| 1 |
SUBROUTINE ZTRSEN( JOB, COMPQ, SELECT, N, T, LDT, Q, LDQ, W, M, S, |
| 2 |
$ SEP, WORK, LWORK, INFO ) |
| 3 |
* |
| 4 |
* -- LAPACK 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 |
* Modified to call ZLACN2 in place of ZLACON, 10 Feb 03, SJH. |
| 10 |
* |
| 11 |
* .. Scalar Arguments .. |
| 12 |
CHARACTER COMPQ, JOB |
| 13 |
INTEGER INFO, LDQ, LDT, LWORK, M, N |
| 14 |
DOUBLE PRECISION S, SEP |
| 15 |
* .. |
| 16 |
* .. Array Arguments .. |
| 17 |
LOGICAL SELECT( * ) |
| 18 |
COMPLEX*16 Q( LDQ, * ), T( LDT, * ), W( * ), WORK( * ) |
| 19 |
* .. |
| 20 |
* |
| 21 |
* Purpose |
| 22 |
* ======= |
| 23 |
* |
| 24 |
* ZTRSEN reorders the Schur factorization of a complex matrix |
| 25 |
* A = Q*T*Q**H, so that a selected cluster of eigenvalues appears in |
| 26 |
* the leading positions on the diagonal of the upper triangular matrix |
| 27 |
* T, and the leading columns of Q form an orthonormal basis of the |
| 28 |
* corresponding right invariant subspace. |
| 29 |
* |
| 30 |
* Optionally the routine computes the reciprocal condition numbers of |
| 31 |
* the cluster of eigenvalues and/or the invariant subspace. |
| 32 |
* |
| 33 |
* Arguments |
| 34 |
* ========= |
| 35 |
* |
| 36 |
* JOB (input) CHARACTER*1 |
| 37 |
* Specifies whether condition numbers are required for the |
| 38 |
* cluster of eigenvalues (S) or the invariant subspace (SEP): |
| 39 |
* = 'N': none; |
| 40 |
* = 'E': for eigenvalues only (S); |
| 41 |
* = 'V': for invariant subspace only (SEP); |
| 42 |
* = 'B': for both eigenvalues and invariant subspace (S and |
| 43 |
* SEP). |
| 44 |
* |
| 45 |
* COMPQ (input) CHARACTER*1 |
| 46 |
* = 'V': update the matrix Q of Schur vectors; |
| 47 |
* = 'N': do not update Q. |
| 48 |
* |
| 49 |
* SELECT (input) LOGICAL array, dimension (N) |
| 50 |
* SELECT specifies the eigenvalues in the selected cluster. To |
| 51 |
* select the j-th eigenvalue, SELECT(j) must be set to .TRUE.. |
| 52 |
* |
| 53 |
* N (input) INTEGER |
| 54 |
* The order of the matrix T. N >= 0. |
| 55 |
* |
| 56 |
* T (input/output) COMPLEX*16 array, dimension (LDT,N) |
| 57 |
* On entry, the upper triangular matrix T. |
| 58 |
* On exit, T is overwritten by the reordered matrix T, with the |
| 59 |
* selected eigenvalues as the leading diagonal elements. |
| 60 |
* |
| 61 |
* LDT (input) INTEGER |
| 62 |
* The leading dimension of the array T. LDT >= max(1,N). |
| 63 |
* |
| 64 |
* Q (input/output) COMPLEX*16 array, dimension (LDQ,N) |
| 65 |
* On entry, if COMPQ = 'V', the matrix Q of Schur vectors. |
| 66 |
