![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to dgsvj1.f CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [local] / rpl / lapack / lapack|
Return to dgsvj1.f CVS log | Up to [local] / rpl / lapack / lapack |
1.15 bertrand 1: *> \brief \b DGSVJ1 pre-processor for the routine dgesvj, applies Jacobi rotations targeting only particular pivots.
1.7 bertrand 2: *
3: * =========== DOCUMENTATION ===========
4: *
5: * Online html documentation available at
6: * http://www.netlib.org/lapack/explore-html/
7: *
8: *> \htmlonly
9: *> Download DGSVJ1 + dependencies
10: *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/dgsvj1.f">
11: *> [TGZ]</a>
12: *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/dgsvj1.f">
13: *> [ZIP]</a>
14: *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/dgsvj1.f">
15: *> [TXT]</a>
16: *> \endhtmlonly
17: *
18: * Definition:
19: * ===========
20: *
21: * SUBROUTINE DGSVJ1( JOBV, M, N, N1, A, LDA, D, SVA, MV, V, LDV,
22: * EPS, SFMIN, TOL, NSWEEP, WORK, LWORK, INFO )
23: *
24: * .. Scalar Arguments ..
25: * DOUBLE PRECISION EPS, SFMIN, TOL
26: * INTEGER INFO, LDA, LDV, LWORK, M, MV, N, N1, NSWEEP
27: * CHARACTER*1 JOBV
28: * ..
29: * .. Array Arguments ..
30: * DOUBLE PRECISION A( LDA, * ), D( N ), SVA( N ), V( LDV, * ),
31: * $ WORK( LWORK )
32: * ..
33: *
34: *
35: *> \par Purpose:
36: * =============
37: *>
38: *> \verbatim
39: *>
1.14 bertrand 40: *> DGSVJ1 is called from DGESVJ as a pre-processor and that is its main
41: *> purpose. It applies Jacobi rotations in the same way as DGESVJ does, but
1.7 bertrand 42: *> it targets only particular pivots and it does not check convergence
43: *> (stopping criterion). Few tunning parameters (marked by [TP]) are
44: *> available for the implementer.
45: *>
46: *> Further Details
47: *> ~~~~~~~~~~~~~~~
48: *> DGSVJ1 applies few sweeps of Jacobi rotations in the column space of
49: *> the input M-by-N matrix A. The pivot pairs are taken from the (1,2)
50: *> off-diagonal block in the corresponding N-by-N Gram matrix A^T * A. The
51: *> block-entries (tiles) of the (1,2) off-diagonal block are marked by the
52: *> [x]'s in the following scheme:
53: *>
54: *> | * * * [x] [x] [x]|
55: *> | * * * [x] [x] [x]| Row-cycling in the nblr-by-nblc [x] blocks.
56: *> | * * * [x] [x] [x]| Row-cyclic pivoting inside each [x] block.
57: *> |[x] [x] [x] * * * |
58: *> |[x] [x] [x] * * * |
59: *> |[x] [x] [x] * * * |
60: *>
61: *> In terms of the columns of A, the first N1 columns are rotated 'against'
62: *> the remaining N-N1 columns, trying to increase the angle between the
63: *> corresponding subspaces. The off-diagonal block is N1-by(N-N1) and it is
64: *> tiled using quadratic tiles of side KBL. Here, KBL is a tunning parmeter.
65: *> The number of sweeps is given in NSWEEP and the orthogonality threshold
66: *> is given in TOL.
67: *> \endverbatim
68: *
69: * Arguments:
70: * ==========
71: *
72: *> \param[in] JOBV
73: *> \verbatim
74: *> JOBV is CHARACTER*1
75: *> Specifies whether the output from this procedure is used
76: *> to compute the matrix V:
77: *> = 'V': the product of the Jacobi rotations is accumulated
78: *> by postmulyiplying the N-by-N array V.
79: *> (See the description of V.)
80: *> = 'A': the product of the Jacobi rotations is accumulated
81: *> by postmulyiplying the MV-by-N array V.
82: *> (See the descriptions of MV and V.)
83: *> = 'N': the Jacobi rotations are not accumulated.
84: *> \endverbatim
85: *>
86: *> \param[in] M
87: *> \verbatim
88: *> M is INTEGER
89: *> The number of rows of the input matrix A. M >= 0.
90: *> \endverbatim
91: *>
92: *> \param[in] N
93: *> \verbatim
94: *> N is INTEGER
95: *> The number of columns of the input matrix A.
96: *> M >= N >= 0.
97: *> \endverbatim
98: *>
99: *> \param[in] N1
100: *> \verbatim
101: *> N1 is INTEGER
102: *> N1 specifies the 2 x 2 block partition, the first N1 columns are
103: *> rotated 'against' the remaining N-N1 columns of A.
