forked from Reference-LAPACK/lapack
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdgesvd.f
3504 lines (3504 loc) · 133 KB
/
dgesvd.f
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
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
*> \brief <b> DGESVD computes the singular value decomposition (SVD) for GE matrices</b>
*
* =========== DOCUMENTATION ===========
*
* Online html documentation available at
* http://www.netlib.org/lapack/explore-html/
*
*> \htmlonly
*> Download DGESVD + dependencies
*> <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/dgesvd.f">
*> [TGZ]</a>
*> <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/dgesvd.f">
*> [ZIP]</a>
*> <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/dgesvd.f">
*> [TXT]</a>
*> \endhtmlonly
*
* Definition:
* ===========
*
* SUBROUTINE DGESVD( JOBU, JOBVT, M, N, A, LDA, S, U, LDU, VT, LDVT,
* WORK, LWORK, INFO )
*
* .. Scalar Arguments ..
* CHARACTER JOBU, JOBVT
* INTEGER INFO, LDA, LDU, LDVT, LWORK, M, N
* ..
* .. Array Arguments ..
* DOUBLE PRECISION A( LDA, * ), S( * ), U( LDU, * ),
* $ VT( LDVT, * ), WORK( * )
* ..
*
*
*> \par Purpose:
* =============
*>
*> \verbatim
*>
*> DGESVD computes the singular value decomposition (SVD) of a real
*> M-by-N matrix A, optionally computing the left and/or right singular
*> vectors. The SVD is written
*>
*> A = U * SIGMA * transpose(V)
*>
*> where SIGMA is an M-by-N matrix which is zero except for its
*> min(m,n) diagonal elements, U is an M-by-M orthogonal matrix, and
*> V is an N-by-N orthogonal matrix. The diagonal elements of SIGMA
*> are the singular values of A; they are real and non-negative, and
*> are returned in descending order. The first min(m,n) columns of
*> U and V are the left and right singular vectors of A.
*>
*> Note that the routine returns V**T, not V.
*> \endverbatim
*
* Arguments:
* ==========
*
*> \param[in] JOBU
*> \verbatim
*> JOBU is CHARACTER*1
*> Specifies options for computing all or part of the matrix U:
*> = 'A': all M columns of U are returned in array U:
*> = 'S': the first min(m,n) columns of U (the left singular
*> vectors) are returned in the array U;
*> = 'O': the first min(m,n) columns of U (the left singular
*> vectors) are overwritten on the array A;
*> = 'N': no columns of U (no left singular vectors) are
*> computed.
*> \endverbatim
*>
*> \param[in] JOBVT
*> \verbatim
*> JOBVT is CHARACTER*1
*> Specifies options for computing all or part of the matrix
*> V**T:
*> = 'A': all N rows of V**T are returned in the array VT;
*> = 'S': the first min(m,n) rows of V**T (the right singular
*> vectors) are returned in the array VT;
*> = 'O': the first min(m,n) rows of V**T (the right singular
*> vectors) are overwritten on the array A;
*> = 'N': no rows of V**T (no right singular vectors) are
*> computed.
*>
*> JOBVT and JOBU cannot both be 'O'.
*> \endverbatim
*>
*> \param[in] M
*> \verbatim
*> M is INTEGER
*> The number of rows of the input matrix A. M >= 0.
*> \endverbatim
*>
*> \param[in] N
*> \verbatim
*> N is INTEGER
*> The number of columns of the input matrix A. N >= 0.
*> \endverbatim
*>
*> \param[in,out] A
*> \verbatim
*> A is DOUBLE PRECISION array, dimension (LDA,N)
*> On entry, the M-by-N matrix A.
*> On exit,
*> if JOBU = 'O', A is overwritten with the first min(m,n)
*> columns of U (the left singular vectors,
*> stored columnwise);
*> if JOBVT = 'O', A is overwritten with the first min(m,n)
*> rows of V**T (the right singular vectors,
*> stored rowwise);
*> if JOBU .ne. 'O' and JOBVT .ne. 'O', the contents of A
*> are destroyed.
*> \endverbatim
*>
*> \param[in] LDA
*> \verbatim
*> LDA is INTEGER
*> The leading dimension of the array A. LDA >= max(1,M).
*> \endverbatim
*>
*> \param[out] S
*> \verbatim
*> S is DOUBLE PRECISION array, dimension (min(M,N))
*> The singular values of A, sorted so that S(i) >= S(i+1).
*> \endverbatim
*>
*> \param[out] U
*> \verbatim
*> U is DOUBLE PRECISION array, dimension (LDU,UCOL)
*> (LDU,M) if JOBU = 'A' or (LDU,min(M,N)) if JOBU = 'S'.
*> If JOBU = 'A', U contains the M-by-M orthogonal matrix U;
*> if JOBU = 'S', U contains the first min(m,n) columns of U
*> (the left singular vectors, stored columnwise);
*> if JOBU = 'N' or 'O', U is not referenced.
*> \endverbatim
*>
*> \param[in] LDU
*> \verbatim
*> LDU is INTEGER
*> The leading dimension of the array U. LDU >= 1; if
*> JOBU = 'S' or 'A', LDU >= M.
*> \endverbatim
*>
*> \param[out] VT
*> \verbatim
*> VT is DOUBLE PRECISION array, dimension (LDVT,N)
*> If JOBVT = 'A', VT contains the N-by-N orthogonal matrix
*> V**T;
*> if JOBVT = 'S', VT contains the first min(m,n) rows of
*> V**T (the right singular vectors, stored rowwise);
*> if JOBVT = 'N' or 'O', VT is not referenced.
