \verbatim
62*> WANTT is LOGICAL
63*> If .TRUE., then the Hessenberg matrix H is fully updated
64*> so that the quasi-triangular Schur factor may be
65*> computed (in cooperation with the calling subroutine).
66*> If .FALSE., then only enough of H is updated to preserve
67*> the eigenvalues.
68*> \endverbatim
69*>
70*> \param[in] WANTZ
71*> \verbatim
72*> WANTZ is LOGICAL
73*> If .TRUE., then the orthogonal matrix Z is updated so
74*> so that the orthogonal Schur factor may be computed
75*> (in cooperation with the calling subroutine).
76*> If .FALSE., then Z is not referenced.
77*> \endverbatim
78*>
79*> \param[in] N
80*> \verbatim
81*> N is INTEGER
82*> The order of the matrix H and (if WANTZ is .TRUE.) the
83*> order of the orthogonal matrix Z.
84*> \endverbatim
85*>
86*> \param[in] KTOP
87*> \verbatim
88*> KTOP is INTEGER
89*> It is assumed that either KTOP = 1 or H(KTOP,KTOP-1)=0.
90*> KBOT and KTOP together determine an isolated block
91*> along the diagonal of the Hessenberg matrix.
92*> \endverbatim
93*>
94*> \param[in] KBOT
95*> \verbatim
96*> KBOT is INTEGER
97*> It is assumed without a check that either
98*> KBOT = N or H(KBOT+1,KBOT)=0. KBOT and KTOP together
99*> determine an isolated block along the diagonal of the
100*> Hessenberg matrix.
101*> \endverbatim
102*>
103*> \param[in] NW
104*> \verbatim
105*> NW is INTEGER
106*> Deflation window size. 1 <= NW <= (KBOT-KTOP+1).
107*> \endverbatim
108*>
109*> \param[in,out] H
110*> \verbatim
111*> H is REAL array, dimension (LDH,N)
112*> On input the initial N-by-N section of H stores the
113*> Hessenberg matrix undergoing aggressive early deflation.
114*> On output H has been transformed by an orthogonal
115*> similarity transformation, perturbed, and the returned
116*> to Hessenberg form that (it is to be hoped) has some
117*> zero subdiagonal entries.
118*> \endverbatim
119*>
120*> \param[in] LDH
121*> \verbatim
122*> LDH is INTEGER
123*> Leading dimension of H just as declared in the calling
124*> subroutine. N <= LDH
125*> \endverbatim
126*>
127*> \param[in] ILOZ
128*> \verbatim
129*> ILOZ is INTEGER
130*> \endverbatim
131*>
132*> \param[in] IHIZ
133*> \verbatim
134*> IHIZ is INTEGER
135*> Specify the rows of Z to which transformations must be
136*> applied if WANTZ is .TRUE.. 1 <= ILOZ <= IHIZ <= N.
137*> \endverbatim
138*>
139*> \param[in,out] Z
140*> \verbatim
141*> Z is REAL array, dimension (LDZ,N)
142*> IF WANTZ is .TRUE., then on output, the orthogonal
143*> similarity transformation mentioned above has been
144*> accumulated into Z(ILOZ:IHIZ,ILOZ:IHIZ) from the right.
145*> If WANTZ is .FALSE., then Z is unreferenced.
146*> \endverbatim
147*>
148*> \param[in] LDZ
149*> \verbatim
150*> LDZ is INTEGER
151*> The leading dimension of Z just as declared in the
152*> calling subroutine. 1 <= LDZ.
153*> \endverbatim
154*>
155*> \param[out] NS
156*> \verbatim
157*> NS is INTEGER
158*> The number of unconverged (ie approximate) eigenvalues
159*> returned in SR and SI that may be used as shifts by the
160*> calling subroutine.
161*> \endverbatim
162*>
163*> \param[out] ND
164*> \verbatim
165*> ND is INTEGER
166*> The number of converged eigenvalues uncovered by this
167*> subroutine.
