 |
LAPACK 3.3.1
Linear Algebra PACKage
|
|
LAPACK 3.3.1
Linear Algebra PACKage
|
00001 SUBROUTINE DGEESX( JOBVS, SORT, SELECT, SENSE, N, A, LDA, SDIM, 00002 $ WR, WI, VS, LDVS, RCONDE, RCONDV, WORK, LWORK, 00003 $ IWORK, LIWORK, BWORK, INFO ) 00004 * 00005 * -- LAPACK driver routine (version 3.2.2) -- 00006 * -- LAPACK is a software package provided by Univ. of Tennessee, -- 00007 * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- 00008 * June 2010 00009 * 00010 * .. Scalar Arguments .. 00011 CHARACTER JOBVS, SENSE, SORT 00012 INTEGER INFO, LDA, LDVS, LIWORK, LWORK, N, SDIM 00013 DOUBLE PRECISION RCONDE, RCONDV 00014 * .. 00015 * .. Array Arguments .. 00016 LOGICAL BWORK( * ) 00017 INTEGER IWORK( * ) 00018 DOUBLE PRECISION A( LDA, * ), VS( LDVS, * ), WI( * ), WORK( * ), 00019 $ WR( * ) 00020 * .. 00021 * .. Function Arguments .. 00022 LOGICAL SELECT 00023 EXTERNAL SELECT 00024 * .. 00025 * 00026 * Purpose 00027 * ======= 00028 * 00029 * DGEESX computes for an N-by-N real nonsymmetric matrix A, the 00030 * eigenvalues, the real Schur form T, and, optionally, the matrix of 00031 * Schur vectors Z. This gives the Schur factorization A = Z*T*(Z**T). 00032 * 00033 * Optionally, it also orders the eigenvalues on the diagonal of the 00034 * real Schur form so that selected eigenvalues are at the top left; 00035 * computes a reciprocal condition number for the average of the 00036 * selected eigenvalues (RCONDE); and computes a reciprocal condition 00037 * number for the right invariant subspace corresponding to the 00038 * selected eigenvalues (RCONDV). The leading columns of Z form an 00039 * orthonormal basis for this invariant subspace. 00040 * 00041 * For further explanation of the reciprocal condition numbers RCONDE 00042 * and RCONDV, see Section 4.10 of the LAPACK Users' Guide (where 00043 * these quantities are called s and sep respectively). 00044 * 00045 * A real matrix is in real Schur form if it is upper quasi-triangular 00046 * with 1-by-1 and 2-by-2 blocks. 2-by-2 blocks will be standardized in 00047 * the form 00048 * [ a b ] 00049 * [ c a ] 00050 * 00051 * where b*c < 0. The eigenvalues of such a block are a +- sqrt(bc). 00052 * 00053 * Arguments 00054 * ========= 00055 * 00056 * JOBVS (input) CHARACTER*1 00057 * = 'N': Schur vectors are not computed; 00058 * = 'V': Schur vectors are computed. 00059 * 00060 * SORT (input) CHARACTER*1 00061 * Specifies whether or not to order the eigenvalues on the 00062 * diagonal of the Schur form. 00063 * = 'N': Eigenvalues are not ordered; 00064 * = 'S': Eigenvalues are ordered (see SELECT). 00065 * 00066 * SELECT (external procedure) LOGICAL FUNCTION of two DOUBLE PRECISION arguments 00067 * SELECT must be declared EXTERNAL in the calling subroutine. 00068 * If SORT = 'S', SELECT is used to select eigenvalues to sort 00069 * to the top left of the Schur form. 00070 * If SORT = 'N', SELECT is not referenced. 