File:
[local] /
rpl /
lapack /
lapack /
zggev.f
Revision
1.7:
download - view:
text,
annotated -
select for diffs -
revision graph
Tue Dec 21 13:53:45 2010 UTC (13 years, 5 months ago) by
bertrand
Branches:
MAIN
CVS tags:
rpl-4_1_3,
rpl-4_1_2,
rpl-4_1_1,
rpl-4_1_0,
rpl-4_0_24,
rpl-4_0_22,
rpl-4_0_21,
rpl-4_0_20,
rpl-4_0,
HEAD
Mise à jour de lapack vers la version 3.3.0.
1: SUBROUTINE ZGGEV( JOBVL, JOBVR, N, A, LDA, B, LDB, ALPHA, BETA,
2: $ VL, LDVL, VR, LDVR, WORK, LWORK, RWORK, INFO )
3: *
4: * -- LAPACK driver routine (version 3.2) --
5: * -- LAPACK is a software package provided by Univ. of Tennessee, --
6: * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
7: * November 2006
8: *
9: * .. Scalar Arguments ..
10: CHARACTER JOBVL, JOBVR
11: INTEGER INFO, LDA, LDB, LDVL, LDVR, LWORK, N
12: * ..
13: * .. Array Arguments ..
14: DOUBLE PRECISION RWORK( * )
15: COMPLEX*16 A( LDA, * ), ALPHA( * ), B( LDB, * ),
16: $ BETA( * ), VL( LDVL, * ), VR( LDVR, * ),
17: $ WORK( * )
18: * ..
19: *
20: * Purpose
21: * =======
22: *
23: * ZGGEV computes for a pair of N-by-N complex nonsymmetric matrices
24: * (A,B), the generalized eigenvalues, and optionally, the left and/or
25: * right generalized eigenvectors.
26: *
27: * A generalized eigenvalue for a pair of matrices (A,B) is a scalar
28: * lambda or a ratio alpha/beta = lambda, such that A - lambda*B is
29: * singular. It is usually represented as the pair (alpha,beta), as
30: * there is a reasonable interpretation for beta=0, and even for both
31: * being zero.
32: *
33: * The right generalized eigenvector v(j) corresponding to the
34: * generalized eigenvalue lambda(j) of (A,B) satisfies
35: *
36: * A * v(j) = lambda(j) * B * v(j).
37: *
38: * The left generalized eigenvector u(j) corresponding to the
39: * generalized eigenvalues lambda(j) of (A,B) satisfies
40: *
41: * u(j)**H * A = lambda(j) * u(j)**H * B
42: *
43: * where u(j)**H is the conjugate-transpose of u(j).
44: *
45: * Arguments
46: * =========
47: *
48: * JOBVL (input) CHARACTER*1
49: * = 'N': do not compute the left generalized eigenvectors;
50: * = 'V': compute the left generalized eigenvectors.
51: *
52: * JOBVR (input) CHARACTER*1
53: * = 'N': do not compute the right generalized eigenvectors;
54: * = 'V': compute the right generalized eigenvectors.
55: *
56: * N (input) INTEGER
57: * The order of the matrices A, B, VL, and VR. N >= 0.
58: *
59: * A (input/output) COMPLEX*16 array, dimension (LDA, N)
60: * On entry, the matrix A in the pair (A,B).
61: * On exit, A has been overwritten.
62: *
63: * LDA (input) INTEGER
64: * The leading dimension of A. LDA >= max(1,N).
65: *
66: * B (input/output) COMPLEX*16 array, dimension (LDB, N)
67: * On entry, the matrix B in the pair (A,B).
68: * On exit, B has been overwritten.
69: *
70: * LDB (input) INTEGER
71: * The leading dimension of B. LDB >= max(1,N).
72: *
73: * ALPHA (output) COMPLEX*16 array, dimension (N)
74: * BETA (output) COMPLEX*16 array, dimension (N)
75: * On exit, ALPHA(j)/BETA(j), j=1,...,N, will be the
76: * generalized eigenvalues.
77: *
78: * Note: the quotients ALPHA(j)/BETA(j) may easily over- or
79: * underflow, and BETA(j) may even be zero. Thus, the user
80: * should avoid naively computing the ratio alpha/beta.