168*> \endverbatim
169*>
170*> \param[out] SR
171*> \verbatim
172*> SR is REAL array, dimension (KBOT)
173*> \endverbatim
174*>
175*> \param[out] SI
176*> \verbatim
177*> SI is REAL array, dimension (KBOT)
178*> On output, the real and imaginary parts of approximate
179*> eigenvalues that may be used for shifts are stored in
180*> SR(KBOT-ND-NS+1) through SR(KBOT-ND) and
181*> SI(KBOT-ND-NS+1) through SI(KBOT-ND), respectively.
182*> The real and imaginary parts of converged eigenvalues
183*> are stored in SR(KBOT-ND+1) through SR(KBOT) and
184*> SI(KBOT-ND+1) through SI(KBOT), respectively.
185*> \endverbatim
186*>
187*> \param[out] V
188*> \verbatim
189*> V is REAL array, dimension (LDV,NW)
190*> An NW-by-NW work array.
191*> \endverbatim
192*>
193*> \param[in] LDV
194*> \verbatim
195*> LDV is INTEGER
196*> The leading dimension of V just as declared in the
197*> calling subroutine. NW <= LDV
198*> \endverbatim
199*>
200*> \param[in] NH
201*> \verbatim
202*> NH is INTEGER
203*> The number of columns of T. NH >= NW.
204*> \endverbatim
205*>
206*> \param[out] T
207*> \verbatim
208*> T is REAL array, dimension (LDT,NW)
209*> \endverbatim
210*>
211*> \param[in] LDT
212*> \verbatim
213*> LDT is INTEGER
214*> The leading dimension of T just as declared in the
215*> calling subroutine. NW <= LDT
216*> \endverbatim
217*>
218*> \param[in] NV
219*> \verbatim
220*> NV is INTEGER
221*> The number of rows of work array WV available for
222*> workspace. NV >= NW.
223*> \endverbatim
224*>
225*> \param[out] WV
226*> \verbatim
227*> WV is REAL array, dimension (LDWV,NW)
228*> \endverbatim
229*>
230*> \param[in] LDWV
231*> \verbatim
232*> LDWV is INTEGER
233*> The leading dimension of W just as declared in the
234*> calling subroutine. NW <= LDV
235*> \endverbatim
236*>
237*> \param[out] WORK
238*> \verbatim
239*> WORK is REAL array, dimension (LWORK)
240*> On exit, WORK(1) is set to an estimate of the optimal value
241*> of LWORK for the given values of N, NW, KTOP and KBOT.
242*> \endverbatim
243*>
244*> \param[in] LWORK
245*> \verbatim
246*> LWORK is INTEGER
247*> The dimension of the work array WORK. LWORK = 2*NW
248*> suffices, but greater efficiency may result from larger
249*> values of LWORK.
250*>
251*> If LWORK = -1, then a workspace query is assumed; SLAQR2
252*> only estimates the optimal workspace size for the given
253*> values of N, NW, KTOP and KBOT. The estimate is returned
254*> in WORK(1). No error message related to LWORK is issued
255*> by XERBLA. Neither H nor Z are accessed.
256*> \endverbatim
257*
258* Authors:
259* ========
260*
261*> \author Univ. of Tennessee
262*> \author Univ. of California Berkeley
263*> \author Univ. of Colorado Denver
264*> \author NAG Ltd.
265*
266*> \ingroup laqr2
267*
268*> \par Contributors:
269* ==================
270*>
271*> Karen Braman and Ralph Byers, Department of Mathematics,
272*> University of Kansas, USA
273*>
274* =====================================================================
275 SUBROUTINE slaqr2( WANTT, WANTZ, N, KTOP, KBOT, NW, H, LDH, ILOZ,
276 $ IHIZ, Z, LDZ, NS, ND, SR, SI, V, LDV, NH, T,
277 $ LDT, NV, WV, LDWV, WORK, LWORK )
278*
279* -- LAPACK auxiliary routine --
280* -- LAPACK is a software package provided by Univ. of Tennessee, --
281* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
282*
283* .. Scalar Arguments ..
284 INTEGER IHIZ, ILOZ, KBOT, KTOP, LDH, LDT, LDV, LDWV,
285 $ LDZ, LWORK, N, ND, NH, NS, NV, NW
286 LOGICAL WANTT, WANTZ
287* ..
288* .. Array Arguments ..