00071 * An eigenvalue WR(j)+sqrt(-1)*WI(j) is selected if 00072 * SELECT(WR(j),WI(j)) is true; i.e., if either one of a 00073 * complex conjugate pair of eigenvalues is selected, then both 00074 * are. Note that a selected complex eigenvalue may no longer 00075 * satisfy SELECT(WR(j),WI(j)) = .TRUE. after ordering, since 00076 * ordering may change the value of complex eigenvalues 00077 * (especially if the eigenvalue is ill-conditioned); in this 00078 * case INFO may be set to N+3 (see INFO below). 00079 * 00080 * SENSE (input) CHARACTER*1 00081 * Determines which reciprocal condition numbers are computed. 00082 * = 'N': None are computed; 00083 * = 'E': Computed for average of selected eigenvalues only; 00084 * = 'V': Computed for selected right invariant subspace only; 00085 * = 'B': Computed for both. 00086 * If SENSE = 'E', 'V' or 'B', SORT must equal 'S'. 00087 * 00088 * N (input) INTEGER 00089 * The order of the matrix A. N >= 0. 00090 * 00091 * A (input/output) DOUBLE PRECISION array, dimension (LDA, N) 00092 * On entry, the N-by-N matrix A. 00093 * On exit, A is overwritten by its real Schur form T. 00094 * 00095 * LDA (input) INTEGER 00096 * The leading dimension of the array A. LDA >= max(1,N). 00097 * 00098 * SDIM (output) INTEGER 00099 * If SORT = 'N', SDIM = 0. 00100 * If SORT = 'S', SDIM = number of eigenvalues (after sorting) 00101 * for which SELECT is true. (Complex conjugate 00102 * pairs for which SELECT is true for either 00103 * eigenvalue count as 2.) 00104 * 00105 * WR (output) DOUBLE PRECISION array, dimension (N) 00106 * WI (output) DOUBLE PRECISION array, dimension (N) 00107 * WR and WI contain the real and imaginary parts, respectively, 00108 * of the computed eigenvalues, in the same order that they 00109 * appear on the diagonal of the output Schur form T. Complex 00110 * conjugate pairs of eigenvalues appear consecutively with the 00111 * eigenvalue having the positive imaginary part first. 00112 * 00113 * VS (output) DOUBLE PRECISION array, dimension (LDVS,N) 00114 * If JOBVS = 'V', VS contains the orthogonal matrix Z of Schur 00115 * vectors. 00116 * If JOBVS = 'N', VS is not referenced. 00117 * 00118 * LDVS (input) INTEGER 00119 * The leading dimension of the array VS. LDVS >= 1, and if 00120 * JOBVS = 'V', LDVS >= N. 00121 * 00122 * RCONDE (output) DOUBLE PRECISION 00123 * If SENSE = 'E' or 'B', RCONDE contains the reciprocal 00124 * condition number for the average of the selected eigenvalues. 00125 * Not referenced if SENSE = 'N' or 'V'. 00126 * 00127 * RCONDV (output) DOUBLE PRECISION 00128 * If SENSE = 'V' or 'B', RCONDV contains the reciprocal 00129 * condition number for the selected right invariant subspace. 00130 * Not referenced if SENSE = 'N' or 'E'. 00131 * 00132 * WORK (workspace/output) DOUBLE PRECISION array, dimension (MAX(1,LWORK)) 00133 * On exit, if INFO = 0, WORK(1) returns the optimal LWORK. 00134 * 00135 * LWORK (input) INTEGER 00136 * The dimension of the array WORK. LWORK >= max(1,3*N). 