Annotation of rpl/lapack/lapack/zhbevx.f, revision 1.18
1.8 bertrand 1: *> \brief <b> ZHBEVX computes the eigenvalues and, optionally, the left and/or right eigenvectors for OTHER matrices</b>
2: *
3: * =========== DOCUMENTATION ===========
4: *
1.15 bertrand 5: * Online html documentation available at
6: * http://www.netlib.org/lapack/explore-html/
1.8 bertrand 7: *
8: *> \htmlonly
1.15 bertrand 9: *> Download ZHBEVX + dependencies
10: *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/zhbevx.f">
11: *> [TGZ]</a>
12: *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/zhbevx.f">
13: *> [ZIP]</a>
14: *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/zhbevx.f">
1.8 bertrand 15: *> [TXT]</a>
1.15 bertrand 16: *> \endhtmlonly
1.8 bertrand 17: *
18: * Definition:
19: * ===========
20: *
21: * SUBROUTINE ZHBEVX( JOBZ, RANGE, UPLO, N, KD, AB, LDAB, Q, LDQ, VL,
22: * VU, IL, IU, ABSTOL, M, W, Z, LDZ, WORK, RWORK,
23: * IWORK, IFAIL, INFO )
1.15 bertrand 24: *
1.8 bertrand 25: * .. Scalar Arguments ..
26: * CHARACTER JOBZ, RANGE, UPLO
27: * INTEGER IL, INFO, IU, KD, LDAB, LDQ, LDZ, M, N
28: * DOUBLE PRECISION ABSTOL, VL, VU
29: * ..
30: * .. Array Arguments ..
31: * INTEGER IFAIL( * ), IWORK( * )
32: * DOUBLE PRECISION RWORK( * ), W( * )
33: * COMPLEX*16 AB( LDAB, * ), Q( LDQ, * ), WORK( * ),
34: * $ Z( LDZ, * )
35: * ..
1.15 bertrand 36: *
1.8 bertrand 37: *
38: *> \par Purpose:
39: * =============
40: *>
41: *> \verbatim
42: *>
43: *> ZHBEVX computes selected eigenvalues and, optionally, eigenvectors
44: *> of a complex Hermitian band matrix A. Eigenvalues and eigenvectors
45: *> can be selected by specifying either a range of values or a range of
46: *> indices for the desired eigenvalues.
47: *> \endverbatim
48: *
49: * Arguments:
50: * ==========
51: *
52: *> \param[in] JOBZ
53: *> \verbatim
54: *> JOBZ is CHARACTER*1
55: *> = 'N': Compute eigenvalues only;
56: *> = 'V': Compute eigenvalues and eigenvectors.
57: *> \endverbatim
58: *>
59: *> \param[in] RANGE
60: *> \verbatim
61: *> RANGE is CHARACTER*1
62: *> = 'A': all eigenvalues will be found;
63: *> = 'V': all eigenvalues in the half-open interval (VL,VU]
64: *> will be found;
65: *> = 'I': the IL-th through IU-th eigenvalues will be found.
66: *> \endverbatim
67: *>
68: *> \param[in] UPLO
69: *> \verbatim
70: *> UPLO is CHARACTER*1
71: *> = 'U': Upper triangle of A is stored;
72: *> = 'L': Lower triangle of A is stored.
73: *> \endverbatim
74: *>
75: *> \param[in] N
76: *> \verbatim
77: *> N is INTEGER
78: *> The order of the matrix A. N >= 0.
79: *> \endverbatim
80: *>
81: *> \param[in] KD
82: *> \verbatim
83: *> KD is INTEGER
84: *> The number of superdiagonals of the matrix A if UPLO = 'U',
85: *> or the number of subdiagonals if UPLO = 'L'. KD >= 0.
86: *> \endverbatim
87: *>
88: *> \param[in,out] AB
89: *> \verbatim
90: *> AB is COMPLEX*16 array, dimension (LDAB, N)
91: *> On entry, the upper or lower triangle of the Hermitian band
92: *> matrix A, stored in the first KD+1 rows of the array. The
93: *> j-th column of A is stored in the j-th column of the array AB
94: *> as follows:
95: *> if UPLO = 'U', AB(kd+1+i-j,j) = A(i,j) for max(1,j-kd)<=i<=j;
96: *> if UPLO = 'L', AB(1+i-j,j) = A(i,j) for j<=i<=min(n,j+kd).