289 REAL H( LDH, * ), SI( * ), SR( * ), T( LDT, * ),
290 $ V( LDV, * ), WORK( * ), WV( LDWV, * ),
291 $ z( ldz, * )
292* ..
293*
294* ================================================================
295* .. Parameters ..
296 REAL ZERO, ONE
297 PARAMETER ( ZERO = 0.0e0, one = 1.0e0 )
298* ..
299* .. Local Scalars ..
300 REAL AA, BB, BETA, CC, CS, DD, EVI, EVK, FOO, S,
301 $ SAFMAX, SAFMIN, SMLNUM, SN, TAU, ULP
302 INTEGER I, IFST, ILST, INFO, INFQR, J, JW, K, KCOL,
303 $ KEND, KLN, KROW, KWTOP, LTOP, LWK1, LWK2,
304 $ lwkopt
305 LOGICAL BULGE, SORTED
306* ..
307* .. External Functions ..
308 REAL SLAMCH, SROUNDUP_LWORK
309 EXTERNAL SLAMCH, SROUNDUP_LWORK
310* ..
311* .. External Subroutines ..
312 EXTERNAL scopy, sgehrd, sgemm, slacpy, slahqr,
314* ..
315* .. Intrinsic Functions ..
316 INTRINSIC abs, int, max, min, real, sqrt
317* ..
318* .. Executable Statements ..
319*
320* ==== Estimate optimal workspace. ====
321*
322 jw = min( nw, kbot-ktop+1 )
323 IF( jw.LE.2 ) THEN
324 lwkopt = 1
325 ELSE
326*
327* ==== Workspace query call to SGEHRD ====
328*
329 CALL sgehrd( jw, 1, jw-1, t, ldt, work, work, -1, info )
330 lwk1 = int( work( 1 ) )
331*
332* ==== Workspace query call to SORMHR ====
333*
334 CALL sormhr( 'R', 'N', jw, jw, 1, jw-1, t, ldt, work, v, ldv,
335 $ work, -1, info )
336 lwk2 = int( work( 1 ) )
337*
338* ==== Optimal workspace ====
339*
340 lwkopt = jw + max( lwk1, lwk2 )
341 END IF
342*
343* ==== Quick return in case of workspace query. ====
344*
345 IF( lwork.EQ.-1 ) THEN
346 work( 1 ) = sroundup_lwork( lwkopt )
347 RETURN
348 END IF
349*
350* ==== Nothing to do ...
351* ... for an empty active block ... ====
352 ns = 0
353 nd = 0
354 work( 1 ) = one
355 IF( ktop.GT.kbot )
356 $ RETURN
357* ... nor for an empty deflation window. ====
358 IF( nw.LT.1 )
359 $ RETURN
360*
361* ==== Machine constants ====
362*
363 safmin = slamch( 'SAFE MINIMUM' )
364 safmax = one / safmin
365 ulp = slamch( 'PRECISION' )
366 smlnum = safmin*( real( n ) / ulp )
367*
368* ==== Setup deflation window ====
369*
370 jw = min( nw, kbot-ktop+1 )
371 kwtop = kbot - jw + 1
372 IF( kwtop.EQ.ktop ) THEN
373 s = zero
374 ELSE
375 s = h( kwtop, kwtop-1 )
376 END IF
377*
378 IF( kbot.EQ.kwtop ) THEN
379*
380* ==== 1-by-1 deflation window: not much to do ====
381*
382 sr( kwtop ) = h( kwtop, kwtop )
383 si( kwtop ) = zero
384 ns = 1
385 nd = 0
386 IF( abs( s ).LE.max( smlnum, ulp*abs( h( kwtop, kwtop ) ) ) )
387 $ THEN
388 ns = 0
389 nd = 1
390 IF( kwtop.GT.ktop )
391 $ h( kwtop, kwtop-1 ) = zero
392 END IF
393 work( 1 ) = one
394 RETURN
395 END IF
396*
397* ==== Convert to spike-triangular form. (In case of a
398* . rare QR failure, this routine continues to do
399* . aggressive early deflation using that part of
400* . the deflation window that converged using INFQR
401* . here and there to keep track.) ====
402*
403 CALL slacpy( 'U', jw, jw, h( kwtop, kwtop ), ldh, t, ldt )
404 CALL scopy( jw-1, h( kwtop+1, kwtop ), ldh+1, t( 2, 1 ), ldt+1 )
405*
406 CALL slaset( 'A', jw, jw, zero, one, v, ldv )
407 CALL slahqr( .true., .true., jw, 1, jw, t, ldt, sr( kwtop ),
408 $ si( kwtop ), 1, jw, v, ldv, infqr )
409*
410* ==== STREXC needs a clean margin near the diagonal ====
411*
412 DO 10 j = 1, jw - 3
413 t( j+2, j ) = zero
414 t( j+3, j ) = zero
415 10 CONTINUE
416 IF( jw.GT.2 )
417 $ t( jw, jw-2 ) = zero
418*
419* ==== Deflation detection loop ====
420*
421 ns = jw
422 ilst = infqr + 1
423 20 CONTINUE
424 IF( ilst.LE.ns ) THEN
425 IF( ns.EQ.1 ) THEN
426 bulge = .false.