00137 * Also, if SENSE = 'E' or 'V' or 'B', 00138 * LWORK >= N+2*SDIM*(N-SDIM), where SDIM is the number of 00139 * selected eigenvalues computed by this routine. Note that 00140 * N+2*SDIM*(N-SDIM) <= N+N*N/2. Note also that an error is only 00141 * returned if LWORK < max(1,3*N), but if SENSE = 'E' or 'V' or 00142 * 'B' this may not be large enough. 00143 * For good performance, LWORK must generally be larger. 00144 * 00145 * If LWORK = -1, then a workspace query is assumed; the routine 00146 * only calculates upper bounds on the optimal sizes of the 00147 * arrays WORK and IWORK, returns these values as the first 00148 * entries of the WORK and IWORK arrays, and no error messages 00149 * related to LWORK or LIWORK are issued by XERBLA. 00150 * 00151 * IWORK (workspace/output) INTEGER array, dimension (MAX(1,LIWORK)) 00152 * On exit, if INFO = 0, IWORK(1) returns the optimal LIWORK. 00153 * 00154 * LIWORK (input) INTEGER 00155 * The dimension of the array IWORK. 00156 * LIWORK >= 1; if SENSE = 'V' or 'B', LIWORK >= SDIM*(N-SDIM). 00157 * Note that SDIM*(N-SDIM) <= N*N/4. Note also that an error is 00158 * only returned if LIWORK < 1, but if SENSE = 'V' or 'B' this 00159 * may not be large enough. 00160 * 00161 * If LIWORK = -1, then a workspace query is assumed; the 00162 * routine only calculates upper bounds on the optimal sizes of 00163 * the arrays WORK and IWORK, returns these values as the first 00164 * entries of the WORK and IWORK arrays, and no error messages 00165 * related to LWORK or LIWORK are issued by XERBLA. 00166 * 00167 * BWORK (workspace) LOGICAL array, dimension (N) 00168 * Not referenced if SORT = 'N'. 00169 * 00170 * INFO (output) INTEGER 00171 * = 0: successful exit 00172 * < 0: if INFO = -i, the i-th argument had an illegal value. 00173 * > 0: if INFO = i, and i is 00174 * <= N: the QR algorithm failed to compute all the 00175 * eigenvalues; elements 1:ILO-1 and i+1:N of WR and WI 00176 * contain those eigenvalues which have converged; if 00177 * JOBVS = 'V', VS contains the transformation which 00178 * reduces A to its partially converged Schur form. 00179 * = N+1: the eigenvalues could not be reordered because some 00180 * eigenvalues were too close to separate (the problem 00181 * is very ill-conditioned); 00182 * = N+2: after reordering, roundoff changed values of some 00183 * complex eigenvalues so that leading eigenvalues in 00184 * the Schur form no longer satisfy SELECT=.TRUE. This 00185 * could also be caused by underflow due to scaling. 00186 * 00187 * ===================================================================== 00188 * 00189 * .. Parameters .. 00190 DOUBLE PRECISION ZERO, ONE 00191 PARAMETER ( ZERO = 0.0D0, ONE = 1.0D0 ) 00192 * .. 00193 * .. Local Scalars .. 00194 LOGICAL CURSL, LASTSL, LQUERY, LST2SL, SCALEA, WANTSB, 00195 $ WANTSE, WANTSN, WANTST, WANTSV, WANTVS 00196 INTEGER HSWORK, I, I1, I2, IBAL, ICOND, IERR, IEVAL, 00197 $ IHI, ILO, INXT, IP, ITAU, IWRK, LIWRK, LWRK, 00198 $ MAXWRK, MINWRK 00199 DOUBLE PRECISION ANRM, BIGNUM, CSCALE, EPS, SMLNUM 00200 * .. 00201 * .. Local Arrays .. 00202 DOUBLE PRECISION DUM( 1 ) 00203 * .. 00204 * .. External Subroutines .. 