104: *> \endverbatim
105: *>
106: *> \param[in,out] A
107: *> \verbatim
108: *> A is DOUBLE PRECISION array, dimension (LDA,N)
109: *> On entry, M-by-N matrix A, such that A*diag(D) represents
110: *> the input matrix.
111: *> On exit,
112: *> A_onexit * D_onexit represents the input matrix A*diag(D)
113: *> post-multiplied by a sequence of Jacobi rotations, where the
114: *> rotation threshold and the total number of sweeps are given in
115: *> TOL and NSWEEP, respectively.
116: *> (See the descriptions of N1, D, TOL and NSWEEP.)
117: *> \endverbatim
118: *>
119: *> \param[in] LDA
120: *> \verbatim
121: *> LDA is INTEGER
122: *> The leading dimension of the array A. LDA >= max(1,M).
123: *> \endverbatim
124: *>
125: *> \param[in,out] D
126: *> \verbatim
127: *> D is DOUBLE PRECISION array, dimension (N)
128: *> The array D accumulates the scaling factors from the fast scaled
129: *> Jacobi rotations.
130: *> On entry, A*diag(D) represents the input matrix.
131: *> On exit, A_onexit*diag(D_onexit) represents the input matrix
132: *> post-multiplied by a sequence of Jacobi rotations, where the
133: *> rotation threshold and the total number of sweeps are given in
134: *> TOL and NSWEEP, respectively.
135: *> (See the descriptions of N1, A, TOL and NSWEEP.)
136: *> \endverbatim
137: *>
138: *> \param[in,out] SVA
139: *> \verbatim
140: *> SVA is DOUBLE PRECISION array, dimension (N)
141: *> On entry, SVA contains the Euclidean norms of the columns of
142: *> the matrix A*diag(D).
143: *> On exit, SVA contains the Euclidean norms of the columns of
144: *> the matrix onexit*diag(D_onexit).
145: *> \endverbatim
146: *>
147: *> \param[in] MV
148: *> \verbatim
149: *> MV is INTEGER
150: *> If JOBV .EQ. 'A', then MV rows of V are post-multipled by a
151: *> sequence of Jacobi rotations.
152: *> If JOBV = 'N', then MV is not referenced.
153: *> \endverbatim
154: *>
155: *> \param[in,out] V
156: *> \verbatim
157: *> V is DOUBLE PRECISION array, dimension (LDV,N)
158: *> If JOBV .EQ. 'V' then N rows of V are post-multipled by a
159: *> sequence of Jacobi rotations.
160: *> If JOBV .EQ. 'A' then MV rows of V are post-multipled by a
161: *> sequence of Jacobi rotations.
162: *> If JOBV = 'N', then V is not referenced.
163: *> \endverbatim
164: *>
165: *> \param[in] LDV
166: *> \verbatim
167: *> LDV is INTEGER
168: *> The leading dimension of the array V, LDV >= 1.
169: *> If JOBV = 'V', LDV .GE. N.
170: *> If JOBV = 'A', LDV .GE. MV.
171: *> \endverbatim
172: *>
173: *> \param[in] EPS
174: *> \verbatim
175: *> EPS is DOUBLE PRECISION
176: *> EPS = DLAMCH('Epsilon')
177: *> \endverbatim
178: *>
179: *> \param[in] SFMIN
180: *> \verbatim
181: *> SFMIN is DOUBLE PRECISION
182: *> SFMIN = DLAMCH('Safe Minimum')
183: *> \endverbatim
184: *>
185: *> \param[in] TOL
186: *> \verbatim
187: *> TOL is DOUBLE PRECISION
188: *> TOL is the threshold for Jacobi rotations. For a pair
189: *> A(:,p), A(:,q) of pivot columns, the Jacobi rotation is
190: *> applied only if DABS(COS(angle(A(:,p),A(:,q)))) .GT. TOL.
191: *> \endverbatim
192: *>
193: *> \param[in] NSWEEP
194: *> \verbatim
195: *> NSWEEP is INTEGER
196: *> NSWEEP is the number of sweeps of Jacobi rotations to be
197: *> performed.
198: *> \endverbatim
199: *>
200: *> \param[out] WORK
201: *> \verbatim
202: *> WORK is DOUBLE PRECISION array, dimension (LWORK)
203: *> \endverbatim
204: *>
205: *> \param[in] LWORK
206: *> \verbatim
207: *> LWORK is INTEGER
208: *> LWORK is the dimension of WORK. LWORK .GE. M.
209: *> \endverbatim
210: *>
211: *> \param[out] INFO
212: *> \verbatim
213: *> INFO is INTEGER
214: *> = 0 : successful exit.
