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
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
|
*> \brief \b SLAQR4
*
* =========== DOCUMENTATION ===========
*
* Online html documentation available at
* http://www.netlib.org/lapack/explore-html/
*
*> Download SLAQR4 + dependencies
*> <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/slaqr4.f">
*> [TGZ]</a>
*> <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/slaqr4.f">
*> [ZIP]</a>
*> <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/slaqr4.f">
*> [TXT]</a>
*
* Definition
* ==========
*
* SUBROUTINE SLAQR4( WANTT, WANTZ, N, ILO, IHI, H, LDH, WR, WI,
* ILOZ, IHIZ, Z, LDZ, WORK, LWORK, INFO )
*
* .. Scalar Arguments ..
* INTEGER IHI, IHIZ, ILO, ILOZ, INFO, LDH, LDZ, LWORK, N
* LOGICAL WANTT, WANTZ
* ..
* .. Array Arguments ..
* REAL H( LDH, * ), WI( * ), WORK( * ), WR( * ),
* $ Z( LDZ, * )
* ..
*
* Purpose
* =======
*
*>\details \b Purpose:
*>\verbatim
*>
*> SLAQR4 implements one level of recursion for SLAQR0.
*> It is a complete implementation of the small bulge multi-shift
*> QR algorithm. It may be called by SLAQR0 and, for large enough
*> deflation window size, it may be called by SLAQR3. This
*> subroutine is identical to SLAQR0 except that it calls SLAQR2
*> instead of SLAQR3.
*>
*> SLAQR4 computes the eigenvalues of a Hessenberg matrix H
*> and, optionally, the matrices T and Z from the Schur decomposition
*> H = Z T Z**T, where T is an upper quasi-triangular matrix (the
*> Schur form), and Z is the orthogonal matrix of Schur vectors.
*>
*> Optionally Z may be postmultiplied into an input orthogonal
*> matrix Q so that this routine can give the Schur factorization
*> of a matrix A which has been reduced to the Hessenberg form H
*> by the orthogonal matrix Q: A = Q*H*Q**T = (QZ)*T*(QZ)**T.
*>
*>\endverbatim
*
* Arguments
* =========
*
*> \param[in] WANTT
*> \verbatim
*> WANTT is LOGICAL
*> = .TRUE. : the full Schur form T is required;
*> = .FALSE.: only eigenvalues are required.
*> \endverbatim
*>
*> \param[in] WANTZ
*> \verbatim
*> WANTZ is LOGICAL
*> = .TRUE. : the matrix of Schur vectors Z is required;
*> = .FALSE.: Schur vectors are not required.
*> \endverbatim
*>
*> \param[in] N
*> \verbatim
*> N is INTEGER
*> The order of the matrix H. N .GE. 0.
*> \endverbatim
*>
*> \param[in] ILO
*> \verbatim
*> ILO is INTEGER
*> \endverbatim
*>
*> \param[in] IHI
*> \verbatim
*> IHI is INTEGER
*> It is assumed that H is already upper triangular in rows
*> and columns 1:ILO-1 and IHI+1:N and, if ILO.GT.1,
*> H(ILO,ILO-1) is zero. ILO and IHI are normally set by a
*> previous call to SGEBAL, and then passed to SGEHRD when the
*> matrix output by SGEBAL is reduced to Hessenberg form.
*> Otherwise, ILO and IHI should be set to 1 and N,
*> respectively. If N.GT.0, then 1.LE.ILO.LE.IHI.LE.N.
*> If N = 0, then ILO = 1 and IHI = 0.
*> \endverbatim
*>
*> \param[in,out] H
*> \verbatim
*> H is REAL array, dimension (LDH,N)
*> On entry, the upper Hessenberg matrix H.
*> On exit, if INFO = 0 and WANTT is .TRUE., then H contains
*> the upper quasi-triangular matrix T from the Schur
*> decomposition (the Schur form); 2-by-2 diagonal blocks
*> (corresponding to complex conjugate pairs of eigenvalues)
*> are returned in standard form, with H(i,i) = H(i+1,i+1)
*> and H(i+1,i)*H(i,i+1).LT.0. If INFO = 0 and WANTT is
*> .FALSE., then the contents of H are unspecified on exit.