00205 EXTERNAL DCOPY, DGEBAK, DGEBAL, DGEHRD, DHSEQR, DLACPY, 00206 $ DLASCL, DORGHR, DSWAP, DTRSEN, XERBLA 00207 * .. 00208 * .. External Functions .. 00209 LOGICAL LSAME 00210 INTEGER ILAENV 00211 DOUBLE PRECISION DLAMCH, DLANGE 00212 EXTERNAL LSAME, ILAENV, DLABAD, DLAMCH, DLANGE 00213 * .. 00214 * .. Intrinsic Functions .. 00215 INTRINSIC MAX, SQRT 00216 * .. 00217 * .. Executable Statements .. 00218 * 00219 * Test the input arguments 00220 * 00221 INFO = 0 00222 WANTVS = LSAME( JOBVS, 'V' ) 00223 WANTST = LSAME( SORT, 'S' ) 00224 WANTSN = LSAME( SENSE, 'N' ) 00225 WANTSE = LSAME( SENSE, 'E' ) 00226 WANTSV = LSAME( SENSE, 'V' ) 00227 WANTSB = LSAME( SENSE, 'B' ) 00228 LQUERY = ( LWORK.EQ.-1 .OR. LIWORK.EQ.-1 ) 00229 * 00230 IF( ( .NOT.WANTVS ) .AND. ( .NOT.LSAME( JOBVS, 'N' ) ) ) THEN 00231 INFO = -1 00232 ELSE IF( ( .NOT.WANTST ) .AND. ( .NOT.LSAME( SORT, 'N' ) ) ) THEN 00233 INFO = -2 00234 ELSE IF( .NOT.( WANTSN .OR. WANTSE .OR. WANTSV .OR. WANTSB ) .OR. 00235 $ ( .NOT.WANTST .AND. .NOT.WANTSN ) ) THEN 00236 INFO = -4 00237 ELSE IF( N.LT.0 ) THEN 00238 INFO = -5 00239 ELSE IF( LDA.LT.MAX( 1, N ) ) THEN 00240 INFO = -7 00241 ELSE IF( LDVS.LT.1 .OR. ( WANTVS .AND. LDVS.LT.N ) ) THEN 00242 INFO = -12 00243 END IF 00244 * 00245 * Compute workspace 00246 * (Note: Comments in the code beginning "RWorkspace:" describe the 00247 * minimal amount of real workspace needed at that point in the 00248 * code, as well as the preferred amount for good performance. 00249 * IWorkspace refers to integer workspace. 00250 * NB refers to the optimal block size for the immediately 00251 * following subroutine, as returned by ILAENV. 00252 * HSWORK refers to the workspace preferred by DHSEQR, as 00253 * calculated below. HSWORK is computed assuming ILO=1 and IHI=N, 00254 * the worst case. 00255 * If SENSE = 'E', 'V' or 'B', then the amount of workspace needed 00256 * depends on SDIM, which is computed by the routine DTRSEN later 00257 * in the code.) 00258 * 00259 IF( INFO.EQ.0 ) THEN 00260 LIWRK = 1 00261 IF( N.EQ.0 ) THEN 00262 MINWRK = 1 00263 LWRK = 1 00264 ELSE 00265 MAXWRK = 2*N + N*ILAENV( 1, 'DGEHRD', ' ', N, 1, N, 0 ) 00266 MINWRK = 3*N 00267 * 00268 CALL DHSEQR( 'S', JOBVS, N, 1, N, A, LDA, WR, WI, VS, LDVS, 00269 $ WORK, -1, IEVAL ) 00270 HSWORK = WORK( 1 ) 00271 * 00272 IF( .NOT.WANTVS ) THEN 00273 MAXWRK = MAX( MAXWRK, N + HSWORK ) 00274 ELSE 00275 MAXWRK = MAX( MAXWRK, 2*N + ( N - 1 )*ILAENV( 1, 00276 $ 'DORGHR', ' ', N, 1, N, -1 ) ) 00277 MAXWRK = MAX( MAXWRK, N + HSWORK ) 00278 END IF 00279 LWRK = MAXWRK 00280 IF( .NOT.WANTSN ) 00281 $ LWRK = MAX( LWRK, N + ( N*N )/2 ) 00282 IF( WANTSV .OR. WANTSB ) 00283 $ LIWRK = ( N*N )/4 00284 END IF 00285 IWORK( 1 ) = LIWRK 00286 WORK( 1 ) = LWRK 00287 * 00288 IF( LWORK.LT.MINWRK .AND. .NOT.LQUERY ) THEN 00289 INFO = -16 00290 ELSE IF( LIWORK.LT.1 .AND. .NOT.LQUERY ) THEN 00291 INFO = -18 00292 END IF 00293 END IF 00294 * 00295 IF( INFO.NE.0 ) THEN 00296 CALL XERBLA( 'DGEESX', -INFO ) 00297 RETURN 00298 ELSE IF( LQUERY ) THEN 00299 RETURN 00300 END IF 00301 * 00302 * Quick return if possible 00303 * 00304 IF( N.EQ.0 ) THEN 00305 SDIM = 0 00306 RETURN 00307 END IF 00308 * 00309 * Get machine constants 00310 * 00311 EPS = DLAMCH( 'P' ) 00312 SMLNUM = DLAMCH( 'S' ) 00313 BIGNUM = ONE / SMLNUM 00314 CALL DLABAD( SMLNUM, BIGNUM ) 00315 SMLNUM = SQRT( SMLNUM ) / EPS 00316 BIGNUM = ONE / SMLNUM 00317 * 00318 * Scale A if max element outside range [SMLNUM,BIGNUM] 00319 * 00320 ANRM = DLANGE( 'M', N, N, A, LDA, DUM ) 00321 SCALEA = .FALSE. 00322 IF( ANRM.GT.ZERO .AND. ANRM.LT.SMLNUM ) THEN 00323 SCALEA = .TRUE. 00324 CSCALE = SMLNUM 00325 ELSE IF( ANRM.GT.BIGNUM ) THEN 00326 SCALEA = .TRUE. 00327 CSCALE = BIGNUM 00328 END IF 00329 IF( SCALEA ) 00330 $ CALL DLASCL( 'G', 0, 0, ANRM, CSCALE, N, N, A, LDA, IERR ) 00331 * 00332 * Permute the matrix to make it more nearly triangular 00333 * (RWorkspace: need N) 00334 * 00335 IBAL = 1 00336 CALL DGEBAL( 'P', N, A, LDA, ILO, IHI, WORK( IBAL ), IERR ) 00337 * 00338 * Reduce to upper Hessenberg form 00339 * (RWorkspace: need 3*N, prefer 2*N+N*NB) 00340 * 00341 ITAU = N + IBAL 00342 IWRK = N + ITAU 00343 CALL DGEHRD( N, ILO, IHI, A, LDA, WORK( ITAU ), WORK( IWRK ), 00344 $ LWORK-IWRK+1, IERR ) 00345 * 00346 IF( WANTVS ) THEN 00347 * 00348 * Copy Householder vectors to VS 00349 * 00350 CALL DLACPY( 'L', N, N, A, LDA, VS, LDVS ) 00351 * 00352 * Generate orthogonal matrix in VS 00353 * (RWorkspace: need 3*N-1, prefer 2*N+(N-1)*NB) 00354 * 00355 CALL DORGHR( N, ILO, IHI, VS, LDVS, WORK( ITAU ), WORK( IWRK ), 00356 $ LWORK-IWRK+1, IERR ) 00357 END IF 00358 * 00359 SDIM = 0 00360 * 00361 * Perform QR iteration, accumulating Schur vectors in VS if desired 00362 * (RWorkspace: need N+1, prefer N+HSWORK (see comments) ) 00363 * 00364 IWRK = ITAU 00365 CALL DHSEQR( 'S', JOBVS, N, ILO, IHI, A, LDA, WR, WI, VS, LDVS, 00366 $ WORK( IWRK ), LWORK-IWRK+1, IEVAL ) 00367 IF( IEVAL.GT.0 ) 00368 $ INFO = IEVAL 00369 * 00370 * Sort eigenvalues if desired 00371 * 00372 IF( WANTST .AND. INFO.EQ.0 ) THEN 00373 IF( SCALEA ) THEN 00374 CALL DLASCL( 'G', 0, 0, CSCALE, ANRM, N, 1, WR, N, IERR ) 00375 CALL DLASCL( 'G', 0, 0, CSCALE, ANRM, N, 1, WI, N, IERR ) 00376 END IF 00377 DO 10 I = 1, N 00378 BWORK( I ) = SELECT( WR( I ), WI( I ) ) 00379 10 CONTINUE 00380 * 00381 * Reorder eigenvalues, transform Schur vectors, and compute 00382 * reciprocal condition numbers 00383 * (RWorkspace: if SENSE is not 'N', need N+2*SDIM*(N-SDIM) 00384 * otherwise, need N ) 00385 * (IWorkspace: if SENSE is 'V' or 'B', need SDIM*(N-SDIM) 00386 * otherwise, need 0 ) 00387 * 00388 CALL DTRSEN( SENSE, JOBVS, BWORK, N, A, LDA, VS, LDVS, WR, WI, 00389 $ SDIM, RCONDE, RCONDV, WORK( IWRK ), LWORK-IWRK+1, 00390 $ IWORK, LIWORK, ICOND ) 00391 IF( .NOT.WANTSN ) 00392 $ MAXWRK = MAX( MAXWRK, N+2*SDIM*( N-SDIM ) ) 00393 IF( ICOND.EQ.-15 ) THEN 00394 * 00395 * Not enough real workspace 00396 * 00397 INFO = -16 00398 ELSE IF( ICOND.EQ.