*> \endverbatim
*>
*> \param[in] LDVT
*> \verbatim
*> LDVT is INTEGER
*> The leading dimension of the array VT. LDVT >= 1; if
*> JOBVT = 'A', LDVT >= N; if JOBVT = 'S', LDVT >= min(M,N).
*> \endverbatim
*>
*> \param[out] WORK
*> \verbatim
*> WORK is DOUBLE PRECISION array, dimension (MAX(1,LWORK))
*> On exit, if INFO = 0, WORK(1) returns the optimal LWORK;
*> if INFO > 0, WORK(2:MIN(M,N)) contains the unconverged
*> superdiagonal elements of an upper bidiagonal matrix B
*> whose diagonal is in S (not necessarily sorted). B
*> satisfies A = U * B * VT, so it has the same singular values
*> as A, and singular vectors related by U and VT.
*> \endverbatim
*>
*> \param[in] LWORK
*> \verbatim
*> LWORK is INTEGER
*> The dimension of the array WORK.
*> LWORK >= MAX(1,5*MIN(M,N)) for the paths (see comments inside code):
*> - PATH 1 (M much larger than N, JOBU='N')
*> - PATH 1t (N much larger than M, JOBVT='N')
*> LWORK >= MAX(1,3*MIN(M,N) + MAX(M,N),5*MIN(M,N)) for the other paths
*> For good performance, LWORK should generally be larger.
*>
*> If LWORK = -1, then a workspace query is assumed; the routine
*> only calculates the optimal size of the WORK array, returns
*> this value as the first entry of the WORK array, and no error
*> message related to LWORK is issued by XERBLA.
*> \endverbatim
*>
*> \param[out] INFO
*> \verbatim
*> INFO is INTEGER
*> = 0: successful exit.
*> < 0: if INFO = -i, the i-th argument had an illegal value.
*> > 0: if DBDSQR did not converge, INFO specifies how many
*> superdiagonals of an intermediate bidiagonal form B
*> did not converge to zero. See the description of WORK
*> above for details.
*> \endverbatim
*
* Authors:
* ========
*
*> \author Univ. of Tennessee
*> \author Univ. of California Berkeley
*> \author Univ. of Colorado Denver
*> \author NAG Ltd.
*
*> \date April 2012
*
*> \ingroup doubleGEsing
*
* =====================================================================
SUBROUTINE DGESVD( JOBU, JOBVT, M, N, A, LDA, S, U, LDU,
$ VT, LDVT, WORK, LWORK, INFO )
*
* -- LAPACK driver routine (version 3.7.0) --
* -- LAPACK is a software package provided by Univ. of Tennessee, --
* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
* April 2012
*
* .. Scalar Arguments ..
CHARACTER JOBU, JOBVT
INTEGER INFO, LDA, LDU, LDVT, LWORK, M, N
* ..
* .. Array Arguments ..
DOUBLE PRECISION A( LDA, * ), S( * ), U( LDU, * ),
$ VT( LDVT, * ), WORK( * )
* ..
*
* =====================================================================
*
* .. Parameters ..
DOUBLE PRECISION ZERO, ONE
PARAMETER ( ZERO = 0.0D0, ONE = 1.0D0 )
* ..
* .. Local Scalars ..
LOGICAL LQUERY, WNTUA, WNTUAS, WNTUN, WNTUO, WNTUS,
$ WNTVA, WNTVAS, WNTVN, WNTVO, WNTVS
INTEGER BDSPAC, BLK, CHUNK, I, IE, IERR, IR, ISCL,
$ ITAU, ITAUP, ITAUQ, IU, IWORK, LDWRKR, LDWRKU,
$ MAXWRK, MINMN, MINWRK, MNTHR, NCU, NCVT, NRU,
$ NRVT, WRKBL
INTEGER LWORK_DGEQRF, LWORK_DORGQR_N, LWORK_DORGQR_M,
$ LWORK_DGEBRD, LWORK_DORGBR_P, LWORK_DORGBR_Q,
$ LWORK_DGELQF, LWORK_DORGLQ_N, LWORK_DORGLQ_M
DOUBLE PRECISION ANRM, BIGNUM, EPS, SMLNUM
* ..
* .. Local Arrays ..
DOUBLE PRECISION DUM( 1 )
* ..
* .. External Subroutines ..
EXTERNAL DBDSQR, DGEBRD, DGELQF, DGEMM, DGEQRF, DLACPY,
$ DLASCL, DLASET, DORGBR, DORGLQ, DORGQR, DORMBR,
$ XERBLA
* ..
* .. External Functions ..
LOGICAL LSAME
INTEGER ILAENV
DOUBLE PRECISION DLAMCH, DLANGE
EXTERNAL LSAME, ILAENV, DLAMCH, DLANGE
* ..
* .. Intrinsic Functions ..
INTRINSIC MAX, MIN, SQRT
* ..
* .. Executable Statements ..
*
* Test the input arguments
*
INFO = 0
MINMN = MIN( M, N )
WNTUA = LSAME( JOBU, 'A' )
WNTUS = LSAME( JOBU, 'S' )
WNTUAS = WNTUA .OR. WNTUS
WNTUO = LSAME( JOBU, 'O' )
WNTUN = LSAME( JOBU, 'N' )
WNTVA = LSAME( JOBVT, 'A' )
WNTVS = LSAME( JOBVT, 'S' )
WNTVAS = WNTVA .OR. WNTVS
WNTVO = LSAME( JOBVT, 'O' )
WNTVN = LSAME( JOBVT, 'N' )
LQUERY = ( LWORK.EQ.-1 )
*
IF( .NOT.( WNTUA .OR. WNTUS .OR. WNTUO .OR. WNTUN ) ) THEN
INFO = -1
ELSE IF( .NOT.( WNTVA .OR. WNTVS .OR. WNTVO .OR. WNTVN ) .OR.