*> (The output value of H when INFO.GT.0 is given under the
*> description of INFO below.)
*> \endverbatim
*> \verbatim
*> This subroutine may explicitly set H(i,j) = 0 for i.GT.j and
*> j = 1, 2, ... ILO-1 or j = IHI+1, IHI+2, ... N.
*> \endverbatim
*>
*> \param[in] LDH
*> \verbatim
*> LDH is INTEGER
*> The leading dimension of the array H. LDH .GE. max(1,N).
*> \endverbatim
*>
*> \param[out] WR
*> \verbatim
*> WR is REAL array, dimension (IHI)
*> \endverbatim
*>
*> \param[out] WI
*> \verbatim
*> WI is REAL array, dimension (IHI)
*> The real and imaginary parts, respectively, of the computed
*> eigenvalues of H(ILO:IHI,ILO:IHI) are stored in WR(ILO:IHI)
*> and WI(ILO:IHI). If two eigenvalues are computed as a
*> complex conjugate pair, they are stored in consecutive
*> elements of WR and WI, say the i-th and (i+1)th, with
*> WI(i) .GT. 0 and WI(i+1) .LT. 0. If WANTT is .TRUE., then
*> the eigenvalues are stored in the same order as on the
*> diagonal of the Schur form returned in H, with
*> WR(i) = H(i,i) and, if H(i:i+1,i:i+1) is a 2-by-2 diagonal
*> block, WI(i) = sqrt(-H(i+1,i)*H(i,i+1)) and
*> WI(i+1) = -WI(i).
*> \endverbatim
*>
*> \param[in] ILOZ
*> \verbatim
*> ILOZ is INTEGER
*> \endverbatim
*>
*> \param[in] IHIZ
*> \verbatim
*> IHIZ is INTEGER
*> Specify the rows of Z to which transformations must be
*> applied if WANTZ is .TRUE..
*> 1 .LE. ILOZ .LE. ILO; IHI .LE. IHIZ .LE. N.
*> \endverbatim
*>
*> \param[in,out] Z
*> \verbatim
*> Z is REAL array, dimension (LDZ,IHI)
*> If WANTZ is .FALSE., then Z is not referenced.
*> If WANTZ is .TRUE., then Z(ILO:IHI,ILOZ:IHIZ) is
*> replaced by Z(ILO:IHI,ILOZ:IHIZ)*U where U is the
*> orthogonal Schur factor of H(ILO:IHI,ILO:IHI).
*> (The output value of Z when INFO.GT.0 is given under
*> the description of INFO below.)
*> \endverbatim
*>
*> \param[in] LDZ
*> \verbatim
*> LDZ is INTEGER
*> The leading dimension of the array Z. if WANTZ is .TRUE.
*> then LDZ.GE.MAX(1,IHIZ). Otherwize, LDZ.GE.1.
*> \endverbatim
*>
*> \param[out] WORK
*> \verbatim
*> WORK is REAL array, dimension LWORK
*> On exit, if LWORK = -1, WORK(1) returns an estimate of
*> the optimal value for LWORK.
*> \endverbatim
*>
*> \param[in] LWORK
*> \verbatim
*> LWORK is INTEGER
*> The dimension of the array WORK. LWORK .GE. max(1,N)
*> is sufficient, but LWORK typically as large as 6*N may
*> be required for optimal performance. A workspace query
*> to determine the optimal workspace size is recommended.
*> \endverbatim
*> \verbatim
*> If LWORK = -1, then SLAQR4 does a workspace query.
*> In this case, SLAQR4 checks the input parameters and
*> estimates the optimal workspace size for the given
*> values of N, ILO and IHI. The estimate is returned
*> in WORK(1). No error message related to LWORK is
*> issued by XERBLA. Neither H nor Z are accessed.
*> \endverbatim
*>
*> \param[out] INFO
*> \verbatim
*> INFO is INTEGER
*> = 0: successful exit
*> .GT. 0: if INFO = i, SLAQR4 failed to compute all of
*> the eigenvalues. Elements 1:ilo-1 and i+1:n of WR
*> and WI contain those eigenvalues which have been
*> successfully computed. (Failures are rare.)