97: *>
98: *> On exit, AB is overwritten by values generated during the
99: *> reduction to tridiagonal form.
100: *> \endverbatim
101: *>
102: *> \param[in] LDAB
103: *> \verbatim
104: *> LDAB is INTEGER
105: *> The leading dimension of the array AB. LDAB >= KD + 1.
106: *> \endverbatim
107: *>
108: *> \param[out] Q
109: *> \verbatim
110: *> Q is COMPLEX*16 array, dimension (LDQ, N)
111: *> If JOBZ = 'V', the N-by-N unitary matrix used in the
112: *> reduction to tridiagonal form.
113: *> If JOBZ = 'N', the array Q is not referenced.
114: *> \endverbatim
115: *>
116: *> \param[in] LDQ
117: *> \verbatim
118: *> LDQ is INTEGER
119: *> The leading dimension of the array Q. If JOBZ = 'V', then
120: *> LDQ >= max(1,N).
121: *> \endverbatim
122: *>
123: *> \param[in] VL
124: *> \verbatim
125: *> VL is DOUBLE PRECISION
1.13 bertrand 126: *> If RANGE='V', the lower bound of the interval to
127: *> be searched for eigenvalues. VL < VU.
128: *> Not referenced if RANGE = 'A' or 'I'.
1.8 bertrand 129: *> \endverbatim
130: *>
131: *> \param[in] VU
132: *> \verbatim
133: *> VU is DOUBLE PRECISION
1.13 bertrand 134: *> If RANGE='V', the upper bound of the interval to
1.8 bertrand 135: *> be searched for eigenvalues. VL < VU.
136: *> Not referenced if RANGE = 'A' or 'I'.
137: *> \endverbatim
138: *>
139: *> \param[in] IL
140: *> \verbatim
141: *> IL is INTEGER
1.13 bertrand 142: *> If RANGE='I', the index of the
143: *> smallest eigenvalue to be returned.
144: *> 1 <= IL <= IU <= N, if N > 0; IL = 1 and IU = 0 if N = 0.
145: *> Not referenced if RANGE = 'A' or 'V'.
1.8 bertrand 146: *> \endverbatim
147: *>
148: *> \param[in] IU
149: *> \verbatim
150: *> IU is INTEGER
1.13 bertrand 151: *> If RANGE='I', the index of the
152: *> largest eigenvalue to be returned.
1.8 bertrand 153: *> 1 <= IL <= IU <= N, if N > 0; IL = 1 and IU = 0 if N = 0.
154: *> Not referenced if RANGE = 'A' or 'V'.
155: *> \endverbatim
156: *>
157: *> \param[in] ABSTOL
158: *> \verbatim
159: *> ABSTOL is DOUBLE PRECISION
160: *> The absolute error tolerance for the eigenvalues.
161: *> An approximate eigenvalue is accepted as converged
162: *> when it is determined to lie in an interval [a,b]
163: *> of width less than or equal to
164: *>
165: *> ABSTOL + EPS * max( |a|,|b| ) ,
166: *>
167: *> where EPS is the machine precision. If ABSTOL is less than
168: *> or equal to zero, then EPS*|T| will be used in its place,
169: *> where |T| is the 1-norm of the tridiagonal matrix obtained
170: *> by reducing AB to tridiagonal form.
171: *>
172: *> Eigenvalues will be computed most accurately when ABSTOL is
173: *> set to twice the underflow threshold 2*DLAMCH('S'), not zero.
174: *> If this routine returns with INFO>0, indicating that some
175: *> eigenvectors did not converge, try setting ABSTOL to
176: *> 2*DLAMCH('S').
177: *>
178: *> See "Computing Small Singular Values of Bidiagonal Matrices
179: *> with Guaranteed High Relative Accuracy," by Demmel and
180: *> Kahan, LAPACK Working Note #3.