$ ( WNTVO .AND. WNTUO ) ) THEN
INFO = -2
ELSE IF( M.LT.0 ) THEN
INFO = -3
ELSE IF( N.LT.0 ) THEN
INFO = -4
ELSE IF( LDA.LT.MAX( 1, M ) ) THEN
INFO = -6
ELSE IF( LDU.LT.1 .OR. ( WNTUAS .AND. LDU.LT.M ) ) THEN
INFO = -9
ELSE IF( LDVT.LT.1 .OR. ( WNTVA .AND. LDVT.LT.N ) .OR.
$ ( WNTVS .AND. LDVT.LT.MINMN ) ) THEN
INFO = -11
END IF
*
* Compute workspace
* (Note: Comments in the code beginning "Workspace:" describe the
* minimal amount of workspace needed at that point in the code,
* as well as the preferred amount for good performance.
* NB refers to the optimal block size for the immediately
* following subroutine, as returned by ILAENV.)
*
IF( INFO.EQ.0 ) THEN
MINWRK = 1
MAXWRK = 1
IF( M.GE.N .AND. MINMN.GT.0 ) THEN
*
* Compute space needed for DBDSQR
*
MNTHR = ILAENV( 6, 'DGESVD', JOBU // JOBVT, M, N, 0, 0 )
BDSPAC = 5*N
* Compute space needed for DGEQRF
CALL DGEQRF( M, N, A, LDA, DUM(1), DUM(1), -1, IERR )
LWORK_DGEQRF = INT( DUM(1) )
* Compute space needed for DORGQR
CALL DORGQR( M, N, N, A, LDA, DUM(1), DUM(1), -1, IERR )
LWORK_DORGQR_N = INT( DUM(1) )
CALL DORGQR( M, M, N, A, LDA, DUM(1), DUM(1), -1, IERR )
LWORK_DORGQR_M = INT( DUM(1) )
* Compute space needed for DGEBRD
CALL DGEBRD( N, N, A, LDA, S, DUM(1), DUM(1),
$ DUM(1), DUM(1), -1, IERR )
LWORK_DGEBRD = INT( DUM(1) )
* Compute space needed for DORGBR P
CALL DORGBR( 'P', N, N, N, A, LDA, DUM(1),
$ DUM(1), -1, IERR )
LWORK_DORGBR_P = INT( DUM(1) )
* Compute space needed for DORGBR Q
CALL DORGBR( 'Q', N, N, N, A, LDA, DUM(1),
$ DUM(1), -1, IERR )
LWORK_DORGBR_Q = INT( DUM(1) )
*
IF( M.GE.MNTHR ) THEN
IF( WNTUN ) THEN
*
* Path 1 (M much larger than N, JOBU='N')
*
MAXWRK = N + LWORK_DGEQRF
MAXWRK = MAX( MAXWRK, 3*N + LWORK_DGEBRD )
IF( WNTVO .OR. WNTVAS )
$ MAXWRK = MAX( MAXWRK, 3*N + LWORK_DORGBR_P )
MAXWRK = MAX( MAXWRK, BDSPAC )
MINWRK = MAX( 4*N, BDSPAC )
ELSE IF( WNTUO .AND. WNTVN ) THEN
*
* Path 2 (M much larger than N, JOBU='O', JOBVT='N')
*
WRKBL = N + LWORK_DGEQRF
WRKBL = MAX( WRKBL, N + LWORK_DORGQR_N )
WRKBL = MAX( WRKBL, 3*N + LWORK_DGEBRD )
WRKBL = MAX( WRKBL, 3*N + LWORK_DORGBR_Q )
WRKBL = MAX( WRKBL, BDSPAC )
MAXWRK = MAX( N*N + WRKBL, N*N + M*N + N )
MINWRK = MAX( 3*N + M, BDSPAC )
ELSE IF( WNTUO .AND. WNTVAS ) THEN
*
* Path 3 (M much larger than N, JOBU='O', JOBVT='S' or
* 'A')
*
WRKBL = N + LWORK_DGEQRF
WRKBL = MAX( WRKBL, N + LWORK_DORGQR_N )
WRKBL = MAX( WRKBL, 3*N + LWORK_DGEBRD )
WRKBL = MAX( WRKBL, 3*N + LWORK_DORGBR_Q )
WRKBL = MAX( WRKBL, 3*N + LWORK_DORGBR_P )
WRKBL = MAX( WRKBL, BDSPAC )
MAXWRK = MAX( N*N + WRKBL, N*N + M*N + N )
MINWRK = MAX( 3*N + M, BDSPAC )
ELSE IF( WNTUS .AND. WNTVN ) THEN
*
* Path 4 (M much larger than N, JOBU='S', JOBVT='N')
*
WRKBL = N + LWORK_DGEQRF
WRKBL = MAX( WRKBL, N + LWORK_DORGQR_N )
WRKBL = MAX( WRKBL, 3*N + LWORK_DGEBRD )
WRKBL = MAX( WRKBL, 3*N + LWORK_DORGBR_Q )
WRKBL = MAX( WRKBL, BDSPAC )
MAXWRK = N*N + WRKBL
MINWRK = MAX( 3*N + M, BDSPAC )
ELSE IF( WNTUS .AND. WNTVO ) THEN
*