81: * However, ALPHA will be always less than and usually
82: * comparable with norm(A) in magnitude, and BETA always less
83: * than and usually comparable with norm(B).
84: *
85: * VL (output) COMPLEX*16 array, dimension (LDVL,N)
86: * If JOBVL = 'V', the left generalized eigenvectors u(j) are
87: * stored one after another in the columns of VL, in the same
88: * order as their eigenvalues.
89: * Each eigenvector is scaled so the largest component has
90: * abs(real part) + abs(imag. part) = 1.
91: * Not referenced if JOBVL = 'N'.
92: *
93: * LDVL (input) INTEGER
94: * The leading dimension of the matrix VL. LDVL >= 1, and
95: * if JOBVL = 'V', LDVL >= N.
96: *
97: * VR (output) COMPLEX*16 array, dimension (LDVR,N)
98: * If JOBVR = 'V', the right generalized eigenvectors v(j) are
99: * stored one after another in the columns of VR, in the same
100: * order as their eigenvalues.
101: * Each eigenvector is scaled so the largest component has
102: * abs(real part) + abs(imag. part) = 1.
103: * Not referenced if JOBVR = 'N'.
104: *
105: * LDVR (input) INTEGER
106: * The leading dimension of the matrix VR. LDVR >= 1, and
107: * if JOBVR = 'V', LDVR >= N.
108: *
109: * WORK (workspace/output) COMPLEX*16 array, dimension (MAX(1,LWORK))
110: * On exit, if INFO = 0, WORK(1) returns the optimal LWORK.
111: *
112: * LWORK (input) INTEGER
113: * The dimension of the array WORK. LWORK >= max(1,2*N).
114: * For good performance, LWORK must generally be larger.
115: *
116: * If LWORK = -1, then a workspace query is assumed; the routine
117: * only calculates the optimal size of the WORK array, returns
118: * this value as the first entry of the WORK array, and no error
119: * message related to LWORK is issued by XERBLA.
120: *
121: * RWORK (workspace/output) DOUBLE PRECISION array, dimension (8*N)
122: *
123: * INFO (output) INTEGER
124: * = 0: successful exit
125: * < 0: if INFO = -i, the i-th argument had an illegal value.
126: * =1,...,N:
127: * The QZ iteration failed. No eigenvectors have been
128: * calculated, but ALPHA(j) and BETA(j) should be
129: * correct for j=INFO+1,...,N.
130: * > N: =N+1: other then QZ iteration failed in DHGEQZ,
131: * =N+2: error return from DTGEVC.
132: *
133: * =====================================================================
134: *
135: * .. Parameters ..
136: DOUBLE PRECISION ZERO, ONE
137: PARAMETER ( ZERO = 0.0D0, ONE = 1.0D0 )
138: COMPLEX*16 CZERO, CONE
139: PARAMETER ( CZERO = ( 0.0D0, 0.0D0 ),
140: $ CONE = ( 1.0D0, 0.0D0 ) )
141: * ..
142: * .. Local Scalars ..
143: LOGICAL ILASCL, ILBSCL, ILV, ILVL, ILVR, LQUERY
144: CHARACTER CHTEMP
145: INTEGER ICOLS, IERR, IHI, IJOBVL, IJOBVR, ILEFT, ILO,
146: $ IN, IRIGHT, IROWS, IRWRK, ITAU, IWRK, JC, JR,
147: $ LWKMIN, LWKOPT
148: DOUBLE PRECISION ANRM, ANRMTO, BIGNUM, BNRM, BNRMTO, EPS,
149: $ SMLNUM, TEMP
150: COMPLEX*16 X
151: * ..
152: * .. Local Arrays ..
153: LOGICAL LDUMMA( 1 )
154: * ..
155: * .. External Subroutines ..
156: EXTERNAL DLABAD, XERBLA, ZGEQRF, ZGGBAK, ZGGBAL, ZGGHRD,
157: $ ZHGEQZ, ZLACPY, ZLASCL, ZLASET, ZTGEVC, ZUNGQR,
158: $ ZUNMQR
159: * ..
160: * .. External Functions ..
161: LOGICAL LSAME
162: INTEGER ILAENV
163: DOUBLE PRECISION DLAMCH, ZLANGE
164: EXTERNAL LSAME, ILAENV, DLAMCH, ZLANGE
165: * ..