* On exit, if COMPQ = 'V', Q has been postmultiplied by the |
| 67 |
* unitary transformation matrix which reorders T; the leading M |
| 68 |
* columns of Q form an orthonormal basis for the specified |
| 69 |
* invariant subspace. |
| 70 |
* If COMPQ = 'N', Q is not referenced. |
| 71 |
* |
| 72 |
* LDQ (input) INTEGER |
| 73 |
* The leading dimension of the array Q. |
| 74 |
* LDQ >= 1; and if COMPQ = 'V', LDQ >= N. |
| 75 |
* |
| 76 |
* W (output) COMPLEX*16 array, dimension (N) |
| 77 |
* The reordered eigenvalues of T, in the same order as they |
| 78 |
* appear on the diagonal of T. |
| 79 |
* |
| 80 |
* M (output) INTEGER |
| 81 |
* The dimension of the specified invariant subspace. |
| 82 |
* 0 <= M <= N. |
| 83 |
* |
| 84 |
* S (output) DOUBLE PRECISION |
| 85 |
* If JOB = 'E' or 'B', S is a lower bound on the reciprocal |
| 86 |
* condition number for the selected cluster of eigenvalues. |
| 87 |
* S cannot underestimate the true reciprocal condition number |
| 88 |
* by more than a factor of sqrt(N). If M = 0 or N, S = 1. |
| 89 |
* If JOB = 'N' or 'V', S is not referenced. |
| 90 |
* |
| 91 |
* SEP (output) DOUBLE PRECISION |
| 92 |
* If JOB = 'V' or 'B', SEP is the estimated reciprocal |
| 93 |
* condition number of the specified invariant subspace. If |
| 94 |
* M = 0 or N, SEP = norm(T). |
| 95 |
* If JOB = 'N' or 'E', SEP is not referenced. |
| 96 |
* |
| 97 |
* WORK (workspace/output) COMPLEX*16 array, dimension (MAX(1,LWORK)) |
| 98 |
* On exit, if INFO = 0, WORK(1) returns the optimal LWORK. |
| 99 |
* |
| 100 |
* LWORK (input) INTEGER |
| 101 |
* The dimension of the array WORK. |
| 102 |
* If JOB = 'N', LWORK >= 1; |
| 103 |
* if JOB = 'E', LWORK = max(1,M*(N-M)); |
| 104 |
* if JOB = 'V' or 'B', LWORK >= max(1,2*M*(N-M)). |
| 105 |
* |
| 106 |
* If LWORK = -1, then a workspace query is assumed; the routine |
| 107 |
* only calculates the optimal size of the WORK array, returns |
| 108 |
* this value as the first entry of the WORK array, and no error |
| 109 |
* message related to LWORK is issued by XERBLA. |
| 110 |
* |
| 111 |
* INFO (output) INTEGER |
| 112 |
* = 0: successful exit |
| 113 |
* < 0: if INFO = -i, the i-th argument had an illegal value |
| 114 |
* |
| 115 |
* Further Details |
| 116 |
* =============== |
| 117 |
* |
| 118 |
* ZTRSEN first collects the selected eigenvalues by computing a unitary |
| 119 |
* transformation Z to move them to the top left corner of T. In other |
| 120 |
* words, the selected eigenvalues are the eigenvalues of T11 in: |
| 121 |
* |
| 122 |
* Z'*T*Z = ( T11 T12 ) n1 |
| 123 |
* ( 0 T22 ) n2 |
