 |
LAPACK 3.3.1
Linear Algebra PACKage
|
|
LAPACK 3.3.1
Linear Algebra PACKage
|
00001 RECURSIVE SUBROUTINE CUNCSD( JOBU1, JOBU2, JOBV1T, JOBV2T, TRANS, 00002 $ SIGNS, M, P, Q, X11, LDX11, X12, 00003 $ LDX12, X21, LDX21, X22, LDX22, THETA, 00004 $ U1, LDU1, U2, LDU2, V1T, LDV1T, V2T, 00005 $ LDV2T, WORK, LWORK, RWORK, LRWORK, 00006 $ IWORK, INFO ) 00007 IMPLICIT NONE 00008 * 00009 * -- LAPACK routine (version 3.3.1) -- 00010 * 00011 * -- Contributed by Brian Sutton of the Randolph-Macon College -- 00012 * -- November 2010 00013 * 00014 * -- LAPACK is a software package provided by Univ. of Tennessee, -- 00015 * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- 00016 * 00017 * @generated c 00018 * 00019 * .. Scalar Arguments .. 00020 CHARACTER JOBU1, JOBU2, JOBV1T, JOBV2T, SIGNS, TRANS 00021 INTEGER INFO, LDU1, LDU2, LDV1T, LDV2T, LDX11, LDX12, 00022 $ LDX21, LDX22, LRWORK, LWORK, M, P, Q 00023 * .. 00024 * .. Array Arguments .. 00025 INTEGER IWORK( * ) 00026 REAL THETA( * ) 00027 REAL RWORK( * ) 00028 COMPLEX U1( LDU1, * ), U2( LDU2, * ), V1T( LDV1T, * ), 00029 $ V2T( LDV2T, * ), WORK( * ), X11( LDX11, * ), 00030 $ X12( LDX12, * ), X21( LDX21, * ), X22( LDX22, 00031 $ * ) 00032 * .. 00033 * 00034 * Purpose 00035 * ======= 00036 * 00037 * CUNCSD computes the CS decomposition of an M-by-M partitioned 00038 * unitary matrix X: 00039 * 00040 * [ I 0 0 | 0 0 0 ] 00041 * [ 0 C 0 | 0 -S 0 ] 00042 * [ X11 | X12 ] [ U1 | ] [ 0 0 0 | 0 0 -I ] [ V1 | ]**H 00043 * X = [-----------] = [---------] [---------------------] [---------] . 00044 * [ X21 | X22 ] [ | U2 ] [ 0 0 0 | I 0 0 ] [ | V2 ] 00045 * [ 0 S 0 | 0 C 0 ] 00046 * [ 0 0 I | 0 0 0 ] 00047 * 00048 * X11 is P-by-Q. The unitary matrices U1, U2, V1, and V2 are P-by-P, 00049 * (M-P)-by-(M-P), Q-by-Q, and (M-Q)-by-(M-Q), respectively. C and S are 00050 * R-by-R nonnegative diagonal matrices satisfying C^2 + S^2 = I, in 00051 * which R = MIN(P,M-P,Q,M-Q). 00052 * 00053 * Arguments 00054 * ========= 00055 * 00056 * JOBU1 (input) CHARACTER 00057 * = 'Y': U1 is computed; 00058 * otherwise: U1 is not computed. 00059 * 00060 * JOBU2 (input) CHARACTER 00061 * = 'Y': U2 is computed; 00062 * otherwise: U2 is not computed. 00063 * 00064 * JOBV1T (input) CHARACTER 00065 * = 'Y': V1T is computed; 00066 * otherwise: V1T is not computed. 00067 * 00068 * JOBV2T (input) CHARACTER 00069 * = 'Y': V2T is computed; 00070 * otherwise: V2T is not computed. 00071 * 00072 * TRANS (input) CHARACTER 00073 * = 'T': X, U1, U2, V1T, and V2T are stored in row-major 00074 * order; 00075 * otherwise: X, U1, U2, V1T, and V2T are stored in column- 00076 * major order. 