* Path 5 (M much larger than N, JOBU='S', JOBVT='O')
*
WRKBL = N + LWORK_DGEQRF
WRKBL = MAX( WRKBL, N + LWORK_DORGQR_N )
WRKBL = MAX( WRKBL, 3*N + LWORK_DGEBRD )
WRKBL = MAX( WRKBL, 3*N + LWORK_DORGBR_Q )
WRKBL = MAX( WRKBL, 3*N + LWORK_DORGBR_P )
WRKBL = MAX( WRKBL, BDSPAC )
MAXWRK = 2*N*N + WRKBL
MINWRK = MAX( 3*N + M, BDSPAC )
ELSE IF( WNTUS .AND. WNTVAS ) THEN
*
* Path 6 (M much larger than N, JOBU='S', JOBVT='S' or
* 'A')
*
WRKBL = N + LWORK_DGEQRF
WRKBL = MAX( WRKBL, N + LWORK_DORGQR_N )
WRKBL = MAX( WRKBL, 3*N + LWORK_DGEBRD )
WRKBL = MAX( WRKBL, 3*N + LWORK_DORGBR_Q )
WRKBL = MAX( WRKBL, 3*N + LWORK_DORGBR_P )
WRKBL = MAX( WRKBL, BDSPAC )
MAXWRK = N*N + WRKBL
MINWRK = MAX( 3*N + M, BDSPAC )
ELSE IF( WNTUA .AND. WNTVN ) THEN
*
* Path 7 (M much larger than N, JOBU='A', JOBVT='N')
*
WRKBL = N + LWORK_DGEQRF
WRKBL = MAX( WRKBL, N + LWORK_DORGQR_M )
WRKBL = MAX( WRKBL, 3*N + LWORK_DGEBRD )
WRKBL = MAX( WRKBL, 3*N + LWORK_DORGBR_Q )
WRKBL = MAX( WRKBL, BDSPAC )
MAXWRK = N*N + WRKBL
MINWRK = MAX( 3*N + M, BDSPAC )
ELSE IF( WNTUA .AND. WNTVO ) THEN
*
* Path 8 (M much larger than N, JOBU='A', JOBVT='O')
*
WRKBL = N + LWORK_DGEQRF
WRKBL = MAX( WRKBL, N + LWORK_DORGQR_M )
WRKBL = MAX( WRKBL, 3*N + LWORK_DGEBRD )
WRKBL = MAX( WRKBL, 3*N + LWORK_DORGBR_Q )
WRKBL = MAX( WRKBL, 3*N + LWORK_DORGBR_P )
WRKBL = MAX( WRKBL, BDSPAC )
MAXWRK = 2*N*N + WRKBL
MINWRK = MAX( 3*N + M, BDSPAC )
ELSE IF( WNTUA .AND. WNTVAS ) THEN
*
* Path 9 (M much larger than N, JOBU='A', JOBVT='S' or
* 'A')
*
WRKBL = N + LWORK_DGEQRF
WRKBL = MAX( WRKBL, N + LWORK_DORGQR_M )
WRKBL = MAX( WRKBL, 3*N + LWORK_DGEBRD )
WRKBL = MAX( WRKBL, 3*N + LWORK_DORGBR_Q )
WRKBL = MAX( WRKBL, 3*N + LWORK_DORGBR_P )
WRKBL = MAX( WRKBL, BDSPAC )
MAXWRK = N*N + WRKBL
MINWRK = MAX( 3*N + M, BDSPAC )
END IF
ELSE
*
* Path 10 (M at least N, but not much larger)
*
CALL DGEBRD( M, N, A, LDA, S, DUM(1), DUM(1),
$ DUM(1), DUM(1), -1, IERR )
LWORK_DGEBRD = INT( DUM(1) )
MAXWRK = 3*N + LWORK_DGEBRD
IF( WNTUS .OR. WNTUO ) THEN
CALL DORGBR( 'Q', M, N, N, A, LDA, DUM(1),
$ DUM(1), -1, IERR )
LWORK_DORGBR_Q = INT( DUM(1) )
MAXWRK = MAX( MAXWRK, 3*N + LWORK_DORGBR_Q )
END IF
IF( WNTUA ) THEN
CALL DORGBR( 'Q', M, M, N, A, LDA, DUM(1),
$ DUM(1), -1, IERR )
LWORK_DORGBR_Q = INT( DUM(1) )
MAXWRK = MAX( MAXWRK, 3*N + LWORK_DORGBR_Q )
END IF
IF( .NOT.WNTVN ) THEN
MAXWRK = MAX( MAXWRK, 3*N + LWORK_DORGBR_P )
END IF
MAXWRK = MAX( MAXWRK, BDSPAC )
MINWRK = MAX( 3*N + M, BDSPAC )
END IF
ELSE IF( MINMN.GT.0 ) THEN
*
* Compute space needed for DBDSQR
*
MNTHR = ILAENV( 6, 'DGESVD', JOBU // JOBVT, M, N, 0, 0 )
BDSPAC = 5*M
* Compute space needed for DGELQF
CALL DGELQF( M, N, A, LDA, DUM(1), DUM(1), -1, IERR )
LWORK_DGELQF = INT( DUM(1) )
* Compute space needed for DORGLQ
CALL DORGLQ( N, N, M, DUM(1), N, DUM(1), DUM(1), -1, IERR )
LWORK_DORGLQ_N = INT( DUM(1) )
CALL DORGLQ( M, N, M, A, LDA, DUM(1), DUM(1), -1, IERR )
LWORK_DORGLQ_M = INT( DUM(1) )
* Compute space needed for DGEBRD