*> \endverbatim
*> \verbatim
*> If INFO .GT. 0 and WANT is .FALSE., then on exit,
*> the remaining unconverged eigenvalues are the eigen-
*> values of the upper Hessenberg matrix rows and
*> columns ILO through INFO of the final, output
*> value of H.
*> \endverbatim
*> \verbatim
*> If INFO .GT. 0 and WANTT is .TRUE., then on exit
*> \endverbatim
*> \verbatim
*> (*) (initial value of H)*U = U*(final value of H)
*> \endverbatim
*> \verbatim
*> where U is an orthogonal matrix. The final
*> value of H is upper Hessenberg and quasi-triangular
*> in rows and columns INFO+1 through IHI.
*> \endverbatim
*> \verbatim
*> If INFO .GT. 0 and WANTZ is .TRUE., then on exit
*> \endverbatim
*> \verbatim
*> (final value of Z(ILO:IHI,ILOZ:IHIZ)
*> = (initial value of Z(ILO:IHI,ILOZ:IHIZ)*U
*> \endverbatim
*> \verbatim
*> where U is the orthogonal matrix in (*) (regard-
*> less of the value of WANTT.)
*> \endverbatim
*> \verbatim
*> If INFO .GT. 0 and WANTZ is .FALSE., then Z is not
*> accessed.
*> \endverbatim
*>
*
* Authors
* =======
*
*> \author Univ. of Tennessee
*> \author Univ. of California Berkeley
*> \author Univ. of Colorado Denver
*> \author NAG Ltd.
*
*> \date November 2011
*
*> \ingroup realOTHERauxiliary
*
*
* Further Details
* ===============
*>\details \b Further \b Details
*> \verbatim
*>
*> Based on contributions by
*> Karen Braman and Ralph Byers, Department of Mathematics,
*> University of Kansas, USA
*>
*> References:
*> K. Braman, R. Byers and R. Mathias, The Multi-Shift QR
*> Algorithm Part I: Maintaining Well Focused Shifts, and Level 3
*> Performance, SIAM Journal of Matrix Analysis, volume 23, pages
*> 929--947, 2002.
*>
*> K. Braman, R. Byers and R. Mathias, The Multi-Shift QR
*> Algorithm Part II: Aggressive Early Deflation, SIAM Journal
*> of Matrix Analysis, volume 23, pages 948--973, 2002.
*>
*> \endverbatim
*>
* =====================================================================
SUBROUTINE SLAQR4( WANTT, WANTZ, N, ILO, IHI, H, LDH, WR, WI,
$ ILOZ, IHIZ, Z, LDZ, WORK, LWORK, INFO )
*
* -- LAPACK auxiliary 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 2011
*
* .. Scalar Arguments ..
INTEGER IHI, IHIZ, ILO, ILOZ, INFO, LDH, LDZ, LWORK, N
LOGICAL WANTT, WANTZ
* ..
* .. Array Arguments ..
REAL H( LDH, * ), WI( * ), WORK( * ), WR( * ),
$ Z( LDZ, * )
* ..
*
* ================================================================
*
* .. Parameters ..
*
* ==== Matrices of order NTINY or smaller must be processed by
* . SLAHQR because of insufficient subdiagonal scratch space.
* . (This is a hard limit.) ====
INTEGER NTINY
PARAMETER ( NTINY = 11 )
*
* ==== Exceptional deflation windows: try to cure rare
* . slow convergence by varying the size of the
* . deflation window after KEXNW iterations. ====
INTEGER KEXNW
PARAMETER ( KEXNW = 5 )
*
* ==== Exceptional shifts: try to cure rare slow convergence
* . with ad-hoc exceptional shifts every KEXSH iterations.
* . ====
INTEGER KEXSH
PARAMETER ( KEXSH = 6 )
*
* ==== The constants WILK1 and WILK2 are used to form the
* . exceptional shifts. ====
REAL WILK1, WILK2
PARAMETER ( WILK1 = 0.75e0, WILK2 = -0.4375e0 )
REAL ZERO, ONE
PARAMETER ( ZERO = 0.0e0, ONE = 1.0e0 )
* ..