427 ELSE
428 bulge = t( ns, ns-1 ).NE.zero
429 END IF
430*
431* ==== Small spike tip test for deflation ====
432*
433 IF( .NOT.bulge ) THEN
434*
435* ==== Real eigenvalue ====
436*
437 foo = abs( t( ns, ns ) )
438 IF( foo.EQ.zero )
439 $ foo = abs( s )
440 IF( abs( s*v( 1, ns ) ).LE.max( smlnum, ulp*foo ) ) THEN
441*
442* ==== Deflatable ====
443*
444 ns = ns - 1
445 ELSE
446*
447* ==== Undeflatable. Move it up out of the way.
448* . (STREXC can not fail in this case.) ====
449*
450 ifst = ns
451 CALL strexc( 'V', jw, t, ldt, v, ldv, ifst, ilst, work,
452 $ info )
453 ilst = ilst + 1
454 END IF
455 ELSE
456*
457* ==== Complex conjugate pair ====
458*
459 foo = abs( t( ns, ns ) ) + sqrt( abs( t( ns, ns-1 ) ) )*
460 $ sqrt( abs( t( ns-1, ns ) ) )
461 IF( foo.EQ.zero )
462 $ foo = abs( s )
463 IF( max( abs( s*v( 1, ns ) ), abs( s*v( 1, ns-1 ) ) ).LE.
464 $ max( smlnum, ulp*foo ) ) THEN
465*
466* ==== Deflatable ====
467*
468 ns = ns - 2
469 ELSE
470*
471* ==== Undeflatable. Move them up out of the way.
472* . Fortunately, STREXC does the right thing with
473* . ILST in case of a rare exchange failure. ====
474*
475 ifst = ns
476 CALL strexc( 'V', jw, t, ldt, v, ldv, ifst, ilst, work,
477 $ info )
478 ilst = ilst + 2
479 END IF
480 END IF
481*
482* ==== End deflation detection loop ====
483*
484 GO TO 20
485 END IF
486*
487* ==== Return to Hessenberg form ====
488*
489 IF( ns.EQ.0 )
490 $ s = zero
491*
492 IF( ns.LT.jw ) THEN
493*
494* ==== sorting diagonal blocks of T improves accuracy for
495* . graded matrices. Bubble sort deals well with
496* . exchange failures. ====
497*
498 sorted = .false.
499 i = ns + 1
500 30 CONTINUE
501 IF( sorted )
502 $ GO TO 50
503 sorted = .true.
504*
505 kend = i - 1
506 i = infqr + 1
507 IF( i.EQ.ns ) THEN
508 k = i + 1
509 ELSE IF( t( i+1, i ).EQ.zero ) THEN
510 k = i + 1
511 ELSE
512 k = i + 2
513 END IF
514 40 CONTINUE
515 IF( k.LE.kend ) THEN
516 IF( k.EQ.i+1 ) THEN
517 evi = abs( t( i, i ) )
518 ELSE
519 evi = abs( t( i, i ) ) + sqrt( abs( t( i+1, i ) ) )*
520 $ sqrt( abs( t( i, i+1 ) ) )
521 END IF
522*
523 IF( k.EQ.kend ) THEN
524 evk = abs( t( k, k ) )
525 ELSE IF( t( k+1, k ).EQ.zero ) THEN
526 evk = abs( t( k, k ) )
527 ELSE
528 evk = abs( t( k, k ) ) + sqrt( abs( t( k+1, k ) ) )*
529 $ sqrt( abs( t( k, k+1 ) ) )
530 END IF
531*
532 IF( evi.GE.evk ) THEN
533 i = k
534 ELSE
535 sorted = .false.