00077 * 00078 * SIGNS (input) CHARACTER 00079 * = 'O': The lower-left block is made nonpositive (the 00080 * "other" convention); 00081 * otherwise: The upper-right block is made nonpositive (the 00082 * "default" convention). 00083 * 00084 * M (input) INTEGER 00085 * The number of rows and columns in X. 00086 * 00087 * P (input) INTEGER 00088 * The number of rows in X11 and X12. 0 <= P <= M. 00089 * 00090 * Q (input) INTEGER 00091 * The number of columns in X11 and X21. 0 <= Q <= M. 00092 * 00093 * X (input/workspace) COMPLEX array, dimension (LDX,M) 00094 * On entry, the unitary matrix whose CSD is desired. 00095 * 00096 * LDX (input) INTEGER 00097 * The leading dimension of X. LDX >= MAX(1,M). 00098 * 00099 * THETA (output) REAL array, dimension (R), in which R = 00100 * MIN(P,M-P,Q,M-Q). 00101 * C = DIAG( COS(THETA(1)), ... , COS(THETA(R)) ) and 00102 * S = DIAG( SIN(THETA(1)), ... , SIN(THETA(R)) ). 00103 * 00104 * U1 (output) COMPLEX array, dimension (P) 00105 * If JOBU1 = 'Y', U1 contains the P-by-P unitary matrix U1. 00106 * 00107 * LDU1 (input) INTEGER 00108 * The leading dimension of U1. If JOBU1 = 'Y', LDU1 >= 00109 * MAX(1,P). 00110 * 00111 * U2 (output) COMPLEX array, dimension (M-P) 00112 * If JOBU2 = 'Y', U2 contains the (M-P)-by-(M-P) unitary 00113 * matrix U2. 00114 * 00115 * LDU2 (input) INTEGER 00116 * The leading dimension of U2. If JOBU2 = 'Y', LDU2 >= 00117 * MAX(1,M-P). 00118 * 00119 * V1T (output) COMPLEX array, dimension (Q) 00120 * If JOBV1T = 'Y', V1T contains the Q-by-Q matrix unitary 00121 * matrix V1**H. 00122 * 00123 * LDV1T (input) INTEGER 00124 * The leading dimension of V1T. If JOBV1T = 'Y', LDV1T >= 00125 * MAX(1,Q). 00126 * 00127 * V2T (output) COMPLEX array, dimension (M-Q) 00128 * If JOBV2T = 'Y', V2T contains the (M-Q)-by-(M-Q) unitary 00129 * matrix V2**H. 00130 * 00131 * LDV2T (input) INTEGER 00132 * The leading dimension of V2T. If JOBV2T = 'Y', LDV2T >= 00133 * MAX(1,M-Q). 00134 * 00135 * WORK (workspace) COMPLEX array, dimension (MAX(1,LWORK)) 00136 * On exit, if INFO = 0, WORK(1) returns the optimal LWORK. 00137 * 00138 * LWORK (input) INTEGER 00139 * The dimension of the array WORK. 00140 * 00141 * If LWORK = -1, then a workspace query is assumed; the routine 00142 * only calculates the optimal size of the WORK array, returns 00143 * this value as the first entry of the work array, and no error 00144 * message related to LWORK is issued by XERBLA. 