215: *> < 0 : if INFO = -i, then the i-th argument had an illegal value
216: *> \endverbatim
217: *
218: * Authors:
219: * ========
220: *
221: *> \author Univ. of Tennessee
222: *> \author Univ. of California Berkeley
223: *> \author Univ. of Colorado Denver
224: *> \author NAG Ltd.
225: *
1.15 bertrand 226: *> \date June 2016
1.7 bertrand 227: *
228: *> \ingroup doubleOTHERcomputational
229: *
230: *> \par Contributors:
231: * ==================
232: *>
233: *> Zlatko Drmac (Zagreb, Croatia) and Kresimir Veselic (Hagen, Germany)
234: *
235: * =====================================================================
1.1 bertrand 236: SUBROUTINE DGSVJ1( JOBV, M, N, N1, A, LDA, D, SVA, MV, V, LDV,
1.6 bertrand 237: $ EPS, SFMIN, TOL, NSWEEP, WORK, LWORK, INFO )
1.1 bertrand 238: *
1.15 bertrand 239: * -- LAPACK computational routine (version 3.6.1) --
1.1 bertrand 240: * -- LAPACK is a software package provided by Univ. of Tennessee, --
241: * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
1.15 bertrand 242: * June 2016
1.1 bertrand 243: *
244: * .. Scalar Arguments ..
245: DOUBLE PRECISION EPS, SFMIN, TOL
246: INTEGER INFO, LDA, LDV, LWORK, M, MV, N, N1, NSWEEP
247: CHARACTER*1 JOBV
248: * ..
249: * .. Array Arguments ..
250: DOUBLE PRECISION A( LDA, * ), D( N ), SVA( N ), V( LDV, * ),
1.6 bertrand 251: $ WORK( LWORK )
1.1 bertrand 252: * ..
253: *
254: * =====================================================================
255: *
256: * .. Local Parameters ..
1.9 bertrand 257: DOUBLE PRECISION ZERO, HALF, ONE
258: PARAMETER ( ZERO = 0.0D0, HALF = 0.5D0, ONE = 1.0D0 )
1.1 bertrand 259: * ..
260: * .. Local Scalars ..
261: DOUBLE PRECISION AAPP, AAPP0, AAPQ, AAQQ, APOAQ, AQOAP, BIG,
1.6 bertrand 262: $ BIGTHETA, CS, LARGE, MXAAPQ, MXSINJ, ROOTBIG,
263: $ ROOTEPS, ROOTSFMIN, ROOTTOL, SMALL, SN, T,
264: $ TEMP1, THETA, THSIGN
1.1 bertrand 265: INTEGER BLSKIP, EMPTSW, i, ibr, igl, IERR, IJBLSK,
1.6 bertrand 266: $ ISWROT, jbc, jgl, KBL, MVL, NOTROT, nblc, nblr,
267: $ p, PSKIPPED, q, ROWSKIP, SWBAND
1.1 bertrand 268: LOGICAL APPLV, ROTOK, RSVEC
269: * ..
270: * .. Local Arrays ..
271: DOUBLE PRECISION FASTR( 5 )
272: * ..
273: * .. Intrinsic Functions ..
1.14 bertrand 274: INTRINSIC DABS, MAX, DBLE, MIN, DSIGN, DSQRT
1.1 bertrand 275: * ..
276: * .. External Functions ..
277: DOUBLE PRECISION DDOT, DNRM2
278: INTEGER IDAMAX
279: LOGICAL LSAME
280: EXTERNAL IDAMAX, LSAME, DDOT, DNRM2
281: * ..
282: * .. External Subroutines ..
283: EXTERNAL DAXPY, DCOPY, DLASCL, DLASSQ, DROTM, DSWAP
284: * ..
285: * .. Executable Statements ..
286: *
287: * Test the input parameters.
288: *
289: APPLV = LSAME( JOBV, 'A' )
290: RSVEC = LSAME( JOBV, 'V' )
291: IF( .NOT.( RSVEC .OR. APPLV .OR. LSAME( JOBV, 'N' ) ) ) THEN
292: INFO = -1
293: ELSE IF( M.LT.0 ) THEN
294: INFO = -2
295: ELSE IF( ( N.LT.0 ) .OR. ( N.GT.M ) ) THEN
296: INFO = -3
297: ELSE IF( N1.LT.0 ) THEN
298: INFO = -4
299: ELSE IF( LDA.LT.M ) THEN
300: INFO = -6
1.4 bertrand 301: ELSE IF( ( RSVEC.OR.APPLV ) .AND. ( MV.LT.0 ) ) THEN
1.1 bertrand 302: INFO = -9
1.4 bertrand 303: ELSE IF( ( RSVEC.AND.( LDV.LT.N ) ).OR.