181: *> \endverbatim
182: *>
183: *> \param[out] M
184: *> \verbatim
185: *> M is INTEGER
186: *> The total number of eigenvalues found. 0 <= M <= N.
187: *> If RANGE = 'A', M = N, and if RANGE = 'I', M = IU-IL+1.
188: *> \endverbatim
189: *>
190: *> \param[out] W
191: *> \verbatim
192: *> W is DOUBLE PRECISION array, dimension (N)
193: *> The first M elements contain the selected eigenvalues in
194: *> ascending order.
195: *> \endverbatim
196: *>
197: *> \param[out] Z
198: *> \verbatim
199: *> Z is COMPLEX*16 array, dimension (LDZ, max(1,M))
200: *> If JOBZ = 'V', then if INFO = 0, the first M columns of Z
201: *> contain the orthonormal eigenvectors of the matrix A
202: *> corresponding to the selected eigenvalues, with the i-th
203: *> column of Z holding the eigenvector associated with W(i).
204: *> If an eigenvector fails to converge, then that column of Z
205: *> contains the latest approximation to the eigenvector, and the
206: *> index of the eigenvector is returned in IFAIL.
207: *> If JOBZ = 'N', then Z is not referenced.
208: *> Note: the user must ensure that at least max(1,M) columns are
209: *> supplied in the array Z; if RANGE = 'V', the exact value of M
210: *> is not known in advance and an upper bound must be used.
211: *> \endverbatim
212: *>
213: *> \param[in] LDZ
214: *> \verbatim
215: *> LDZ is INTEGER
216: *> The leading dimension of the array Z. LDZ >= 1, and if
217: *> JOBZ = 'V', LDZ >= max(1,N).
218: *> \endverbatim
219: *>
220: *> \param[out] WORK
221: *> \verbatim
222: *> WORK is COMPLEX*16 array, dimension (N)
223: *> \endverbatim
224: *>
225: *> \param[out] RWORK
226: *> \verbatim
227: *> RWORK is DOUBLE PRECISION array, dimension (7*N)
228: *> \endverbatim
229: *>
230: *> \param[out] IWORK
231: *> \verbatim
232: *> IWORK is INTEGER array, dimension (5*N)
233: *> \endverbatim
234: *>
235: *> \param[out] IFAIL
236: *> \verbatim
237: *> IFAIL is INTEGER array, dimension (N)
238: *> If JOBZ = 'V', then if INFO = 0, the first M elements of
239: *> IFAIL are zero. If INFO > 0, then IFAIL contains the
240: *> indices of the eigenvectors that failed to converge.
241: *> If JOBZ = 'N', then IFAIL is not referenced.
242: *> \endverbatim
243: *>
244: *> \param[out] INFO
245: *> \verbatim
246: *> INFO is INTEGER
247: *> = 0: successful exit
248: *> < 0: if INFO = -i, the i-th argument had an illegal value
249: *> > 0: if INFO = i, then i eigenvectors failed to converge.
250: *> Their indices are stored in array IFAIL.
251: *> \endverbatim
252: *
253: * Authors:
254: * ========
255: *
1.15 bertrand 256: *> \author Univ. of Tennessee
257: *> \author Univ. of California Berkeley
258: *> \author Univ. of Colorado Denver
259: *> \author NAG Ltd.
1.8 bertrand 260: *
261: *> \ingroup complex16OTHEReigen
262: *
263: * =====================================================================
1.1 bertrand 264: SUBROUTINE ZHBEVX( JOBZ, RANGE, UPLO, N, KD, AB, LDAB, Q, LDQ, VL,
265: $ VU, IL, IU, ABSTOL, M, W, Z, LDZ, WORK, RWORK,
266: $ IWORK, IFAIL, INFO )
267: *
1.18 ! bertrand 268: * -- LAPACK driver routine --
1.1 bertrand 269: * -- LAPACK is a software package provided by Univ. of Tennessee, --
270: * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
271: *
272: * .. Scalar Arguments ..
273: CHARACTER JOBZ, RANGE, UPLO
274: INTEGER IL, INFO, IU, KD, LDAB, LDQ, LDZ, M, N
275: DOUBLE PRECISION ABSTOL, VL, VU
276: * ..