| 124 |
* n1 n2 |
| 125 |
* |
| 126 |
* where N = n1+n2 and Z' means the conjugate transpose of Z. The first |
| 127 |
* n1 columns of Z span the specified invariant subspace of T. |
| 128 |
* |
| 129 |
* If T has been obtained from the Schur factorization of a matrix |
| 130 |
* A = Q*T*Q', then the reordered Schur factorization of A is given by |
| 131 |
* A = (Q*Z)*(Z'*T*Z)*(Q*Z)', and the first n1 columns of Q*Z span the |
| 132 |
* corresponding invariant subspace of A. |
| 133 |
* |
| 134 |
* The reciprocal condition number of the average of the eigenvalues of |
| 135 |
* T11 may be returned in S. S lies between 0 (very badly conditioned) |
| 136 |
* and 1 (very well conditioned). It is computed as follows. First we |
| 137 |
* compute R so that |
| 138 |
* |
| 139 |
* P = ( I R ) n1 |
| 140 |
* ( 0 0 ) n2 |
| 141 |
* n1 n2 |
| 142 |
* |
| 143 |
* is the projector on the invariant subspace associated with T11. |
| 144 |
* R is the solution of the Sylvester equation: |
| 145 |
* |
| 146 |
* T11*R - R*T22 = T12. |
| 147 |
* |
| 148 |
* Let F-norm(M) denote the Frobenius-norm of M and 2-norm(M) denote |
| 149 |
* the two-norm of M. Then S is computed as the lower bound |
| 150 |
* |
| 151 |
* (1 + F-norm(R)**2)**(-1/2) |
| 152 |
* |
| 153 |
* on the reciprocal of 2-norm(P), the true reciprocal condition number. |
| 154 |
* S cannot underestimate 1 / 2-norm(P) by more than a factor of |
| 155 |
* sqrt(N). |
| 156 |
* |
| 157 |
* An approximate error bound for the computed average of the |
| 158 |
* eigenvalues of T11 is |
| 159 |
* |
| 160 |
* EPS * norm(T) / S |
| 161 |
* |
| 162 |
* where EPS is the machine precision. |
| 163 |
* |
| 164 |
* The reciprocal condition number of the right invariant subspace |
| 165 |
* spanned by the first n1 columns of Z (or of Q*Z) is returned in SEP. |
| 166 |
* SEP is defined as the separation of T11 and T22: |
| 167 |
* |
| 168 |
* sep( T11, T22 ) = sigma-min( C ) |
| 169 |
* |
| 170 |
* where sigma-min(C) is the smallest singular value of the |
| 171 |
* n1*n2-by-n1*n2 matrix |
| 172 |
* |
| 173 |
* C = kprod( I(n2), T11 ) - kprod( transpose(T22), I(n1) ) |
| 174 |
* |
| 175 |
* I(m) is an m by m identity matrix, and kprod denotes the Kronecker |
| 176 |
* product. We estimate sigma-min(C) by the reciprocal of an estimate of |
| 177 |
* the 1-norm of inverse(C). The true reciprocal 1-norm of inverse(C) |
| 178 |
* cannot differ from sigma-min(C) by more than a factor of sqrt(n1*n2). |
| 179 |
* |
| 180 |
* When SEP is small, small changes in T can cause large changes in |
| 181 |