536 ifst = i
537 ilst = k
538 CALL strexc( 'V', jw, t, ldt, v, ldv, ifst, ilst, work,
539 $ info )
540 IF( info.EQ.0 ) THEN
541 i = ilst
542 ELSE
543 i = k
544 END IF
545 END IF
546 IF( i.EQ.kend ) THEN
547 k = i + 1
548 ELSE IF( t( i+1, i ).EQ.zero ) THEN
549 k = i + 1
550 ELSE
551 k = i + 2
552 END IF
553 GO TO 40
554 END IF
555 GO TO 30
556 50 CONTINUE
557 END IF
558*
559* ==== Restore shift/eigenvalue array from T ====
560*
561 i = jw
562 60 CONTINUE
563 IF( i.GE.infqr+1 ) THEN
564 IF( i.EQ.infqr+1 ) THEN
565 sr( kwtop+i-1 ) = t( i, i )
566 si( kwtop+i-1 ) = zero
567 i = i - 1
568 ELSE IF( t( i, i-1 ).EQ.zero ) THEN
569 sr( kwtop+i-1 ) = t( i, i )
570 si( kwtop+i-1 ) = zero
571 i = i - 1
572 ELSE
573 aa = t( i-1, i-1 )
574 cc = t( i, i-1 )
575 bb = t( i-1, i )
576 dd = t( i, i )
577 CALL slanv2( aa, bb, cc, dd, sr( kwtop+i-2 ),
578 $ si( kwtop+i-2 ), sr( kwtop+i-1 ),
579 $ si( kwtop+i-1 ), cs, sn )
580 i = i - 2
581 END IF
582 GO TO 60
583 END IF
584*
585 IF( ns.LT.jw .OR. s.EQ.zero ) THEN
586 IF( ns.GT.1 .AND. s.NE.zero ) THEN
587*
588* ==== Reflect spike back into lower triangle ====
589*
590 CALL scopy( ns, v, ldv, work, 1 )
591 beta = work( 1 )
592 CALL slarfg( ns, beta, work( 2 ), 1, tau )
593 work( 1 ) = one
594*
595 CALL slaset( 'L', jw-2, jw-2, zero, zero, t( 3, 1 ), ldt )
596*
597 CALL slarf( 'L', ns, jw, work, 1, tau, t, ldt,
598 $ work( jw+1 ) )
599 CALL slarf( 'R', ns, ns, work, 1, tau, t, ldt,
600 $ work( jw+1 ) )
601 CALL slarf( 'R', jw, ns, work, 1, tau, v, ldv,
602 $ work( jw+1 ) )
603*
604 CALL sgehrd( jw, 1, ns, t, ldt, work, work( jw+1 ),
605 $ lwork-jw, info )
606 END IF
607*
608* ==== Copy updated reduced window into place ====
609*
610 IF( kwtop.GT.1 )
611 $ h( kwtop, kwtop-1 ) = s*v( 1, 1 )
612 CALL slacpy( 'U', jw, jw, t, ldt, h( kwtop, kwtop ), ldh )
613 CALL scopy( jw-1, t( 2, 1 ), ldt+1, h( kwtop+1, kwtop ),
614 $ ldh+1 )
615*
616* ==== Accumulate orthogonal matrix in order update
617* . H and Z, if requested. ====
618*
619 IF( ns.GT.1 .AND. s.NE.zero )
620 $ CALL sormhr( 'R', 'N', jw, ns, 1, ns, t, ldt, work, v, ldv,
621 $ work( jw+1 ), lwork-jw, info )
622*
623* ==== Update vertical slab in H ====
624*
625 IF( wantt ) THEN
626 ltop = 1
627 ELSE
628 ltop = ktop
629 END IF
630 DO 70 krow = ltop, kwtop - 1, nv
631 kln = min( nv, kwtop-krow )
632 CALL sgemm( 'N', 'N', kln, jw, jw, one, h( krow, kwtop ),
633 $ ldh, v, ldv, zero, wv, ldwv )
634 CALL slacpy( 'A', kln, jw, wv, ldwv, h( krow, kwtop ), ldh )
635 70 CONTINUE
636*
637* ==== Update horizontal slab in H ====
638*
639 IF( wantt ) THEN
640 DO 80 kcol = kbot + 1, n, nh
641 kln = min( nh, n-kcol+1 )
642 CALL sgemm( 'C', 'N', jw, kln, jw, one, v, ldv,