1.6 bertrand 304: $ ( APPLV.AND.( LDV.LT.MV ) ) ) THEN
1.1 bertrand 305: INFO = -11
306: ELSE IF( TOL.LE.EPS ) THEN
307: INFO = -14
308: ELSE IF( NSWEEP.LT.0 ) THEN
309: INFO = -15
310: ELSE IF( LWORK.LT.M ) THEN
311: INFO = -17
312: ELSE
313: INFO = 0
314: END IF
315: *
316: * #:(
317: IF( INFO.NE.0 ) THEN
318: CALL XERBLA( 'DGSVJ1', -INFO )
319: RETURN
320: END IF
321: *
322: IF( RSVEC ) THEN
323: MVL = N
324: ELSE IF( APPLV ) THEN
325: MVL = MV
326: END IF
327: RSVEC = RSVEC .OR. APPLV
328:
329: ROOTEPS = DSQRT( EPS )
330: ROOTSFMIN = DSQRT( SFMIN )
331: SMALL = SFMIN / EPS
332: BIG = ONE / SFMIN
333: ROOTBIG = ONE / ROOTSFMIN
334: LARGE = BIG / DSQRT( DBLE( M*N ) )
335: BIGTHETA = ONE / ROOTEPS
336: ROOTTOL = DSQRT( TOL )
337: *
338: * .. Initialize the right singular vector matrix ..
339: *
340: * RSVEC = LSAME( JOBV, 'Y' )
341: *
342: EMPTSW = N1*( N-N1 )
343: NOTROT = 0
344: FASTR( 1 ) = ZERO
345: *
346: * .. Row-cyclic pivot strategy with de Rijk's pivoting ..
347: *
1.14 bertrand 348: KBL = MIN( 8, N )
1.1 bertrand 349: NBLR = N1 / KBL
350: IF( ( NBLR*KBL ).NE.N1 )NBLR = NBLR + 1
351:
352: * .. the tiling is nblr-by-nblc [tiles]
353:
354: NBLC = ( N-N1 ) / KBL
355: IF( ( NBLC*KBL ).NE.( N-N1 ) )NBLC = NBLC + 1
356: BLSKIP = ( KBL**2 ) + 1
357: *[TP] BLKSKIP is a tuning parameter that depends on SWBAND and KBL.
358:
1.14 bertrand 359: ROWSKIP = MIN( 5, KBL )
1.1 bertrand 360: *[TP] ROWSKIP is a tuning parameter.
361: SWBAND = 0
362: *[TP] SWBAND is a tuning parameter. It is meaningful and effective
363: * if SGESVJ is used as a computational routine in the preconditioned
364: * Jacobi SVD algorithm SGESVJ.
365: *
366: *
367: * | * * * [x] [x] [x]|
368: * | * * * [x] [x] [x]| Row-cycling in the nblr-by-nblc [x] blocks.
369: * | * * * [x] [x] [x]| Row-cyclic pivoting inside each [x] block.
370: * |[x] [x] [x] * * * |
371: * |[x] [x] [x] * * * |
372: * |[x] [x] [x] * * * |
373: *
374: *
375: DO 1993 i = 1, NSWEEP
376: * .. go go go ...
377: *
378: MXAAPQ = ZERO
379: MXSINJ = ZERO
380: ISWROT = 0
381: *
382: NOTROT = 0
383: PSKIPPED = 0
384: *
385: DO 2000 ibr = 1, NBLR
386:
387: igl = ( ibr-1 )*KBL + 1
388: *
389: *
390: *........................................................
391: * ... go to the off diagonal blocks
392:
393: igl = ( ibr-1 )*KBL + 1
394:
395: DO 2010 jbc = 1, NBLC
396:
397: jgl = N1 + ( jbc-1 )*KBL + 1
398:
399: * doing the block at ( ibr, jbc )
400:
401: IJBLSK = 0
1.14 bertrand 402: DO 2100 p = igl, MIN( igl+KBL-1, N1 )
1.1 bertrand 403:
404: AAPP = SVA( p )
405:
406: IF( AAPP.GT.ZERO ) THEN
407:
408: PSKIPPED = 0
409:
1.14 bertrand 410: DO 2200 q = jgl, MIN( jgl+KBL-1, N )
1.1 bertrand 411: *
412: AAQQ = SVA( q )
413:
414: IF( AAQQ.GT.ZERO ) THEN
415: AAPP0 = AAPP
416: *
417: * .. M x 2 Jacobi SVD ..
418: *
419: * .. Safe Gram matrix computation ..