* .. Local Scalars ..
REAL AA, BB, CC, CS, DD, SN, SS, SWAP
INTEGER I, INF, IT, ITMAX, K, KACC22, KBOT, KDU, KS,
$ KT, KTOP, KU, KV, KWH, KWTOP, KWV, LD, LS,
$ LWKOPT, NDEC, NDFL, NH, NHO, NIBBLE, NMIN, NS,
$ NSMAX, NSR, NVE, NW, NWMAX, NWR, NWUPBD
LOGICAL SORTED
CHARACTER JBCMPZ*2
* ..
* .. External Functions ..
INTEGER ILAENV
EXTERNAL ILAENV
* ..
* .. Local Arrays ..
REAL ZDUM( 1, 1 )
* ..
* .. External Subroutines ..
EXTERNAL SLACPY, SLAHQR, SLANV2, SLAQR2, SLAQR5
* ..
* .. Intrinsic Functions ..
INTRINSIC ABS, INT, MAX, MIN, MOD, REAL
* ..
* .. Executable Statements ..
INFO = 0
*
* ==== Quick return for N = 0: nothing to do. ====
*
IF( N.EQ.0 ) THEN
WORK( 1 ) = ONE
RETURN
END IF
*
IF( N.LE.NTINY ) THEN
*
* ==== Tiny matrices must use SLAHQR. ====
*
LWKOPT = 1
IF( LWORK.NE.-1 )
$ CALL SLAHQR( WANTT, WANTZ, N, ILO, IHI, H, LDH, WR, WI,
$ ILOZ, IHIZ, Z, LDZ, INFO )
ELSE
*
* ==== Use small bulge multi-shift QR with aggressive early
* . deflation on larger-than-tiny matrices. ====
*
* ==== Hope for the best. ====
*
INFO = 0
*
* ==== Set up job flags for ILAENV. ====
*
IF( WANTT ) THEN
JBCMPZ( 1: 1 ) = 'S'
ELSE
JBCMPZ( 1: 1 ) = 'E'
END IF
IF( WANTZ ) THEN
JBCMPZ( 2: 2 ) = 'V'
ELSE
JBCMPZ( 2: 2 ) = 'N'
END IF
*
* ==== NWR = recommended deflation window size. At this
* . point, N .GT. NTINY = 11, so there is enough
* . subdiagonal workspace for NWR.GE.2 as required.
* . (In fact, there is enough subdiagonal space for
* . NWR.GE.3.) ====
*
NWR = ILAENV( 13, 'SLAQR4', JBCMPZ, N, ILO, IHI, LWORK )
NWR = MAX( 2, NWR )
NWR = MIN( IHI-ILO+1, ( N-1 ) / 3, NWR )
*
* ==== NSR = recommended number of simultaneous shifts.