* the invariant subspace. An approximate bound on the maximum angular |
| 182 |
* error in the computed right invariant subspace is |
| 183 |
* |
| 184 |
* EPS * norm(T) / SEP |
| 185 |
* |
| 186 |
* ===================================================================== |
| 187 |
* |
| 188 |
* .. Parameters .. |
| 189 |
DOUBLE PRECISION ZERO, ONE |
| 190 |
PARAMETER ( ZERO = 0.0D+0, ONE = 1.0D+0 ) |
| 191 |
* .. |
| 192 |
* .. Local Scalars .. |
| 193 |
LOGICAL LQUERY, WANTBH, WANTQ, WANTS, WANTSP |
| 194 |
INTEGER IERR, K, KASE, KS, LWMIN, N1, N2, NN |
| 195 |
DOUBLE PRECISION EST, RNORM, SCALE |
| 196 |
* .. |
| 197 |
* .. Local Arrays .. |
| 198 |
INTEGER ISAVE( 3 ) |
| 199 |
DOUBLE PRECISION RWORK( 1 ) |
| 200 |
* .. |
| 201 |
* .. External Functions .. |
| 202 |
LOGICAL LSAME |
| 203 |
DOUBLE PRECISION ZLANGE |
| 204 |
EXTERNAL LSAME, ZLANGE |
| 205 |
* .. |
| 206 |
* .. External Subroutines .. |
| 207 |
EXTERNAL XERBLA, ZLACN2, ZLACPY, ZTREXC, ZTRSYL |
| 208 |
* .. |
| 209 |
* .. Intrinsic Functions .. |
| 210 |
INTRINSIC MAX, SQRT |
| 211 |
* .. |
| 212 |
* .. Executable Statements .. |
| 213 |
* |
| 214 |
* Decode and test the input parameters. |
| 215 |
* |
| 216 |
WANTBH = LSAME( JOB, 'B' ) |
| 217 |
WANTS = LSAME( JOB, 'E' ) .OR. WANTBH |
| 218 |
WANTSP = LSAME( JOB, 'V' ) .OR. WANTBH |
| 219 |
WANTQ = LSAME( COMPQ, 'V' ) |
| 220 |
* |
| 221 |
* Set M to the number of selected eigenvalues. |
| 222 |
* |
| 223 |
M = 0 |
| 224 |
DO 10 K = 1, N |
| 225 |
IF( SELECT( K ) ) |
| 226 |
$ M = M + 1 |
| 227 |
10 CONTINUE |
| 228 |
* |
| 229 |
N1 = M |
| 230 |
N2 = N - M |
| 231 |
NN = N1*N2 |
| 232 |
* |
| 233 |
INFO = 0 |
| 234 |
LQUERY = ( LWORK.EQ.-1 ) |
| 235 |
* |
| 236 |
IF( WANTSP ) THEN |
| 237 |
LWMIN = MAX( 1, 2*NN ) |
| 238 |
ELSE IF( LSAME( JOB, 'N' ) ) THEN |
| 239 |
LWMIN = 1 |
| 240 |
ELSE IF( LSAME( JOB, 'E' ) ) THEN |
| 241 |
LWMIN = MAX( 1, NN ) |
| 242 |
END IF |
| 243 |
* |
| 244 |
IF( .NOT.LSAME( JOB, 'N' ) .AND. .NOT.WANTS .AND. .NOT.WANTSP ) |
| 245 |
$ THEN |
| 246 |
INFO = -1 |
| 247 |
ELSE IF( .NOT.LSAME( COMPQ, 'N' ) .AND. .NOT.WANTQ ) THEN |
| 248 |
INFO = -2 |
| 249 |
ELSE IF( N.LT.0 ) THEN |
| 250 |
INFO = -4 |
| 251 |
ELSE IF( LDT.LT.MAX( 1, N ) ) THEN |
| 252 |
INFO = -6 |
| 253 |
ELSE IF( LDQ.LT.1 .OR. ( WANTQ .AND. LDQ.LT.N ) ) THEN |
| 254 |
INFO = -8 |
| 255 |
ELSE IF( LWORK.LT.LWMIN .AND. .NOT.LQUERY ) THEN |
| 256 |
INFO = -14 |
| 257 |
END IF |
| 258 |
* |
| 259 |
IF( INFO.EQ.0 ) THEN |
| 260 |
WORK( 1 ) = LWMIN |
| 261 |
END IF |
| 262 |
* |
| 263 |
IF( INFO.NE.0 ) THEN |
| 264 |