420: *
421: IF( AAQQ.GE.ONE ) THEN
422: IF( AAPP.GE.AAQQ ) THEN
423: ROTOK = ( SMALL*AAPP ).LE.AAQQ
424: ELSE
425: ROTOK = ( SMALL*AAQQ ).LE.AAPP
426: END IF
427: IF( AAPP.LT.( BIG / AAQQ ) ) THEN
428: AAPQ = ( DDOT( M, A( 1, p ), 1, A( 1,
1.6 bertrand 429: $ q ), 1 )*D( p )*D( q ) / AAQQ )
430: $ / AAPP
1.1 bertrand 431: ELSE
432: CALL DCOPY( M, A( 1, p ), 1, WORK, 1 )
433: CALL DLASCL( 'G', 0, 0, AAPP, D( p ),
1.6 bertrand 434: $ M, 1, WORK, LDA, IERR )
1.1 bertrand 435: AAPQ = DDOT( M, WORK, 1, A( 1, q ),
1.6 bertrand 436: $ 1 )*D( q ) / AAQQ
1.1 bertrand 437: END IF
438: ELSE
439: IF( AAPP.GE.AAQQ ) THEN
440: ROTOK = AAPP.LE.( AAQQ / SMALL )
441: ELSE
442: ROTOK = AAQQ.LE.( AAPP / SMALL )
443: END IF
444: IF( AAPP.GT.( SMALL / AAQQ ) ) THEN
445: AAPQ = ( DDOT( M, A( 1, p ), 1, A( 1,
1.6 bertrand 446: $ q ), 1 )*D( p )*D( q ) / AAQQ )
447: $ / AAPP
1.1 bertrand 448: ELSE
449: CALL DCOPY( M, A( 1, q ), 1, WORK, 1 )
450: CALL DLASCL( 'G', 0, 0, AAQQ, D( q ),
1.6 bertrand 451: $ M, 1, WORK, LDA, IERR )
1.1 bertrand 452: AAPQ = DDOT( M, WORK, 1, A( 1, p ),
1.6 bertrand 453: $ 1 )*D( p ) / AAPP
1.1 bertrand 454: END IF
455: END IF
456:
1.14 bertrand 457: MXAAPQ = MAX( MXAAPQ, DABS( AAPQ ) )
1.1 bertrand 458:
459: * TO rotate or NOT to rotate, THAT is the question ...
460: *
461: IF( DABS( AAPQ ).GT.TOL ) THEN
462: NOTROT = 0
463: * ROTATED = ROTATED + 1
464: PSKIPPED = 0
465: ISWROT = ISWROT + 1
466: *
467: IF( ROTOK ) THEN
468: *
469: AQOAP = AAQQ / AAPP
470: APOAQ = AAPP / AAQQ
1.6 bertrand 471: THETA = -HALF*DABS(AQOAP-APOAQ) / AAPQ
1.1 bertrand 472: IF( AAQQ.GT.AAPP0 )THETA = -THETA
473:
474: IF( DABS( THETA ).GT.BIGTHETA ) THEN
475: T = HALF / THETA
476: FASTR( 3 ) = T*D( p ) / D( q )
477: FASTR( 4 ) = -T*D( q ) / D( p )
478: CALL DROTM( M, A( 1, p ), 1,
1.6 bertrand 479: $ A( 1, q ), 1, FASTR )
1.1 bertrand 480: IF( RSVEC )CALL DROTM( MVL,
1.6 bertrand 481: $ V( 1, p ), 1,
482: $ V( 1, q ), 1,
483: $ FASTR )
1.14 bertrand 484: SVA( q ) = AAQQ*DSQRT( MAX( ZERO,
1.6 bertrand 485: $ ONE+T*APOAQ*AAPQ ) )
1.14 bertrand 486: AAPP = AAPP*DSQRT( MAX( ZERO,
1.6 bertrand 487: $ ONE-T*AQOAP*AAPQ ) )
1.14 bertrand 488: MXSINJ = MAX( MXSINJ, DABS( T ) )
1.1 bertrand 489: ELSE
490: *
491: * .. choose correct signum for THETA and rotate
492: *
493: THSIGN = -DSIGN( ONE, AAPQ )
494: IF( AAQQ.GT.AAPP0 )THSIGN = -THSIGN
495: T = ONE / ( THETA+THSIGN*
1.6 bertrand 496: $ DSQRT( ONE+THETA*THETA ) )
1.1 bertrand 497: CS = DSQRT( ONE / ( ONE+T*T ) )
498: SN = T*CS
1.14 bertrand 499: MXSINJ = MAX( MXSINJ, DABS( SN ) )
500: SVA( q ) = AAQQ*DSQRT( MAX( ZERO,
1.6 bertrand 501: $ ONE+T*APOAQ*AAPQ ) )
1.14 bertrand 502: AAPP = AAPP*DSQRT( MAX( ZERO,
1.6 bertrand 503: $ ONE-T*AQOAP*AAPQ ) )
1.1 bertrand 504:
505: APOAQ = D( p ) / D( q )
506: AQOAP = D( q ) / D( p )
507: IF( D( p ).GE.ONE ) THEN
508: *
509: IF( D( q ).GE.ONE ) THEN
510: FASTR( 3 ) = T*APOAQ
511: FASTR( 4 ) = -T*AQOAP
512: D( p ) = D( p )*CS
513: D( q ) = D( q )*CS
514: CALL DROTM( M, A( 1, p ), 1,
1.6 bertrand 515: $ A( 1, q ), 1,
516: $ FASTR )
1.1 bertrand 517: IF( RSVEC )CALL DROTM( MVL,
1.6 bertrand 518: $ V( 1, p ), 1, V( 1, q ),
519: $ 1, FASTR )
1.1 bertrand 520: ELSE
521: CALL DAXPY( M, -T*AQOAP,
1.6 bertrand 522: $ A( 1, q ), 1,
523: $ A( 1, p ), 1 )
1.1 bertrand 524: CALL DAXPY( M, CS*SN*APOAQ,
1.6 bertrand 525: $ A( 1, p ), 1,
526: $ A( 1, q ), 1 )
1.1 bertrand 527: IF( RSVEC ) THEN
528: CALL DAXPY( MVL, -T*AQOAP,
1.6 bertrand 529: $ V( 1, q ), 1,
530: $ V( 1, p ), 1 )
1.1 bertrand 531: CALL DAXPY( MVL,
1.6 bertrand 532: $ CS*SN*APOAQ,
533: $ V( 1, p ), 1,
534: $ V( 1, q ), 1 )
1.1 bertrand 535: END IF
536: D( p ) = D( p )*CS
537: D( q ) = D( q ) / CS
538: END IF
539: ELSE
540: IF( D( q ).GE.ONE ) THEN
541: CALL DAXPY( M, T*APOAQ,
1.6 bertrand 542: $ A( 1, p ), 1,
543: $ A( 1, q ), 1 )
1.1 bertrand 544: CALL DAXPY( M, -CS*SN*AQOAP,
1.6 bertrand 545: $ A( 1, q ), 1,
546: $ A( 1, p ), 1 )
1.1 bertrand 547: IF( RSVEC ) THEN
548: CALL DAXPY( MVL, T*APOAQ,
1.6 bertrand 549: $ V( 1, p ), 1,
550: $ V( 1, q ), 1 )
1.1 bertrand 551: CALL DAXPY( MVL,
1.6 bertrand 552: $ -CS*SN*AQOAP,
553: $ V( 1, q ), 1,
554: $ V( 1, p ), 1 )
1.1 bertrand 555: END IF
556: D( p ) = D( p ) / CS
557: D( q ) = D( q )*CS
558: ELSE
559: IF( D( p ).GE.D( q ) ) THEN
560: CALL DAXPY( M, -T*AQOAP,
1.6 bertrand 561: $ A( 1, q ), 1,
562: $ A( 1, p ), 1 )
1.1 bertrand 563: CALL DAXPY( M, CS*SN*APOAQ,
1.6 bertrand 564: $ A( 1, p ), 1,
565: $ A( 1, q ), 1 )
1.1 bertrand 566: D( p ) = D( p )*CS
567: D( q ) = D( q ) / CS
568: IF( RSVEC ) THEN
569: CALL DAXPY( MVL,
1.6 bertrand 570: $ -T*AQOAP,
571: $ V( 1, q ), 1,
572: $ V( 1, p ), 1 )
1.1 bertrand 573: CALL DAXPY( MVL,
1.6 bertrand 574: $ CS*SN*APOAQ,
575: $ V( 1, p ), 1,
576: $ V( 1, q ), 1 )
1.1 bertrand 577: END IF
578: ELSE
579: CALL DAXPY( M, T*APOAQ,
1.6 bertrand 580: $ A( 1, p ), 1,
581: $ A( 1, q ), 1 )
1.1 bertrand 582: CALL DAXPY( M,
1.6 bertrand 583: $ -CS*SN*AQOAP,
584: $ A( 1, q ), 1,
585: $ A( 1, p ), 1 )
1.1 bertrand 586: D( p ) = D( p ) / CS
587: D( q ) = D( q )*CS
588: IF( RSVEC ) THEN
589: CALL DAXPY( MVL,
1.6 bertrand 590: $ T*APOAQ, V( 1, p ),
591: $ 1, V( 1, q ), 1 )
1.1 bertrand 592: CALL DAXPY( MVL,
1.6 bertrand 593: $ -CS*SN*AQOAP,
594: $ V( 1, q ), 1,
595: $ V( 1, p ), 1 )
1.1 bertrand 596: END IF
597: END IF
598: END IF
599: END IF
600: END IF
601:
602: ELSE
603: IF( AAPP.GT.AAQQ ) THEN
604: CALL DCOPY( M, A( 1, p ), 1, WORK,
1.6 bertrand 605: $ 1 )
1.1 bertrand 606: CALL DLASCL( 'G', 0, 0, AAPP, ONE,