* . At this point N .GT. NTINY = 11, so there is at
* . enough subdiagonal workspace for NSR to be even
* . and greater than or equal to two as required. ====
*
NSR = ILAENV( 15, 'SLAQR4', JBCMPZ, N, ILO, IHI, LWORK )
NSR = MIN( NSR, ( N+6 ) / 9, IHI-ILO )
NSR = MAX( 2, NSR-MOD( NSR, 2 ) )
*
* ==== Estimate optimal workspace ====
*
* ==== Workspace query call to SLAQR2 ====
*
CALL SLAQR2( WANTT, WANTZ, N, ILO, IHI, NWR+1, H, LDH, ILOZ,
$ IHIZ, Z, LDZ, LS, LD, WR, WI, H, LDH, N, H, LDH,
$ N, H, LDH, WORK, -1 )
*
* ==== Optimal workspace = MAX(SLAQR5, SLAQR2) ====
*
LWKOPT = MAX( 3*NSR / 2, INT( WORK( 1 ) ) )
*
* ==== Quick return in case of workspace query. ====
*
IF( LWORK.EQ.-1 ) THEN
WORK( 1 ) = REAL( LWKOPT )
RETURN
END IF
*
* ==== SLAHQR/SLAQR0 crossover point ====
*
NMIN = ILAENV( 12, 'SLAQR4', JBCMPZ, N, ILO, IHI, LWORK )
NMIN = MAX( NTINY, NMIN )
*
* ==== Nibble crossover point ====
*
NIBBLE = ILAENV( 14, 'SLAQR4', JBCMPZ, N, ILO, IHI, LWORK )
NIBBLE = MAX( 0, NIBBLE )
*
* ==== Accumulate reflections during ttswp? Use block
* . 2-by-2 structure during matrix-matrix multiply? ====
*
KACC22 = ILAENV( 16, 'SLAQR4', JBCMPZ, N, ILO, IHI, LWORK )
KACC22 = MAX( 0, KACC22 )
KACC22 = MIN( 2, KACC22 )
*
* ==== NWMAX = the largest possible deflation window for
* . which there is sufficient workspace. ====
*
NWMAX = MIN( ( N-1 ) / 3, LWORK / 2 )
NW = NWMAX
*
* ==== NSMAX = the Largest number of simultaneous shifts
* . for which there is sufficient workspace. ====
*
NSMAX = MIN( ( N+6 ) / 9, 2*LWORK / 3 )
NSMAX = NSMAX - MOD( NSMAX, 2 )
*
* ==== NDFL: an iteration count restarted at deflation. ====
*
NDFL = 1
*
* ==== ITMAX = iteration limit ====
*
ITMAX = MAX( 30, 2*KEXSH )*MAX( 10, ( IHI-ILO+1 ) )
*
* ==== Last row and column in the active block ====
*
KBOT = IHI
*
* ==== Main Loop ====
*
DO 80 IT = 1, ITMAX
*
* ==== Done when KBOT falls below ILO ====
*
IF( KBOT.LT.ILO )
$ GO TO 90
*
* ==== Locate active block ====
*
DO 10 K = KBOT, ILO + 1, -1
IF( H( K, K-1 ).EQ.ZERO )
$ GO TO 20
10 CONTINUE
K = ILO
20 CONTINUE
KTOP = K
*
* ==== Select deflation window size:
* . Typical Case:
* . If possible and advisable, nibble the entire
* . active block. If not, use size MIN(NWR,NWMAX)
* . or MIN(NWR+1,NWMAX) depending upon which has
* . the smaller corresponding subdiagonal entry
* . (a heuristic).
* .
* . Exceptional Case:
* . If there have been no deflations in KEXNW or
* . more iterations, then vary the deflation window
* . size. At first, because, larger windows are,
* . in general, more powerful than smaller ones,
* . rapidly increase the window to the maximum possible.
* . Then, gradually reduce the window size. ====
*
NH = KBOT - KTOP + 1
NWUPBD = MIN( NH, NWMAX )
IF( NDFL.LT.KEXNW ) THEN
NW = MIN( NWUPBD, NWR )
ELSE
NW = MIN( NWUPBD, 2*NW )
END IF
IF( NW.LT.NWMAX ) THEN
IF( NW.GE.NH-1 ) THEN
NW = NH
ELSE
KWTOP = KBOT - NW + 1
IF( ABS( H( KWTOP, KWTOP-1 ) ).GT.
$ ABS( H( KWTOP-1, KWTOP-2 ) ) )NW = NW + 1
END IF
END IF
IF( NDFL.LT.KEXNW ) THEN
NDEC = -1
ELSE IF( NDEC.GE.0 .OR. NW.GE.NWUPBD ) THEN
NDEC = NDEC + 1
IF( NW-NDEC.LT.2 )
$ NDEC = 0
NW = NW - NDEC
END IF
*
* ==== Aggressive early deflation:
* . split workspace under the subdiagonal into
* . - an nw-by-nw work array V in the lower
* . left-hand-corner,
* . - an NW-by-at-least-NW-but-more-is-better
* . (NW-by-NHO) horizontal work array along
* . the bottom edge,
* . - an at-least-NW-but-more-is-better (NHV-by-NW)
* . vertical work array along the left-hand-edge.