CALL DGEBRD( M, M, A, LDA, S, DUM(1), DUM(1),
$ DUM(1), DUM(1), -1, IERR )
LWORK_DGEBRD = INT( DUM(1) )
* Compute space needed for DORGBR P
CALL DORGBR( 'P', M, M, M, A, N, DUM(1),
$ DUM(1), -1, IERR )
LWORK_DORGBR_P = INT( DUM(1) )
* Compute space needed for DORGBR Q
CALL DORGBR( 'Q', M, M, M, A, N, DUM(1),
$ DUM(1), -1, IERR )
LWORK_DORGBR_Q = INT( DUM(1) )
IF( N.GE.MNTHR ) THEN
IF( WNTVN ) THEN
*
* Path 1t(N much larger than M, JOBVT='N')
*
MAXWRK = M + LWORK_DGELQF
MAXWRK = MAX( MAXWRK, 3*M + LWORK_DGEBRD )
IF( WNTUO .OR. WNTUAS )
$ MAXWRK = MAX( MAXWRK, 3*M + LWORK_DORGBR_Q )
MAXWRK = MAX( MAXWRK, BDSPAC )
MINWRK = MAX( 4*M, BDSPAC )
ELSE IF( WNTVO .AND. WNTUN ) THEN
*
* Path 2t(N much larger than M, JOBU='N', JOBVT='O')
*
WRKBL = M + LWORK_DGELQF
WRKBL = MAX( WRKBL, M + LWORK_DORGLQ_M )
WRKBL = MAX( WRKBL, 3*M + LWORK_DGEBRD )
WRKBL = MAX( WRKBL, 3*M + LWORK_DORGBR_P )
WRKBL = MAX( WRKBL, BDSPAC )
MAXWRK = MAX( M*M + WRKBL, M*M + M*N + M )
MINWRK = MAX( 3*M + N, BDSPAC )
ELSE IF( WNTVO .AND. WNTUAS ) THEN
*
* Path 3t(N much larger than M, JOBU='S' or 'A',
* JOBVT='O')
*
WRKBL = M + LWORK_DGELQF
WRKBL = MAX( WRKBL, M + LWORK_DORGLQ_M )
WRKBL = MAX( WRKBL, 3*M + LWORK_DGEBRD )
WRKBL = MAX( WRKBL, 3*M + LWORK_DORGBR_P )
WRKBL = MAX( WRKBL, 3*M + LWORK_DORGBR_Q )
WRKBL = MAX( WRKBL, BDSPAC )
MAXWRK = MAX( M*M + WRKBL, M*M + M*N + M )
MINWRK = MAX( 3*M + N, BDSPAC )
ELSE IF( WNTVS .AND. WNTUN ) THEN
*
* Path 4t(N much larger than M, JOBU='N', JOBVT='S')
*
WRKBL = M + LWORK_DGELQF
WRKBL = MAX( WRKBL, M + LWORK_DORGLQ_M )
WRKBL = MAX( WRKBL, 3*M + LWORK_DGEBRD )
WRKBL = MAX( WRKBL, 3*M + LWORK_DORGBR_P )
WRKBL = MAX( WRKBL, BDSPAC )
MAXWRK = M*M + WRKBL
MINWRK = MAX( 3*M + N, BDSPAC )
ELSE IF( WNTVS .AND. WNTUO ) THEN
*
* Path 5t(N much larger than M, JOBU='O', JOBVT='S')
*
WRKBL = M + LWORK_DGELQF
WRKBL = MAX( WRKBL, M + LWORK_DORGLQ_M )
WRKBL = MAX( WRKBL, 3*M + LWORK_DGEBRD )
WRKBL = MAX( WRKBL, 3*M + LWORK_DORGBR_P )
WRKBL = MAX( WRKBL, 3*M + LWORK_DORGBR_Q )
WRKBL = MAX( WRKBL, BDSPAC )
MAXWRK = 2*M*M + WRKBL
MINWRK = MAX( 3*M + N, BDSPAC )
ELSE IF( WNTVS .AND. WNTUAS ) THEN
*
* Path 6t(N much larger than M, JOBU='S' or 'A',
* JOBVT='S')
*
WRKBL = M + LWORK_DGELQF
WRKBL = MAX( WRKBL, M + LWORK_DORGLQ_M )
WRKBL = MAX( WRKBL, 3*M + LWORK_DGEBRD )
WRKBL = MAX( WRKBL, 3*M + LWORK_DORGBR_P )
WRKBL = MAX( WRKBL, 3*M + LWORK_DORGBR_Q )
WRKBL = MAX( WRKBL, BDSPAC )
MAXWRK = M*M + WRKBL
MINWRK = MAX( 3*M + N, BDSPAC )
ELSE IF( WNTVA .AND. WNTUN ) THEN
*
* Path 7t(N much larger than M, JOBU='N', JOBVT='A')
*
WRKBL = M + LWORK_DGELQF
WRKBL = MAX( WRKBL, M + LWORK_DORGLQ_N )
WRKBL = MAX( WRKBL, 3*M + LWORK_DGEBRD )
WRKBL = MAX( WRKBL, 3*M + LWORK_DORGBR_P )
WRKBL = MAX( WRKBL, BDSPAC )
MAXWRK = M*M + WRKBL
MINWRK = MAX( 3*M + N, BDSPAC )
ELSE IF( WNTVA .AND. WNTUO ) THEN
*
* Path 8t(N much larger than M, JOBU='O', JOBVT='A')
*
WRKBL = M + LWORK_DGELQF
WRKBL = MAX( WRKBL, M + LWORK_DORGLQ_N )