CALL XERBLA( 'ZTRSEN', -INFO ) |
| 265 |
RETURN |
| 266 |
ELSE IF( LQUERY ) THEN |
| 267 |
RETURN |
| 268 |
END IF |
| 269 |
* |
| 270 |
* Quick return if possible |
| 271 |
* |
| 272 |
IF( M.EQ.N .OR. M.EQ.0 ) THEN |
| 273 |
IF( WANTS ) |
| 274 |
$ S = ONE |
| 275 |
IF( WANTSP ) |
| 276 |
$ SEP = ZLANGE( '1', N, N, T, LDT, RWORK ) |
| 277 |
GO TO 40 |
| 278 |
END IF |
| 279 |
* |
| 280 |
* Collect the selected eigenvalues at the top left corner of T. |
| 281 |
* |
| 282 |
KS = 0 |
| 283 |
DO 20 K = 1, N |
| 284 |
IF( SELECT( K ) ) THEN |
| 285 |
KS = KS + 1 |
| 286 |
* |
| 287 |
* Swap the K-th eigenvalue to position KS. |
| 288 |
* |
| 289 |
IF( K.NE.KS ) |
| 290 |
$ CALL ZTREXC( COMPQ, N, T, LDT, Q, LDQ, K, KS, IERR ) |
| 291 |
END IF |
| 292 |
20 CONTINUE |
| 293 |
* |
| 294 |
IF( WANTS ) THEN |
| 295 |
* |
| 296 |
* Solve the Sylvester equation for R: |
| 297 |
* |
| 298 |
* T11*R - R*T22 = scale*T12 |
| 299 |
* |
| 300 |
CALL ZLACPY( 'F', N1, N2, T( 1, N1+1 ), LDT, WORK, N1 ) |
| 301 |
CALL ZTRSYL( 'N', 'N', -1, N1, N2, T, LDT, T( N1+1, N1+1 ), |
| 302 |
$ LDT, WORK, N1, SCALE, IERR ) |
| 303 |
* |
| 304 |
* Estimate the reciprocal of the condition number of the cluster |
| 305 |
* of eigenvalues. |
| 306 |
* |
| 307 |
RNORM = ZLANGE( 'F', N1, N2, WORK, N1, RWORK ) |
| 308 |
IF( RNORM.EQ.ZERO ) THEN |
| 309 |
S = ONE |
| 310 |
ELSE |
| 311 |
S = SCALE / ( SQRT( SCALE*SCALE / RNORM+RNORM )* |
| 312 |
$ SQRT( RNORM ) ) |
| 313 |
END IF |
| 314 |
END IF |
| 315 |
* |
| 316 |
IF( WANTSP ) THEN |
| 317 |
* |
| 318 |
* Estimate sep(T11,T22). |
| 319 |
* |
| 320 |
EST = ZERO |
| 321 |
KASE = 0 |
| 322 |
30 CONTINUE |
| 323 |
CALL ZLACN2( NN, WORK( NN+1 ), WORK, EST, KASE, ISAVE ) |
| 324 |
IF( KASE.NE.0 ) THEN |
| 325 |
IF( KASE.EQ.1 ) THEN |
| 326 |
* |
| 327 |
* Solve T11*R - R*T22 = scale*X. |
| 328 |
* |
| 329 |
CALL ZTRSYL( 'N', 'N', -1, N1, N2, T, LDT, |
| 330 |
$ T( N1+1, N1+1 ), LDT, WORK, N1, SCALE, |
| 331 |
$ IERR ) |
| 332 |
ELSE |
| 333 |
* |
| 334 |
* Solve T11'*R - R*T22' = scale*X. |
| 335 |
* |
| 336 |
CALL ZTRSYL( 'C', 'C', -1, N1, N2, T, LDT, |
| 337 |
$ T( N1+1, N1+1 ), LDT, WORK, N1, SCALE, |
| 338 |
$ IERR ) |
| 339 |
END IF |
| 340 |
GO TO 30 |
| 341 |
END IF |
| 342 |
* |
| 343 |
SEP = SCALE / EST |
| 344 |
END IF |
| 345 |
* |
| 346 |
40 CONTINUE |
| 347 |
* |
| 348 |
* Copy reordered eigenvalues to W. |
| 349 |
* |
| 350 |
DO 50 K = 1, N |
| 351 |
W( K ) = T( K, K ) |
| 352 |
50 CONTINUE |
| 353 |
* |
| 354 |
WORK( 1 ) = LWMIN |
| 355 |
* |
| 356 |
RETURN |
| 357 |
* |
| 358 |
* End of ZTRSEN |
| 359 |
* |
| 360 |
END |