1.6 bertrand 607: $ M, 1, WORK, LDA, IERR )
1.1 bertrand 608: CALL DLASCL( 'G', 0, 0, AAQQ, ONE,
1.6 bertrand 609: $ M, 1, A( 1, q ), LDA,
610: $ IERR )
1.1 bertrand 611: TEMP1 = -AAPQ*D( p ) / D( q )
612: CALL DAXPY( M, TEMP1, WORK, 1,
1.6 bertrand 613: $ A( 1, q ), 1 )
1.1 bertrand 614: CALL DLASCL( 'G', 0, 0, ONE, AAQQ,
1.6 bertrand 615: $ M, 1, A( 1, q ), LDA,
616: $ IERR )
1.14 bertrand 617: SVA( q ) = AAQQ*DSQRT( MAX( ZERO,
1.6 bertrand 618: $ ONE-AAPQ*AAPQ ) )
1.14 bertrand 619: MXSINJ = MAX( MXSINJ, SFMIN )
1.1 bertrand 620: ELSE
621: CALL DCOPY( M, A( 1, q ), 1, WORK,
1.6 bertrand 622: $ 1 )
1.1 bertrand 623: CALL DLASCL( 'G', 0, 0, AAQQ, ONE,
1.6 bertrand 624: $ M, 1, WORK, LDA, IERR )
1.1 bertrand 625: CALL DLASCL( 'G', 0, 0, AAPP, ONE,
1.6 bertrand 626: $ M, 1, A( 1, p ), LDA,
627: $ IERR )
1.1 bertrand 628: TEMP1 = -AAPQ*D( q ) / D( p )
629: CALL DAXPY( M, TEMP1, WORK, 1,
1.6 bertrand 630: $ A( 1, p ), 1 )
1.1 bertrand 631: CALL DLASCL( 'G', 0, 0, ONE, AAPP,
1.6 bertrand 632: $ M, 1, A( 1, p ), LDA,
633: $ IERR )
1.14 bertrand 634: SVA( p ) = AAPP*DSQRT( MAX( ZERO,
1.6 bertrand 635: $ ONE-AAPQ*AAPQ ) )
1.14 bertrand 636: MXSINJ = MAX( MXSINJ, SFMIN )
1.1 bertrand 637: END IF
638: END IF
639: * END IF ROTOK THEN ... ELSE
640: *
641: * In the case of cancellation in updating SVA(q)
642: * .. recompute SVA(q)
643: IF( ( SVA( q ) / AAQQ )**2.LE.ROOTEPS )
1.6 bertrand 644: $ THEN
1.1 bertrand 645: IF( ( AAQQ.LT.ROOTBIG ) .AND.
1.6 bertrand 646: $ ( AAQQ.GT.ROOTSFMIN ) ) THEN
1.1 bertrand 647: SVA( q ) = DNRM2( M, A( 1, q ), 1 )*
1.6 bertrand 648: $ D( q )
1.1 bertrand 649: ELSE
650: T = ZERO
1.4 bertrand 651: AAQQ = ONE
1.1 bertrand 652: CALL DLASSQ( M, A( 1, q ), 1, T,
1.6 bertrand 653: $ AAQQ )
1.1 bertrand 654: SVA( q ) = T*DSQRT( AAQQ )*D( q )
655: END IF
656: END IF
657: IF( ( AAPP / AAPP0 )**2.LE.ROOTEPS ) THEN
658: IF( ( AAPP.LT.ROOTBIG ) .AND.
1.6 bertrand 659: $ ( AAPP.GT.ROOTSFMIN ) ) THEN
1.1 bertrand 660: AAPP = DNRM2( M, A( 1, p ), 1 )*
1.6 bertrand 661: $ D( p )
1.1 bertrand 662: ELSE
663: T = ZERO
1.4 bertrand 664: AAPP = ONE
1.1 bertrand 665: CALL DLASSQ( M, A( 1, p ), 1, T,
1.6 bertrand 666: $ AAPP )
1.1 bertrand 667: AAPP = T*DSQRT( AAPP )*D( p )
668: END IF
669: SVA( p ) = AAPP
670: END IF
671: * end of OK rotation
672: ELSE
673: NOTROT = NOTROT + 1
674: * SKIPPED = SKIPPED + 1
675: PSKIPPED = PSKIPPED + 1
676: IJBLSK = IJBLSK + 1
677: END IF
678: ELSE
679: NOTROT = NOTROT + 1
680: PSKIPPED = PSKIPPED + 1
681: IJBLSK = IJBLSK + 1
682: END IF
683:
684: * IF ( NOTROT .GE. EMPTSW ) GO TO 2011
685: IF( ( i.LE.SWBAND ) .AND. ( IJBLSK.GE.BLSKIP ) )
1.6 bertrand 686: $ THEN
1.1 bertrand 687: SVA( p ) = AAPP
688: NOTROT = 0
689: GO TO 2011
690: END IF
691: IF( ( i.LE.SWBAND ) .AND.