WRKBL = MAX( WRKBL, 3*M + LWORK_DGEBRD )
WRKBL = MAX( WRKBL, 3*M + LWORK_DORGBR_P )
WRKBL = MAX( WRKBL, 3*M + LWORK_DORGBR_Q )
WRKBL = MAX( WRKBL, BDSPAC )
MAXWRK = 2*M*M + WRKBL
MINWRK = MAX( 3*M + N, BDSPAC )
ELSE IF( WNTVA .AND. WNTUAS ) THEN
*
* Path 9t(N much larger than M, JOBU='S' or 'A',
* JOBVT='A')
*
WRKBL = M + LWORK_DGELQF
WRKBL = MAX( WRKBL, M + LWORK_DORGLQ_N )
WRKBL = MAX( WRKBL, 3*M + LWORK_DGEBRD )
WRKBL = MAX( WRKBL, 3*M + LWORK_DORGBR_P )
WRKBL = MAX( WRKBL, 3*M + LWORK_DORGBR_Q )
WRKBL = MAX( WRKBL, BDSPAC )
MAXWRK = M*M + WRKBL
MINWRK = MAX( 3*M + N, BDSPAC )
END IF
ELSE
*
* Path 10t(N greater than M, but not much larger)
*
CALL DGEBRD( M, N, A, LDA, S, DUM(1), DUM(1),
$ DUM(1), DUM(1), -1, IERR )
LWORK_DGEBRD = INT( DUM(1) )
MAXWRK = 3*M + LWORK_DGEBRD
IF( WNTVS .OR. WNTVO ) THEN
* Compute space needed for DORGBR P
CALL DORGBR( 'P', M, N, M, A, N, DUM(1),
$ DUM(1), -1, IERR )
LWORK_DORGBR_P = INT( DUM(1) )
MAXWRK = MAX( MAXWRK, 3*M + LWORK_DORGBR_P )
END IF
IF( WNTVA ) THEN
CALL DORGBR( 'P', N, N, M, A, N, DUM(1),
$ DUM(1), -1, IERR )
LWORK_DORGBR_P = INT( DUM(1) )
MAXWRK = MAX( MAXWRK, 3*M + LWORK_DORGBR_P )
END IF
IF( .NOT.WNTUN ) THEN
MAXWRK = MAX( MAXWRK, 3*M + LWORK_DORGBR_Q )
END IF
MAXWRK = MAX( MAXWRK, BDSPAC )
MINWRK = MAX( 3*M + N, BDSPAC )
END IF
END IF
MAXWRK = MAX( MAXWRK, MINWRK )
WORK( 1 ) = MAXWRK
*
IF( LWORK.LT.MINWRK .AND. .NOT.LQUERY ) THEN
INFO = -13
END IF
END IF
*
IF( INFO.NE.0 ) THEN
CALL XERBLA( 'DGESVD', -INFO )
RETURN
ELSE IF( LQUERY ) THEN
RETURN
END IF
*
* Quick return if possible
*
IF( M.EQ.0 .OR. N.EQ.0 ) THEN
RETURN
END IF
*
* Get machine constants
*
EPS = DLAMCH( 'P' )
SMLNUM = SQRT( DLAMCH( 'S' ) ) / EPS
BIGNUM = ONE / SMLNUM
*
* Scale A if max element outside range [SMLNUM,BIGNUM]
*
ANRM = DLANGE( 'M', M, N, A, LDA, DUM )
ISCL = 0
IF( ANRM.GT.ZERO .AND. ANRM.LT.SMLNUM ) THEN
ISCL = 1
CALL DLASCL( 'G', 0, 0, ANRM, SMLNUM, M, N, A, LDA, IERR )
ELSE IF( ANRM.GT.BIGNUM ) THEN
ISCL = 1
CALL DLASCL( 'G', 0, 0, ANRM, BIGNUM, M, N, A, LDA, IERR )
END IF
*
IF( M.GE.N ) THEN
*
* A has at least as many rows as columns. If A has sufficiently
* more rows than columns, first reduce using the QR
* decomposition (if sufficient workspace available)
*
IF( M.GE.MNTHR ) THEN
*
IF( WNTUN ) THEN
*
* Path 1 (M much larger than N, JOBU='N')
* No left singular vectors to be computed
*
ITAU = 1
IWORK = ITAU + N
*
* Compute A=Q*R
* (Workspace: need 2*N, prefer N + N*NB)
*
CALL DGEQRF( M, N, A, LDA, WORK( ITAU ), WORK( IWORK ),
$ LWORK-IWORK+1, IERR )
*
* Zero out below R
*
IF( N .GT. 1 ) THEN
CALL DLASET( 'L', N-1, N-1, ZERO, ZERO, A( 2, 1 ),
$ LDA )
END IF
IE = 1
ITAUQ = IE + N
ITAUP = ITAUQ + N
IWORK = ITAUP + N
*
* Bidiagonalize R in A
* (Workspace: need 4*N, prefer 3*N + 2*N*NB)
*
CALL DGEBRD( N, N, A, LDA, S, WORK( IE ), WORK( ITAUQ ),
$ WORK( ITAUP ), WORK( IWORK ), LWORK-IWORK+1,
$ IERR )
NCVT = 0
IF( WNTVO .OR. WNTVAS ) THEN
*
* If right singular vectors desired, generate P'.