166: * .. Intrinsic Functions ..
167: INTRINSIC ABS, DBLE, DIMAG, MAX, SQRT
168: * ..
169: * .. Statement Functions ..
170: DOUBLE PRECISION ABS1
171: * ..
172: * .. Statement Function definitions ..
173: ABS1( X ) = ABS( DBLE( X ) ) + ABS( DIMAG( X ) )
174: * ..
175: * .. Executable Statements ..
176: *
177: * Decode the input arguments
178: *
179: IF( LSAME( JOBVL, 'N' ) ) THEN
180: IJOBVL = 1
181: ILVL = .FALSE.
182: ELSE IF( LSAME( JOBVL, 'V' ) ) THEN
183: IJOBVL = 2
184: ILVL = .TRUE.
185: ELSE
186: IJOBVL = -1
187: ILVL = .FALSE.
188: END IF
189: *
190: IF( LSAME( JOBVR, 'N' ) ) THEN
191: IJOBVR = 1
192: ILVR = .FALSE.
193: ELSE IF( LSAME( JOBVR, 'V' ) ) THEN
194: IJOBVR = 2
195: ILVR = .TRUE.
196: ELSE
197: IJOBVR = -1
198: ILVR = .FALSE.
199: END IF
200: ILV = ILVL .OR. ILVR
201: *
202: * Test the input arguments
203: *
204: INFO = 0
205: LQUERY = ( LWORK.EQ.-1 )
206: IF( IJOBVL.LE.0 ) THEN
207: INFO = -1
208: ELSE IF( IJOBVR.LE.0 ) THEN
209: INFO = -2
210: ELSE IF( N.LT.0 ) THEN
211: INFO = -3
212: ELSE IF( LDA.LT.MAX( 1, N ) ) THEN
213: INFO = -5
214: ELSE IF( LDB.LT.MAX( 1, N ) ) THEN
215: INFO = -7
216: ELSE IF( LDVL.LT.1 .OR. ( ILVL .AND. LDVL.LT.N ) ) THEN
217: INFO = -11
218: ELSE IF( LDVR.LT.1 .OR. ( ILVR .AND. LDVR.LT.N ) ) THEN
219: INFO = -13
220: END IF
221: *
222: * Compute workspace
223: * (Note: Comments in the code beginning "Workspace:" describe the
224: * minimal amount of workspace needed at that point in the code,
225: * as well as the preferred amount for good performance.
226: * NB refers to the optimal block size for the immediately
227: * following subroutine, as returned by ILAENV. The workspace is
228: * computed assuming ILO = 1 and IHI = N, the worst case.)
229: *
230: IF( INFO.EQ.0 ) THEN
231: LWKMIN = MAX( 1, 2*N )
232: LWKOPT = MAX( 1, N + N*ILAENV( 1, 'ZGEQRF', ' ', N, 1, N, 0 ) )
233: LWKOPT = MAX( LWKOPT, N +
234: $ N*ILAENV( 1, 'ZUNMQR', ' ', N, 1, N, 0 ) )
235: IF( ILVL ) THEN
236: LWKOPT = MAX( LWKOPT, N +
237: $ N*ILAENV( 1, 'ZUNGQR', ' ', N, 1, N, -1 ) )
238: END IF
239: WORK( 1 ) = LWKOPT
240: *
241: IF( LWORK.LT.LWKMIN .AND. .NOT.LQUERY )
242: $ INFO = -15
243: END IF
244: *
245: IF( INFO.NE.0 ) THEN
246: CALL XERBLA( 'ZGGEV ', -INFO )
247: RETURN
248: ELSE IF( LQUERY ) THEN
249: RETURN
250: END IF
251: *
252: * Quick return if possible
253: *
254: IF( N.EQ.0 )
255: $ RETURN
256: *
257: * Get machine constants
258: *
259: EPS = DLAMCH( 'E' )*DLAMCH( 'B' )
260: SMLNUM = DLAMCH( 'S' )
261: BIGNUM = ONE / SMLNUM
262: CALL DLABAD( SMLNUM, BIGNUM )
263: SMLNUM = SQRT( SMLNUM ) / EPS
264: BIGNUM = ONE / SMLNUM
265: *
266: * Scale A if max element outside range [SMLNUM,BIGNUM]
267: *
268: ANRM = ZLANGE( 'M', N, N, A, LDA, RWORK )
269: ILASCL = .FALSE.