1.6 bertrand 692: $ ( PSKIPPED.GT.ROWSKIP ) ) THEN
1.1 bertrand 693: AAPP = -AAPP
694: NOTROT = 0
695: GO TO 2203
696: END IF
697:
698: *
699: 2200 CONTINUE
700: * end of the q-loop
701: 2203 CONTINUE
702:
703: SVA( p ) = AAPP
704: *
705: ELSE
706: IF( AAPP.EQ.ZERO )NOTROT = NOTROT +
1.14 bertrand 707: $ MIN( jgl+KBL-1, N ) - jgl + 1
1.1 bertrand 708: IF( AAPP.LT.ZERO )NOTROT = 0
709: *** IF ( NOTROT .GE. EMPTSW ) GO TO 2011
710: END IF
711:
712: 2100 CONTINUE
713: * end of the p-loop
714: 2010 CONTINUE
715: * end of the jbc-loop
716: 2011 CONTINUE
717: *2011 bailed out of the jbc-loop
1.14 bertrand 718: DO 2012 p = igl, MIN( igl+KBL-1, N )
1.1 bertrand 719: SVA( p ) = DABS( SVA( p ) )
720: 2012 CONTINUE
721: *** IF ( NOTROT .GE. EMPTSW ) GO TO 1994
722: 2000 CONTINUE
723: *2000 :: end of the ibr-loop
724: *
725: * .. update SVA(N)
726: IF( ( SVA( N ).LT.ROOTBIG ) .AND. ( SVA( N ).GT.ROOTSFMIN ) )
1.6 bertrand 727: $ THEN
1.1 bertrand 728: SVA( N ) = DNRM2( M, A( 1, N ), 1 )*D( N )
729: ELSE
730: T = ZERO
1.4 bertrand 731: AAPP = ONE
1.1 bertrand 732: CALL DLASSQ( M, A( 1, N ), 1, T, AAPP )
733: SVA( N ) = T*DSQRT( AAPP )*D( N )
734: END IF
735: *
736: * Additional steering devices
737: *
738: IF( ( i.LT.SWBAND ) .AND. ( ( MXAAPQ.LE.ROOTTOL ) .OR.
1.6 bertrand 739: $ ( ISWROT.LE.N ) ) )SWBAND = i
1.1 bertrand 740:
741: IF( ( i.GT.SWBAND+1 ) .AND. ( MXAAPQ.LT.DBLE( N )*TOL ) .AND.
1.6 bertrand 742: $ ( DBLE( N )*MXAAPQ*MXSINJ.LT.TOL ) ) THEN
1.1 bertrand 743: GO TO 1994
744: END IF
745:
746: *
747: IF( NOTROT.GE.EMPTSW )GO TO 1994
748:
749: 1993 CONTINUE
750: * end i=1:NSWEEP loop
751: * #:) Reaching this point means that the procedure has completed the given
752: * number of sweeps.
753: INFO = NSWEEP - 1
754: GO TO 1995
755: 1994 CONTINUE
756: * #:) Reaching this point means that during the i-th sweep all pivots were
757: * below the given threshold, causing early exit.
758:
759: INFO = 0
760: * #:) INFO = 0 confirms successful iterations.
761: 1995 CONTINUE
762: *
763: * Sort the vector D
764: *
765: DO 5991 p = 1, N - 1
766: q = IDAMAX( N-p+1, SVA( p ), 1 ) + p - 1
767: IF( p.NE.q ) THEN
768: TEMP1 = SVA( p )
769: SVA( p ) = SVA( q )
770: SVA( q ) = TEMP1
771: TEMP1 = D( p )
772: D( p ) = D( q )
773: D( q ) = TEMP1
774: CALL DSWAP( M, A( 1, p ), 1, A( 1, q ), 1 )
775: IF( RSVEC )CALL DSWAP( MVL, V( 1, p ), 1, V( 1, q ), 1 )
776: END IF
777: 5991 CONTINUE
778: *
779: RETURN
780: * ..
781: * .. END OF DGSVJ1
782: * ..
783: END