* (Workspace: need 4*N-1, prefer 3*N + (N-1)*NB)
*
CALL DORGBR( 'P', N, N, N, A, LDA, WORK( ITAUP ),
$ WORK( IWORK ), LWORK-IWORK+1, IERR )
NCVT = N
END IF
IWORK = IE + N
*
* Perform bidiagonal QR iteration, computing right
* singular vectors of A in A if desired
* (Workspace: need BDSPAC)
*
CALL DBDSQR( 'U', N, NCVT, 0, 0, S, WORK( IE ), A, LDA,
$ DUM, 1, DUM, 1, WORK( IWORK ), INFO )
*
* If right singular vectors desired in VT, copy them there
*
IF( WNTVAS )
$ CALL DLACPY( 'F', N, N, A, LDA, VT, LDVT )
*
ELSE IF( WNTUO .AND. WNTVN ) THEN
*
* Path 2 (M much larger than N, JOBU='O', JOBVT='N')
* N left singular vectors to be overwritten on A and
* no right singular vectors to be computed
*
IF( LWORK.GE.N*N+MAX( 4*N, BDSPAC ) ) THEN
*
* Sufficient workspace for a fast algorithm
*
IR = 1
IF( LWORK.GE.MAX( WRKBL, LDA*N + N ) + LDA*N ) THEN
*
* WORK(IU) is LDA by N, WORK(IR) is LDA by N
*
LDWRKU = LDA
LDWRKR = LDA
ELSE IF( LWORK.GE.MAX( WRKBL, LDA*N + N ) + N*N ) THEN
*
* WORK(IU) is LDA by N, WORK(IR) is N by N
*
LDWRKU = LDA
LDWRKR = N
ELSE
*
* WORK(IU) is LDWRKU by N, WORK(IR) is N by N
*
LDWRKU = ( LWORK-N*N-N ) / N
LDWRKR = N
END IF
ITAU = IR + LDWRKR*N
IWORK = ITAU + N
*
* Compute A=Q*R
* (Workspace: need N*N + 2*N, prefer N*N + N + N*NB)
*
CALL DGEQRF( M, N, A, LDA, WORK( ITAU ),
$ WORK( IWORK ), LWORK-IWORK+1, IERR )
*
* Copy R to WORK(IR) and zero out below it
*
CALL DLACPY( 'U', N, N, A, LDA, WORK( IR ), LDWRKR )
CALL DLASET( 'L', N-1, N-1, ZERO, ZERO, WORK( IR+1 ),
$ LDWRKR )
*
* Generate Q in A
* (Workspace: need N*N + 2*N, prefer N*N + N + N*NB)
*
CALL DORGQR( M, N, N, A, LDA, WORK( ITAU ),
$ WORK( IWORK ), LWORK-IWORK+1, IERR )
IE = ITAU
ITAUQ = IE + N
ITAUP = ITAUQ + N
IWORK = ITAUP + N
*
* Bidiagonalize R in WORK(IR)
* (Workspace: need N*N + 4*N, prefer N*N + 3*N + 2*N*NB)
*
CALL DGEBRD( N, N, WORK( IR ), LDWRKR, S, WORK( IE ),
$ WORK( ITAUQ ), WORK( ITAUP ),
$ WORK( IWORK ), LWORK-IWORK+1, IERR )
*
* Generate left vectors bidiagonalizing R
* (Workspace: need N*N + 4*N, prefer N*N + 3*N + N*NB)
*
CALL DORGBR( 'Q', N, N, N, WORK( IR ), LDWRKR,
$ WORK( ITAUQ ), WORK( IWORK ),
$ LWORK-IWORK+1, IERR )
IWORK = IE + N
*
* Perform bidiagonal QR iteration, computing left
* singular vectors of R in WORK(IR)
* (Workspace: need N*N + BDSPAC)
*
CALL DBDSQR( 'U', N, 0, N, 0, S, WORK( IE ), DUM, 1,
$ WORK( IR ), LDWRKR, DUM, 1,
$ WORK( IWORK ), INFO )
IU = IE + N
*
* Multiply Q in A by left singular vectors of R in
* WORK(IR), storing result in WORK(IU) and copying to A
* (Workspace: need N*N + 2*N, prefer N*N + M*N + N)
*
DO 10 I = 1, M, LDWRKU
CHUNK = MIN( M-I+1, LDWRKU )
CALL DGEMM( 'N', 'N', CHUNK, N, N, ONE, A( I, 1 ),
$ LDA, WORK( IR ), LDWRKR, ZERO,
$ WORK( IU ), LDWRKU )
CALL DLACPY( 'F', CHUNK, N, WORK( IU ), LDWRKU,
$ A( I, 1 ), LDA )
10 CONTINUE
*
ELSE
*
* Insufficient workspace for a fast algorithm
*
IE = 1
ITAUQ = IE + N
ITAUP = ITAUQ + N
IWORK = ITAUP + N
*
* Bidiagonalize A
* (Workspace: need 3*N + M, prefer 3*N + (M + N)*NB)
*
CALL DGEBRD( M, N, A, LDA, S, WORK( IE ),
$ WORK( ITAUQ ), WORK( ITAUP ),
$ WORK( IWORK ), LWORK-IWORK+1, IERR )
*
* Generate left vectors bidiagonalizing A
* (Workspace: need 4*N, prefer 3*N + N*NB)
*
CALL DORGBR( 'Q', M, N, N, A, LDA, WORK( ITAUQ ),
$ WORK( IWORK ), LWORK-IWORK+1, IERR )
IWORK = IE + N
*
* Perform bidiagonal QR iteration, computing left
* singular vectors of A in A
* (Workspace: need BDSPAC)
*
CALL DBDSQR( 'U', N, 0, M, 0, S, WORK( IE ), DUM, 1,
$ A, LDA, DUM, 1, WORK( IWORK ), INFO )
*
END IF
*
ELSE IF( WNTUO .AND. WNTVAS ) THEN
*