* . ====
*
KV = N - NW + 1
KT = NW + 1
NHO = ( N-NW-1 ) - KT + 1
KWV = NW + 2
NVE = ( N-NW ) - KWV + 1
*
* ==== Aggressive early deflation ====
*
CALL SLAQR2( WANTT, WANTZ, N, KTOP, KBOT, NW, H, LDH, ILOZ,
$ IHIZ, Z, LDZ, LS, LD, WR, WI, H( KV, 1 ), LDH,
$ NHO, H( KV, KT ), LDH, NVE, H( KWV, 1 ), LDH,
$ WORK, LWORK )
*
* ==== Adjust KBOT accounting for new deflations. ====
*
KBOT = KBOT - LD
*
* ==== KS points to the shifts. ====
*
KS = KBOT - LS + 1
*
* ==== Skip an expensive QR sweep if there is a (partly
* . heuristic) reason to expect that many eigenvalues
* . will deflate without it. Here, the QR sweep is
* . skipped if many eigenvalues have just been deflated
* . or if the remaining active block is small.
*
IF( ( LD.EQ.0 ) .OR. ( ( 100*LD.LE.NW*NIBBLE ) .AND. ( KBOT-
$ KTOP+1.GT.MIN( NMIN, NWMAX ) ) ) ) THEN
*
* ==== NS = nominal number of simultaneous shifts.
* . This may be lowered (slightly) if SLAQR2
* . did not provide that many shifts. ====
*
NS = MIN( NSMAX, NSR, MAX( 2, KBOT-KTOP ) )
NS = NS - MOD( NS, 2 )
*
* ==== If there have been no deflations
* . in a multiple of KEXSH iterations,
* . then try exceptional shifts.
* . Otherwise use shifts provided by
* . SLAQR2 above or from the eigenvalues
* . of a trailing principal submatrix. ====
*
IF( MOD( NDFL, KEXSH ).EQ.0 ) THEN
KS = KBOT - NS + 1
DO 30 I = KBOT, MAX( KS+1, KTOP+2 ), -2
SS = ABS( H( I, I-1 ) ) + ABS( H( I-1, I-2 ) )
AA = WILK1*SS + H( I, I )
BB = SS
CC = WILK2*SS
DD = AA
CALL SLANV2( AA, BB, CC, DD, WR( I-1 ), WI( I-1 ),
$ WR( I ), WI( I ), CS, SN )
30 CONTINUE
IF( KS.EQ.KTOP ) THEN
WR( KS+1 ) = H( KS+1, KS+1 )
WI( KS+1 ) = ZERO
WR( KS ) = WR( KS+1 )
WI( KS ) = WI( KS+1 )
END IF
ELSE
*
* ==== Got NS/2 or fewer shifts? Use SLAHQR
* . on a trailing principal submatrix to
* . get more. (Since NS.LE.NSMAX.LE.(N+6)/9,
* . there is enough space below the subdiagonal
* . to fit an NS-by-NS scratch array.) ====
*
IF( KBOT-KS+1.LE.NS / 2 ) THEN
KS = KBOT - NS + 1
KT = N - NS + 1
CALL SLACPY( 'A', NS, NS, H( KS, KS ), LDH,
$ H( KT, 1 ), LDH )
CALL SLAHQR( .false., .false., NS, 1, NS,
$ H( KT, 1 ), LDH, WR( KS ), WI( KS ),
$ 1, 1, ZDUM, 1, INF )
KS = KS + INF
*
* ==== In case of a rare QR failure use
* . eigenvalues of the trailing 2-by-2
* . principal submatrix. ====
*
IF( KS.GE.KBOT ) THEN
AA = H( KBOT-1, KBOT-1 )
CC = H( KBOT, KBOT-1 )
BB = H( KBOT-1, KBOT )
DD = H( KBOT, KBOT )
CALL SLANV2( AA, BB, CC, DD, WR( KBOT-1 ),
$ WI( KBOT-1 ), WR( KBOT ),
$ WI( KBOT ), CS, SN )
KS = KBOT - 1
END IF
END IF
*
IF( KBOT-KS+1.GT.NS ) THEN
*
* ==== Sort the shifts (Helps a little)
* . Bubble sort keeps complex conjugate
* . pairs together. ====
*
SORTED = .false.