270: IF( ANRM.GT.ZERO .AND. ANRM.LT.SMLNUM ) THEN
271: ANRMTO = SMLNUM
272: ILASCL = .TRUE.
273: ELSE IF( ANRM.GT.BIGNUM ) THEN
274: ANRMTO = BIGNUM
275: ILASCL = .TRUE.
276: END IF
277: IF( ILASCL )
278: $ CALL ZLASCL( 'G', 0, 0, ANRM, ANRMTO, N, N, A, LDA, IERR )
279: *
280: * Scale B if max element outside range [SMLNUM,BIGNUM]
281: *
282: BNRM = ZLANGE( 'M', N, N, B, LDB, RWORK )
283: ILBSCL = .FALSE.
284: IF( BNRM.GT.ZERO .AND. BNRM.LT.SMLNUM ) THEN
285: BNRMTO = SMLNUM
286: ILBSCL = .TRUE.
287: ELSE IF( BNRM.GT.BIGNUM ) THEN
288: BNRMTO = BIGNUM
289: ILBSCL = .TRUE.
290: END IF
291: IF( ILBSCL )
292: $ CALL ZLASCL( 'G', 0, 0, BNRM, BNRMTO, N, N, B, LDB, IERR )
293: *
294: * Permute the matrices A, B to isolate eigenvalues if possible
295: * (Real Workspace: need 6*N)
296: *
297: ILEFT = 1
298: IRIGHT = N + 1
299: IRWRK = IRIGHT + N
300: CALL ZGGBAL( 'P', N, A, LDA, B, LDB, ILO, IHI, RWORK( ILEFT ),
301: $ RWORK( IRIGHT ), RWORK( IRWRK ), IERR )
302: *
303: * Reduce B to triangular form (QR decomposition of B)
304: * (Complex Workspace: need N, prefer N*NB)
305: *
306: IROWS = IHI + 1 - ILO
307: IF( ILV ) THEN
308: ICOLS = N + 1 - ILO
309: ELSE
310: ICOLS = IROWS
311: END IF
312: ITAU = 1
313: IWRK = ITAU + IROWS
314: CALL ZGEQRF( IROWS, ICOLS, B( ILO, ILO ), LDB, WORK( ITAU ),
315: $ WORK( IWRK ), LWORK+1-IWRK, IERR )
316: *
317: * Apply the orthogonal transformation to matrix A
318: * (Complex Workspace: need N, prefer N*NB)
319: *
320: CALL ZUNMQR( 'L', 'C', IROWS, ICOLS, IROWS, B( ILO, ILO ), LDB,
321: $ WORK( ITAU ), A( ILO, ILO ), LDA, WORK( IWRK ),
322: $ LWORK+1-IWRK, IERR )
323: *
324: * Initialize VL
325: * (Complex Workspace: need N, prefer N*NB)
326: *
327: IF( ILVL ) THEN
328: CALL ZLASET( 'Full', N, N, CZERO, CONE, VL, LDVL )
329: IF( IROWS.GT.1 ) THEN
330: CALL ZLACPY( 'L', IROWS-1, IROWS-1, B( ILO+1, ILO ), LDB,
331: $ VL( ILO+1, ILO ), LDVL )
332: END IF
333: CALL ZUNGQR( IROWS, IROWS, IROWS, VL( ILO, ILO ), LDVL,
334: $ WORK( ITAU ), WORK( IWRK ), LWORK+1-IWRK, IERR )
335: END IF
336: *
337: * Initialize VR
338: *
339: IF( ILVR )
340: $ CALL ZLASET( 'Full', N, N, CZERO, CONE, VR, LDVR )
341: *
342: * Reduce to generalized Hessenberg form
343: *
344: IF( ILV ) THEN
345: *
346: * Eigenvectors requested -- work on whole matrix.