277: * .. Array Arguments ..
278: INTEGER IFAIL( * ), IWORK( * )
279: DOUBLE PRECISION RWORK( * ), W( * )
280: COMPLEX*16 AB( LDAB, * ), Q( LDQ, * ), WORK( * ),
281: $ Z( LDZ, * )
282: * ..
283: *
284: * =====================================================================
285: *
286: * .. Parameters ..
287: DOUBLE PRECISION ZERO, ONE
288: PARAMETER ( ZERO = 0.0D0, ONE = 1.0D0 )
289: COMPLEX*16 CZERO, CONE
290: PARAMETER ( CZERO = ( 0.0D0, 0.0D0 ),
291: $ CONE = ( 1.0D0, 0.0D0 ) )
292: * ..
293: * .. Local Scalars ..
294: LOGICAL ALLEIG, INDEIG, LOWER, TEST, VALEIG, WANTZ
295: CHARACTER ORDER
296: INTEGER I, IINFO, IMAX, INDD, INDE, INDEE, INDIBL,
297: $ INDISP, INDIWK, INDRWK, INDWRK, ISCALE, ITMP1,
298: $ J, JJ, NSPLIT
299: DOUBLE PRECISION ABSTLL, ANRM, BIGNUM, EPS, RMAX, RMIN, SAFMIN,
300: $ SIGMA, SMLNUM, TMP1, VLL, VUU
301: COMPLEX*16 CTMP1
302: * ..
303: * .. External Functions ..
304: LOGICAL LSAME
305: DOUBLE PRECISION DLAMCH, ZLANHB
306: EXTERNAL LSAME, DLAMCH, ZLANHB
307: * ..
308: * .. External Subroutines ..
309: EXTERNAL DCOPY, DSCAL, DSTEBZ, DSTERF, XERBLA, ZCOPY,
310: $ ZGEMV, ZHBTRD, ZLACPY, ZLASCL, ZSTEIN, ZSTEQR,
311: $ ZSWAP
312: * ..
313: * .. Intrinsic Functions ..
314: INTRINSIC DBLE, MAX, MIN, SQRT
315: * ..
316: * .. Executable Statements ..
317: *
318: * Test the input parameters.
319: *
320: WANTZ = LSAME( JOBZ, 'V' )
321: ALLEIG = LSAME( RANGE, 'A' )
322: VALEIG = LSAME( RANGE, 'V' )
323: INDEIG = LSAME( RANGE, 'I' )
324: LOWER = LSAME( UPLO, 'L' )
325: *
326: INFO = 0
327: IF( .NOT.( WANTZ .OR. LSAME( JOBZ, 'N' ) ) ) THEN
328: INFO = -1
329: ELSE IF( .NOT.( ALLEIG .OR. VALEIG .OR. INDEIG ) ) THEN
330: INFO = -2
331: ELSE IF( .NOT.( LOWER .OR. LSAME( UPLO, 'U' ) ) ) THEN
332: INFO = -3
333: ELSE IF( N.LT.0 ) THEN
334: INFO = -4
335: ELSE IF( KD.LT.0 ) THEN
336: INFO = -5
337: ELSE IF( LDAB.LT.KD+1 ) THEN
338: INFO = -7
339: ELSE IF( WANTZ .AND. LDQ.LT.MAX( 1, N ) ) THEN
340: INFO = -9
341: ELSE
342: IF( VALEIG ) THEN
343: IF( N.GT.0 .AND. VU.LE.VL )
344: $ INFO = -11
345: ELSE IF( INDEIG ) THEN
346: IF( IL.LT.1 .OR. IL.GT.MAX( 1, N ) ) THEN
347: INFO = -12
348: ELSE IF( IU.LT.MIN( N, IL ) .OR. IU.GT.N ) THEN
349: INFO = -13
350: END IF
351: END IF
352: END IF
353: IF( INFO.EQ.0 ) THEN
354: IF( LDZ.LT.1 .OR. ( WANTZ .AND. LDZ.LT.N ) )
355: $ INFO = -18
356: END IF
357: *
358: IF( INFO.NE.0 ) THEN
359: CALL XERBLA( 'ZHBEVX', -INFO )
360: RETURN
361: END IF
362: *
363: * Quick return if possible
364: *
365: M = 0
366: IF( N.EQ.0 )
367: $ RETURN
368: *
369: IF( N.EQ.1 ) THEN
370: M = 1
371: IF( LOWER ) THEN
372: CTMP1 = AB( 1, 1 )
373: ELSE
374: CTMP1 = AB( KD+1, 1 )
375: END IF
376: TMP1 = DBLE( CTMP1 )
377: IF( VALEIG ) THEN
378: IF( .NOT.( VL.LT.TMP1 .AND. VU.GE.TMP1 ) )
379: $ M = 0
380: END IF
381: IF( M.EQ.1 ) THEN
1.18 ! bertrand 382: W( 1 ) = DBLE( CTMP1 )
1.1 bertrand 383: IF( WANTZ )
384: $ Z( 1, 1 ) = CONE
385: END IF
386: RETURN
387: END IF
388: *
389: * Get machine constants.