-17 ) THEN 00399 * 00400 * Not enough integer workspace 00401 * 00402 INFO = -18 00403 ELSE IF( ICOND.GT.0 ) THEN 00404 * 00405 * DTRSEN failed to reorder or to restore standard Schur form 00406 * 00407 INFO = ICOND + N 00408 END IF 00409 END IF 00410 * 00411 IF( WANTVS ) THEN 00412 * 00413 * Undo balancing 00414 * (RWorkspace: need N) 00415 * 00416 CALL DGEBAK( 'P', 'R', N, ILO, IHI, WORK( IBAL ), N, VS, LDVS, 00417 $ IERR ) 00418 END IF 00419 * 00420 IF( SCALEA ) THEN 00421 * 00422 * Undo scaling for the Schur form of A 00423 * 00424 CALL DLASCL( 'H', 0, 0, CSCALE, ANRM, N, N, A, LDA, IERR ) 00425 CALL DCOPY( N, A, LDA+1, WR, 1 ) 00426 IF( ( WANTSV .OR. WANTSB ) .AND. INFO.EQ.0 ) THEN 00427 DUM( 1 ) = RCONDV 00428 CALL DLASCL( 'G', 0, 0, CSCALE, ANRM, 1, 1, DUM, 1, IERR ) 00429 RCONDV = DUM( 1 ) 00430 END IF 00431 IF( CSCALE.EQ.SMLNUM ) THEN 00432 * 00433 * If scaling back towards underflow, adjust WI if an 00434 * offdiagonal element of a 2-by-2 block in the Schur form 00435 * underflows. 00436 * 00437 IF( IEVAL.GT.0 ) THEN 00438 I1 = IEVAL + 1 00439 I2 = IHI - 1 00440 CALL DLASCL( 'G', 0, 0, CSCALE, ANRM, ILO-1, 1, WI, N, 00441 $ IERR ) 00442 ELSE IF( WANTST ) THEN 00443 I1 = 1 00444 I2 = N - 1 00445 ELSE 00446 I1 = ILO 00447 I2 = IHI - 1 00448 END IF 00449 INXT = I1 - 1 00450 DO 20 I = I1, I2 00451 IF( I.LT.INXT ) 00452 $ GO TO 20 00453 IF( WI( I ).EQ.ZERO ) THEN 00454 INXT = I + 1 00455 ELSE 00456 IF( A( I+1, I ).EQ.ZERO ) THEN 00457 WI( I ) = ZERO 00458 WI( I+1 ) = ZERO 00459 ELSE IF( A( I+1, I ).NE.ZERO .AND. A( I, I+1 ).EQ. 00460 $ ZERO ) THEN 00461 WI( I ) = ZERO 00462 WI( I+1 ) = ZERO 00463 IF( I.GT.1 ) 00464 $ CALL DSWAP( I-1, A( 1, I ), 1, A( 1, I+1 ), 1 ) 00465 IF( N.GT.I+1 ) 00466 $ CALL DSWAP( N-I-1, A( I, I+2 ), LDA, 00467 $ A( I+1, I+2 ), LDA ) 00468 CALL DSWAP( N, VS( 1, I ), 1, VS( 1, I+1 ), 1 ) 00469 A( I, I+1 ) = A( I+1, I ) 00470 A( I+1, I ) = ZERO 00471 END IF 00472 INXT = I + 2 00473 END IF 00474 20 CONTINUE 00475 END IF 00476 CALL DLASCL( 'G', 0, 0, CSCALE, ANRM, N-IEVAL, 1, 00477 $ WI( IEVAL+1 ), MAX( N-IEVAL, 1 ), IERR ) 00478 END IF 00479 * 00480 IF( WANTST .AND. INFO.EQ.0 ) THEN 00481 * 00482 * Check if reordering successful 00483 * 00484 LASTSL = .TRUE. 00485 LST2SL = .TRUE. 00486 SDIM = 0 00487 IP = 0 00488 DO 30 I = 1, N 00489 CURSL = SELECT( WR( I ), WI( I ) ) 00490 IF( WI( I ).EQ.ZERO ) THEN 00491 IF( CURSL ) 00492 $ SDIM = SDIM + 1 00493 IP = 0 00494 IF( CURSL .AND. .NOT.LASTSL ) 00495 $ INFO = N + 2 00496 ELSE 00497 IF( IP.EQ.1 ) THEN 00498 * 00499 * Last eigenvalue of conjugate pair 00500 * 00501 CURSL = CURSL .OR. LASTSL 00502 LASTSL = CURSL 00503 IF( CURSL ) 00504 $ SDIM = SDIM + 2 00505 IP = -1 00506 IF( CURSL .AND. .NOT.LST2SL ) 00507 $ INFO = N + 2 00508 ELSE 00509 * 00510 * First eigenvalue of conjugate pair 00511 * 00512 IP = 1 00513 END IF 00514 END IF 00515 LST2SL = LASTSL 00516 LASTSL = CURSL 00517 30 CONTINUE 00518 END IF 00519 * 00520 WORK( 1 ) = MAXWRK 00521 IF( WANTSV .OR. WANTSB ) THEN 00522 IWORK( 1 ) = MAX( 1, SDIM*( N-SDIM ) ) 00523 ELSE 00524 IWORK( 1 ) = 1 00525 END IF 00526 * 00527 RETURN 00528 * 00529 * End of DGEESX 00530 * 00531 END
1.7.3 