00145 * 00146 * RWORK (workspace) REAL array, dimension MAX(1,LRWORK) 00147 * On exit, if INFO = 0, RWORK(1) returns the optimal LRWORK. 00148 * If INFO > 0 on exit, RWORK(2:R) contains the values PHI(1), 00149 * ..., PHI(R-1) that, together with THETA(1), ..., THETA(R), 00150 * define the matrix in intermediate bidiagonal-block form 00151 * remaining after nonconvergence. INFO specifies the number 00152 * of nonzero PHI's. 00153 * 00154 * LRWORK (input) INTEGER 00155 * The dimension of the array RWORK. 00156 * 00157 * If LRWORK = -1, then a workspace query is assumed; the routine 00158 * only calculates the optimal size of the RWORK array, returns 00159 * this value as the first entry of the work array, and no error 00160 * message related to LRWORK is issued by XERBLA. 00161 * 00162 * IWORK (workspace) INTEGER array, dimension (M-MIN(P,M-P,Q,M-Q)) 00163 * 00164 * INFO (output) INTEGER 00165 * = 0: successful exit. 00166 * < 0: if INFO = -i, the i-th argument had an illegal value. 00167 * > 0: CBBCSD did not converge. See the description of RWORK 00168 * above for details. 00169 * 00170 * Reference 00171 * ========= 00172 * 00173 * [1] Brian D. Sutton. Computing the complete CS decomposition. Numer. 00174 * Algorithms, 50(1):33-65, 2009. 00175 * 00176 * =================================================================== 00177 * 00178 * .. Parameters .. 00179 REAL REALONE 00180 PARAMETER ( REALONE = 1.0E0 ) 00181 COMPLEX NEGONE, ONE, PIOVER2, ZERO 00182 PARAMETER ( NEGONE = (-1.0E0,0.0E0), ONE = (1.0E0,0.0E0), 00183 $ PIOVER2 = 1.57079632679489662E0, 00184 $ ZERO = (0.0E0,0.0E0) ) 00185 * .. 00186 * .. Local Scalars .. 00187 CHARACTER TRANST, SIGNST 00188 INTEGER CHILDINFO, I, IB11D, IB11E, IB12D, IB12E, 00189 $ IB21D, IB21E, IB22D, IB22E, IBBCSD, IORBDB, 00190 $ IORGLQ, IORGQR, IPHI, ITAUP1, ITAUP2, ITAUQ1, 00191 $ ITAUQ2, J, LBBCSDWORK, LBBCSDWORKMIN, 00192 $ LBBCSDWORKOPT, LORBDBWORK, LORBDBWORKMIN, 00193 $ LORBDBWORKOPT, LORGLQWORK, LORGLQWORKMIN, 00194 $ LORGLQWORKOPT, LORGQRWORK, LORGQRWORKMIN, 00195 $ LORGQRWORKOPT, LWORKMIN, LWORKOPT 00196 LOGICAL COLMAJOR, DEFAULTSIGNS, LQUERY, WANTU1, WANTU2, 00197 $ WANTV1T, WANTV2T 00198 INTEGER LRWORKMIN, LRWORKOPT 00199 LOGICAL LRQUERY 00200 * .. 00201 * .. External Subroutines .. 00202 EXTERNAL XERBLA, CBBCSD, CLACPY, CLAPMR, CLAPMT, CLASCL, 00203 $ CLASET, CUNBDB, CUNGLQ, CUNGQR 00204 * .. 00205 * .. External Functions .. 00206 LOGICAL LSAME 00207 EXTERNAL LSAME 00208 * .. 00209 * .. Intrinsic Functions 00210 INTRINSIC COS, INT, MAX, MIN, SIN 00211 * .. 00212 * .. Executable Statements .. 