643 $ h( kwtop, kcol ), ldh, zero, t, ldt )
644 CALL slacpy( 'A', jw, kln, t, ldt, h( kwtop, kcol ),
645 $ ldh )
646 80 CONTINUE
647 END IF
648*
649* ==== Update vertical slab in Z ====
650*
651 IF( wantz ) THEN
652 DO 90 krow = iloz, ihiz, nv
653 kln = min( nv, ihiz-krow+1 )
654 CALL sgemm( 'N', 'N', kln, jw, jw, one, z( krow, kwtop ),
655 $ ldz, v, ldv, zero, wv, ldwv )
656 CALL slacpy( 'A', kln, jw, wv, ldwv, z( krow, kwtop ),
657 $ ldz )
658 90 CONTINUE
659 END IF
660 END IF
661*
662* ==== Return the number of deflations ... ====
663*
664 nd = jw - ns
665*
666* ==== ... and the number of shifts. (Subtracting
667* . INFQR from the spike length takes care
668* . of the case of a rare QR failure while
669* . calculating eigenvalues of the deflation
670* . window.) ====
671*
672 ns = ns - infqr
673*
674* ==== Return optimal workspace. ====
675*
676 work( 1 ) = sroundup_lwork( lwkopt )
677*
678* ==== End of SLAQR2 ====
679*
680 END
subroutine scopy(n, sx, incx, sy, incy)
SCOPY
Definition scopy.f:82
subroutine sgehrd(n, ilo, ihi, a, lda, tau, work, lwork, info)
SGEHRD
Definition sgehrd.f:167
subroutine sgemm(transa, transb, m, n, k, alpha, a, lda, b, ldb, beta, c, ldc)
SGEMM
Definition sgemm.f:188
subroutine slacpy(uplo, m, n, a, lda, b, ldb)
SLACPY copies all or part of one two-dimensional array to another.
Definition slacpy.f:103
subroutine slahqr(wantt, wantz, n, ilo, ihi, h, ldh, wr, wi, iloz, ihiz, z, ldz, info)
SLAHQR computes the eigenvalues and Schur factorization of an upper Hessenberg matrix,...
Definition slahqr.f:207
subroutine slanv2(a, b, c, d, rt1r, rt1i, rt2r, rt2i, cs, sn)
SLANV2 computes the Schur factorization of a real 2-by-2 nonsymmetric matrix in standard form.
Definition slanv2.f:127
subroutine slaqr2(wantt, wantz, n, ktop, kbot, nw, h, ldh, iloz, ihiz, z, ldz, ns, nd, sr, si, v, ldv, nh, t, ldt, nv, wv, ldwv, work, lwork)
SLAQR2 performs the orthogonal similarity transformation of a Hessenberg matrix to detect and deflate...
Definition slaqr2.f:278
subroutine slarf(side, m, n, v, incv, tau, c, ldc, work)
SLARF applies an elementary reflector to a general rectangular matrix.
Definition slarf.f:124
subroutine slarfg(n, alpha, x, incx, tau)
SLARFG generates an elementary reflector (Householder matrix).
Definition slarfg.f:106
subroutine slaset(uplo, m, n, alpha, beta, a, lda)
SLASET initializes the off-diagonal elements and the diagonal elements of a matrix to given values.
Definition slaset.f:110
subroutine strexc(compq, n, t, ldt, q, ldq, ifst, ilst, work, info)
STREXC
Definition strexc.f:148
subroutine sormhr(side, trans, m, n, ilo, ihi, a, lda, tau, c, ldc, work, lwork, info)
SORMHR
Definition sormhr.f:179