DO 50 K = KBOT, KS + 1, -1
IF( SORTED )
$ GO TO 60
SORTED = .true.
DO 40 I = KS, K - 1
IF( ABS( WR( I ) )+ABS( WI( I ) ).LT.
$ ABS( WR( I+1 ) )+ABS( WI( I+1 ) ) ) THEN
SORTED = .false.
*
SWAP = WR( I )
WR( I ) = WR( I+1 )
WR( I+1 ) = SWAP
*
SWAP = WI( I )
WI( I ) = WI( I+1 )
WI( I+1 ) = SWAP
END IF
40 CONTINUE
50 CONTINUE
60 CONTINUE
END IF
*
* ==== Shuffle shifts into pairs of real shifts
* . and pairs of complex conjugate shifts
* . assuming complex conjugate shifts are
* . already adjacent to one another. (Yes,
* . they are.) ====
*
DO 70 I = KBOT, KS + 2, -2
IF( WI( I ).NE.-WI( I-1 ) ) THEN
*
SWAP = WR( I )
WR( I ) = WR( I-1 )
WR( I-1 ) = WR( I-2 )
WR( I-2 ) = SWAP
*
SWAP = WI( I )
WI( I ) = WI( I-1 )
WI( I-1 ) = WI( I-2 )
WI( I-2 ) = SWAP
END IF
70 CONTINUE
END IF
*
* ==== If there are only two shifts and both are
* . real, then use only one. ====
*
IF( KBOT-KS+1.EQ.2 ) THEN
IF( WI( KBOT ).EQ.ZERO ) THEN
IF( ABS( WR( KBOT )-H( KBOT, KBOT ) ).LT.
$ ABS( WR( KBOT-1 )-H( KBOT, KBOT ) ) ) THEN
WR( KBOT-1 ) = WR( KBOT )
ELSE
WR( KBOT ) = WR( KBOT-1 )
END IF
END IF
END IF
*
* ==== Use up to NS of the the smallest magnatiude
* . shifts. If there aren't NS shifts available,
* . then use them all, possibly dropping one to
* . make the number of shifts even. ====
*
NS = MIN( NS, KBOT-KS+1 )
NS = NS - MOD( NS, 2 )
KS = KBOT - NS + 1
*
* ==== Small-bulge multi-shift QR sweep:
* . split workspace under the subdiagonal into
* . - a KDU-by-KDU work array U in the lower
* . left-hand-corner,
* . - a KDU-by-at-least-KDU-but-more-is-better
* . (KDU-by-NHo) horizontal work array WH along
* . the bottom edge,
* . - and an at-least-KDU-but-more-is-better-by-KDU
* . (NVE-by-KDU) vertical work WV arrow along
* . the left-hand-edge. ====
*
KDU = 3*NS - 3
KU = N - KDU + 1
KWH = KDU + 1
NHO = ( N-KDU+1-4 ) - ( KDU+1 ) + 1
KWV = KDU + 4
NVE = N - KDU - KWV + 1
*
* ==== Small-bulge multi-shift QR sweep ====
*
CALL SLAQR5( WANTT, WANTZ, KACC22, N, KTOP, KBOT, NS,
$ WR( KS ), WI( KS ), H, LDH, ILOZ, IHIZ, Z,
$ LDZ, WORK, 3, H( KU, 1 ), LDH, NVE,
$ H( KWV, 1 ), LDH, NHO, H( KU, KWH ), LDH )
END IF
*
* ==== Note progress (or the lack of it). ====
*
IF( LD.GT.0 ) THEN
NDFL = 1
ELSE
NDFL = NDFL + 1
END IF
*
* ==== End of main loop ====
80 CONTINUE
*
* ==== Iteration limit exceeded. Set INFO to show where
* . the problem occurred and exit. ====
*
INFO = KBOT
90 CONTINUE
END IF
*
* ==== Return the optimal value of LWORK. ====
*
WORK( 1 ) = REAL( LWKOPT )
*
* ==== End of SLAQR4 ====
*
END
|