* Path 3 (M much larger than N, JOBU='O', JOBVT='S' or 'A')
* N left singular vectors to be overwritten on A and
* N right singular vectors to be computed in VT
*
IF( LWORK.GE.N*N+MAX( 4*N, BDSPAC ) ) THEN
*
* Sufficient workspace for a fast algorithm
*
IR = 1
IF( LWORK.GE.MAX( WRKBL, LDA*N + N ) + LDA*N ) THEN
*
* WORK(IU) is LDA by N and WORK(IR) is LDA by N
*
LDWRKU = LDA
LDWRKR = LDA
ELSE IF( LWORK.GE.MAX( WRKBL, LDA*N + N ) + N*N ) THEN
*
* WORK(IU) is LDA by N and WORK(IR) is N by N
*
LDWRKU = LDA
LDWRKR = N
ELSE
*
* WORK(IU) is LDWRKU by N and WORK(IR) is N by N
*
LDWRKU = ( LWORK-N*N-N ) / N
LDWRKR = N
END IF
ITAU = IR + LDWRKR*N
IWORK = ITAU + N
*
* Compute A=Q*R
* (Workspace: need N*N + 2*N, prefer N*N + N + N*NB)
*
CALL DGEQRF( M, N, A, LDA, WORK( ITAU ),
$ WORK( IWORK ), LWORK-IWORK+1, IERR )
*
* Copy R to VT, zeroing out below it
*
CALL DLACPY( 'U', N, N, A, LDA, VT, LDVT )
IF( N.GT.1 )
$ CALL DLASET( 'L', N-1, N-1, ZERO, ZERO,
$ VT( 2, 1 ), LDVT )
*
* Generate Q in A
* (Workspace: need N*N + 2*N, prefer N*N + N + N*NB)
*
CALL DORGQR( M, N, N, A, LDA, WORK( ITAU ),
$ WORK( IWORK ), LWORK-IWORK+1, IERR )
IE = ITAU
ITAUQ = IE + N
ITAUP = ITAUQ + N
IWORK = ITAUP + N
*
* Bidiagonalize R in VT, copying result to WORK(IR)
* (Workspace: need N*N + 4*N, prefer N*N + 3*N + 2*N*NB)
*
CALL DGEBRD( N, N, VT, LDVT, S, WORK( IE ),
$ WORK( ITAUQ ), WORK( ITAUP ),
$ WORK( IWORK ), LWORK-IWORK+1, IERR )
CALL DLACPY( 'L', N, N, VT, LDVT, WORK( IR ), LDWRKR )
*
* Generate left vectors bidiagonalizing R in WORK(IR)
* (Workspace: need N*N + 4*N, prefer N*N + 3*N + N*NB)
*
CALL DORGBR( 'Q', N, N, N, WORK( IR ), LDWRKR,
$ WORK( ITAUQ ), WORK( IWORK ),
$ LWORK-IWORK+1, IERR )
*
* Generate right vectors bidiagonalizing R in VT
* (Workspace: need N*N + 4*N-1, prefer N*N + 3*N + (N-1)*NB)
*
CALL DORGBR( 'P', N, N, N, VT, LDVT, WORK( ITAUP ),
$ WORK( IWORK ), LWORK-IWORK+1, IERR )
IWORK = IE + N
*
* Perform bidiagonal QR iteration, computing left
* singular vectors of R in WORK(IR) and computing right
* singular vectors of R in VT
* (Workspace: need N*N + BDSPAC)
*
CALL DBDSQR( 'U', N, N, N, 0, S, WORK( IE ), VT, LDVT,
$ WORK( IR ), LDWRKR, DUM, 1,
$ WORK( IWORK ), INFO )
IU = IE + N
*
* Multiply Q in A by left singular vectors of R in
* WORK(IR), storing result in WORK(IU) and copying to A
* (Workspace: need N*N + 2*N, prefer N*N + M*N + N)
*
DO 20 I = 1, M, LDWRKU
CHUNK = MIN( M-I+1, LDWRKU )
CALL DGEMM( 'N', 'N', CHUNK, N, N, ONE, A( I, 1 ),
$ LDA, WORK( IR ), LDWRKR, ZERO,
$ WORK( IU ), LDWRKU )
CALL DLACPY( 'F', CHUNK, N, WORK( IU ), LDWRKU,
$ A( I, 1 ), LDA )
20 CONTINUE
*
ELSE
*
* Insufficient workspace for a fast algorithm
*
ITAU = 1
IWORK = ITAU + N
*
* Compute A=Q*R
* (Workspace: need 2*N, prefer N + N*NB)
*
CALL DGEQRF( M, N, A, LDA, WORK( ITAU ),
$ WORK( IWORK ), LWORK-IWORK+1, IERR )
*
* Copy R to VT, zeroing out below it
*
CALL DLACPY( 'U', N, N, A, LDA, VT, LDVT )
IF( N.GT.1 )
$ CALL DLASET( 'L', N-1, N-1, ZERO, ZERO,
$ VT( 2, 1 ), LDVT )
*
* Generate Q in A
* (Workspace: need 2*N, prefer N + N*NB)
*
CALL DORGQR( M, N, N, A, LDA, WORK( ITAU ),
$ WORK( IWORK ), LWORK-IWORK+1, IERR )
IE = ITAU
ITAUQ = IE + N
ITAUP = ITAUQ + N
IWORK = ITAUP + N
*
* Bidiagonalize R in VT
* (Workspace: need 4*N, prefer 3*N + 2*N*NB)
*
CALL DGEBRD( N, N, VT, LDVT, S, WORK( IE ),
$ WORK( ITAUQ ), WORK( ITAUP ),
$ WORK( IWORK ), LWORK-IWORK+1, IERR )
*
* Multiply Q in A by left vectors bidiagonalizing R
* (Workspace: need 3*N + M, prefer 3*N + M*NB)
*
CALL DORMBR( 'Q', 'R', 'N', M, N, N, VT, LDVT,
$ WORK( ITAUQ ), A, LDA, WORK( IWORK ),