390: *
391: SAFMIN = DLAMCH( 'Safe minimum' )
392: EPS = DLAMCH( 'Precision' )
393: SMLNUM = SAFMIN / EPS
394: BIGNUM = ONE / SMLNUM
395: RMIN = SQRT( SMLNUM )
396: RMAX = MIN( SQRT( BIGNUM ), ONE / SQRT( SQRT( SAFMIN ) ) )
397: *
398: * Scale matrix to allowable range, if necessary.
399: *
400: ISCALE = 0
401: ABSTLL = ABSTOL
402: IF( VALEIG ) THEN
403: VLL = VL
404: VUU = VU
405: ELSE
406: VLL = ZERO
407: VUU = ZERO
408: END IF
409: ANRM = ZLANHB( 'M', UPLO, N, KD, AB, LDAB, RWORK )
410: IF( ANRM.GT.ZERO .AND. ANRM.LT.RMIN ) THEN
411: ISCALE = 1
412: SIGMA = RMIN / ANRM
413: ELSE IF( ANRM.GT.RMAX ) THEN
414: ISCALE = 1
415: SIGMA = RMAX / ANRM
416: END IF
417: IF( ISCALE.EQ.1 ) THEN
418: IF( LOWER ) THEN
419: CALL ZLASCL( 'B', KD, KD, ONE, SIGMA, N, N, AB, LDAB, INFO )
420: ELSE
421: CALL ZLASCL( 'Q', KD, KD, ONE, SIGMA, N, N, AB, LDAB, INFO )
422: END IF
423: IF( ABSTOL.GT.0 )
424: $ ABSTLL = ABSTOL*SIGMA
425: IF( VALEIG ) THEN
426: VLL = VL*SIGMA
427: VUU = VU*SIGMA
428: END IF
429: END IF
430: *
431: * Call ZHBTRD to reduce Hermitian band matrix to tridiagonal form.
432: *
433: INDD = 1
434: INDE = INDD + N
435: INDRWK = INDE + N
436: INDWRK = 1
437: CALL ZHBTRD( JOBZ, UPLO, N, KD, AB, LDAB, RWORK( INDD ),
438: $ RWORK( INDE ), Q, LDQ, WORK( INDWRK ), IINFO )
439: *
440: * If all eigenvalues are desired and ABSTOL is less than or equal
441: * to zero, then call DSTERF or ZSTEQR. If this fails for some
442: * eigenvalue, then try DSTEBZ.
443: *
444: TEST = .FALSE.
445: IF (INDEIG) THEN
446: IF (IL.EQ.1 .AND. IU.EQ.N) THEN
447: TEST = .TRUE.