347: *
348: CALL ZGGHRD( JOBVL, JOBVR, N, ILO, IHI, A, LDA, B, LDB, VL,
349: $ LDVL, VR, LDVR, IERR )
350: ELSE
351: CALL ZGGHRD( 'N', 'N', IROWS, 1, IROWS, A( ILO, ILO ), LDA,
352: $ B( ILO, ILO ), LDB, VL, LDVL, VR, LDVR, IERR )
353: END IF
354: *
355: * Perform QZ algorithm (Compute eigenvalues, and optionally, the
356: * Schur form and Schur vectors)
357: * (Complex Workspace: need N)
358: * (Real Workspace: need N)
359: *
360: IWRK = ITAU
361: IF( ILV ) THEN
362: CHTEMP = 'S'
363: ELSE
364: CHTEMP = 'E'
365: END IF
366: CALL ZHGEQZ( CHTEMP, JOBVL, JOBVR, N, ILO, IHI, A, LDA, B, LDB,
367: $ ALPHA, BETA, VL, LDVL, VR, LDVR, WORK( IWRK ),
368: $ LWORK+1-IWRK, RWORK( IRWRK ), IERR )
369: IF( IERR.NE.0 ) THEN
370: IF( IERR.GT.0 .AND. IERR.LE.N ) THEN
371: INFO = IERR
372: ELSE IF( IERR.GT.N .AND. IERR.LE.2*N ) THEN
373: INFO = IERR - N
374: ELSE
375: INFO = N + 1
376: END IF
377: GO TO 70
378: END IF
379: *
380: * Compute Eigenvectors
381: * (Real Workspace: need 2*N)
382: * (Complex Workspace: need 2*N)
383: *
384: IF( ILV ) THEN
385: IF( ILVL ) THEN
386: IF( ILVR ) THEN
387: CHTEMP = 'B'
388: ELSE
389: CHTEMP = 'L'
390: END IF
391: ELSE
392: CHTEMP = 'R'
393: END IF
394: *
395: CALL ZTGEVC( CHTEMP, 'B', LDUMMA, N, A, LDA, B, LDB, VL, LDVL,
396: $ VR, LDVR, N, IN, WORK( IWRK ), RWORK( IRWRK ),
397: $ IERR )
398: IF( IERR.NE.0 ) THEN
399: INFO = N + 2
400: GO TO 70
401: END IF
402: *
403: * Undo balancing on VL and VR and normalization
404: * (Workspace: none needed)
405: *
406: IF( ILVL ) THEN
407: CALL ZGGBAK( 'P', 'L', N, ILO, IHI, RWORK( ILEFT ),
408: $ RWORK( IRIGHT ), N, VL, LDVL, IERR )
409: DO 30 JC = 1, N
410: TEMP = ZERO
411: DO 10 JR = 1, N
412: TEMP = MAX( TEMP, ABS1( VL( JR, JC ) ) )
413: 10 CONTINUE
414: IF( TEMP.LT.SMLNUM )
415: $ GO TO 30
416: TEMP = ONE / TEMP
417: DO 20 JR = 1, N
418: VL( JR, JC ) = VL( JR, JC )*TEMP
419: 20 CONTINUE
420: 30 CONTINUE
421: END IF
422: IF( ILVR ) THEN
423: CALL ZGGBAK( 'P', 'R', N, ILO, IHI, RWORK( ILEFT ),
424: $ RWORK( IRIGHT ), N, VR, LDVR, IERR )
425: DO 60 JC = 1, N
426: TEMP = ZERO
427: DO 40 JR = 1, N
428: TEMP = MAX( TEMP, ABS1( VR( JR, JC ) ) )
429: 40 CONTINUE
430: IF( TEMP.LT.SMLNUM )
431: $ GO TO 60
432: TEMP = ONE / TEMP
433: DO 50 JR = 1, N
434: VR( JR, JC ) = VR( JR, JC )*TEMP
435: 50 CONTINUE
436: 60 CONTINUE
437: END IF
438: END IF
439: *
440: * Undo scaling if necessary
441: *
442: IF( ILASCL )
443: $ CALL ZLASCL( 'G', 0, 0, ANRMTO, ANRM, N, 1, ALPHA, N, IERR )
444: *
445: IF( ILBSCL )
446: $ CALL ZLASCL( 'G', 0, 0, BNRMTO, BNRM, N, 1, BETA, N, IERR )
447: *
448: 70 CONTINUE
449: WORK( 1 ) = LWKOPT
450: *
451: RETURN
452: *
453: * End of ZGGEV
454: *
455: END
CVSweb interface <joel.bertrand@systella.fr>