448: END IF
449: END IF
450: IF ((ALLEIG .OR. TEST) .AND. (ABSTOL.LE.ZERO)) THEN
451: CALL DCOPY( N, RWORK( INDD ), 1, W, 1 )
452: INDEE = INDRWK + 2*N
453: IF( .NOT.WANTZ ) THEN
454: CALL DCOPY( N-1, RWORK( INDE ), 1, RWORK( INDEE ), 1 )
455: CALL DSTERF( N, W, RWORK( INDEE ), INFO )
456: ELSE
457: CALL ZLACPY( 'A', N, N, Q, LDQ, Z, LDZ )
458: CALL DCOPY( N-1, RWORK( INDE ), 1, RWORK( INDEE ), 1 )
459: CALL ZSTEQR( JOBZ, N, W, RWORK( INDEE ), Z, LDZ,
460: $ RWORK( INDRWK ), INFO )
461: IF( INFO.EQ.0 ) THEN
462: DO 10 I = 1, N
463: IFAIL( I ) = 0
464: 10 CONTINUE
465: END IF
466: END IF
467: IF( INFO.EQ.0 ) THEN
468: M = N
469: GO TO 30
470: END IF
471: INFO = 0
472: END IF
473: *
474: * Otherwise, call DSTEBZ and, if eigenvectors are desired, ZSTEIN.
475: *
476: IF( WANTZ ) THEN
477: ORDER = 'B'
478: ELSE
479: ORDER = 'E'
480: END IF
481: INDIBL = 1
482: INDISP = INDIBL + N
483: INDIWK = INDISP + N
484: CALL DSTEBZ( RANGE, ORDER, N, VLL, VUU, IL, IU, ABSTLL,
485: $ RWORK( INDD ), RWORK( INDE ), M, NSPLIT, W,
486: $ IWORK( INDIBL ), IWORK( INDISP ), RWORK( INDRWK ),
487: $ IWORK( INDIWK ), INFO )
488: *
489: IF( WANTZ ) THEN
490: CALL ZSTEIN( N, RWORK( INDD ), RWORK( INDE ), M, W,
491: $ IWORK( INDIBL ), IWORK( INDISP ), Z, LDZ,
492: $ RWORK( INDRWK ), IWORK( INDIWK ), IFAIL, INFO )
493: *
494: * Apply unitary matrix used in reduction to tridiagonal
495: * form to eigenvectors returned by ZSTEIN.
496: *
497: DO 20 J = 1, M
498: CALL ZCOPY( N, Z( 1, J ), 1, WORK( 1 ), 1 )
499: CALL ZGEMV( 'N', N, N, CONE, Q, LDQ, WORK, 1, CZERO,
500: $ Z( 1, J ), 1 )
501: 20 CONTINUE
502: END IF
503: *
504: * If matrix was scaled, then rescale eigenvalues appropriately.
505: *
506: 30 CONTINUE
507: IF( ISCALE.EQ.1 ) THEN
508: IF( INFO.EQ.0 ) THEN
509: IMAX = M
510: ELSE
511: IMAX = INFO - 1
512: END IF
513: CALL DSCAL( IMAX, ONE / SIGMA, W, 1 )
514: END IF
515: *
516: * If eigenvalues are not in order, then sort them, along with
517: * eigenvectors.
518: *
519: IF( WANTZ ) THEN
520: DO 50 J = 1, M - 1
521: I = 0
522: TMP1 = W( J )
523: DO 40 JJ = J + 1, M
524: IF( W( JJ ).LT.TMP1 ) THEN
525: I = JJ
526: TMP1 = W( JJ )
527: END IF
528: 40 CONTINUE
529: *
530: IF( I.NE.0 ) THEN
531: ITMP1 = IWORK( INDIBL+I-1 )
532: W( I ) = W( J )
533: IWORK( INDIBL+I-1 ) = IWORK( INDIBL+J-1 )
534: W( J ) = TMP1
535: IWORK( INDIBL+J-1 ) = ITMP1
536: CALL ZSWAP( N, Z( 1, I ), 1, Z( 1, J ), 1 )
537: IF( INFO.NE.0 ) THEN
538: ITMP1 = IFAIL( I )
539: IFAIL( I ) = IFAIL( J )
540: IFAIL( J ) = ITMP1
541: END IF
542: END IF
543: 50 CONTINUE
544: END IF
545: *
546: RETURN
547: *
548: * End of ZHBEVX
549: *
550: END
CVSweb interface <joel.bertrand@systella.fr>
![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to zhbevx.f CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [local] / rpl / lapack / lapack