00213 * 00214 * Test input arguments 00215 * 00216 INFO = 0 00217 WANTU1 = LSAME( JOBU1, 'Y' ) 00218 WANTU2 = LSAME( JOBU2, 'Y' ) 00219 WANTV1T = LSAME( JOBV1T, 'Y' ) 00220 WANTV2T = LSAME( JOBV2T, 'Y' ) 00221 COLMAJOR = .NOT. LSAME( TRANS, 'T' ) 00222 DEFAULTSIGNS = .NOT. LSAME( SIGNS, 'O' ) 00223 LQUERY = LWORK .EQ. -1 00224 LRQUERY = LRWORK .EQ. -1 00225 IF( M .LT. 0 ) THEN 00226 INFO = -7 00227 ELSE IF( P .LT. 0 .OR. P .GT. M ) THEN 00228 INFO = -8 00229 ELSE IF( Q .LT. 0 .OR. Q .GT. M ) THEN 00230 INFO = -9 00231 ELSE IF( ( COLMAJOR .AND. LDX11 .LT. MAX(1,P) ) .OR. 00232 $ ( .NOT.COLMAJOR .AND. LDX11 .LT. MAX(1,Q) ) ) THEN 00233 INFO = -11 00234 ELSE IF( WANTU1 .AND. LDU1 .LT. P ) THEN 00235 INFO = -14 00236 ELSE IF( WANTU2 .AND. LDU2 .LT. M-P ) THEN 00237 INFO = -16 00238 ELSE IF( WANTV1T .AND. LDV1T .LT. Q ) THEN 00239 INFO = -18 00240 ELSE IF( WANTV2T .AND. LDV2T .LT. M-Q ) THEN 00241 INFO = -20 00242 END IF 00243 * 00244 * Work with transpose if convenient 00245 * 00246 IF( INFO .EQ. 0 .AND. MIN( P, M-P ) .LT. MIN( Q, M-Q ) ) THEN 00247 IF( COLMAJOR ) THEN 00248 TRANST = 'T' 00249 ELSE 00250 TRANST = 'N' 00251 END IF 00252 IF( DEFAULTSIGNS ) THEN 00253 SIGNST = 'O' 00254 ELSE 00255 SIGNST = 'D' 00256 END IF 00257 CALL CUNCSD( JOBV1T, JOBV2T, JOBU1, JOBU2, TRANST, SIGNST, M, 00258 $ Q, P, X11, LDX11, X21, LDX21, X12, LDX12, X22, 00259 $ LDX22, THETA, V1T, LDV1T, V2T, LDV2T, U1, LDU1, 00260 $ U2, LDU2, WORK, LWORK, RWORK, LRWORK, IWORK, 00261 $ INFO ) 00262 RETURN 00263 END IF 00264 * 00265 * Work with permutation [ 0 I; I 0 ] * X * [ 0 I; I 0 ] if 00266 * convenient 00267 * 00268 IF( INFO .EQ. 0 .AND. M-Q .LT. Q ) THEN 00269 IF( DEFAULTSIGNS ) THEN 00270 SIGNST = 'O' 00271 ELSE 00272 SIGNST = 'D' 00273 END IF 00274 CALL CUNCSD( JOBU2, JOBU1, JOBV2T, JOBV1T, TRANS, SIGNST, M, 00275 $ M-P, M-Q, X22, LDX22, X21, LDX21, X12, LDX12, X11, 00276 $ LDX11, THETA, U2, LDU2, U1, LDU1, V2T, LDV2T, V1T, 00277 $ LDV1T, WORK, LWORK, RWORK, LRWORK, IWORK, INFO ) 00278 RETURN 00279 END IF 00280 * 00281 * Compute workspace 00282 * 00283 IF( INFO .EQ. 0 ) THEN 00284 * 00285 * Real workspace 00286 * 00287 IPHI = 2 00288 IB11D = IPHI + MAX( 1, Q - 1 ) 00289 IB11E = IB11D + MAX( 1, Q ) 00290 IB12D = IB11E + MAX( 1, Q - 1 ) 00291 IB12E = IB12D + MAX( 1, Q ) 00292 IB21D = IB12E + MAX( 1, Q - 1 ) 00293 IB21E = IB21D + MAX( 1, Q ) 00294 IB22D = IB21E + MAX( 1, Q - 1 ) 00295 IB22E = IB22D + MAX( 1, Q ) 00296 IBBCSD = IB22E + MAX( 1, Q - 1 ) 00297 CALL CBBCSD( JOBU1, JOBU2, JOBV1T, JOBV2T, TRANS, M, P, Q, 0, 00298 $ 0, U1, LDU1, U2, LDU2, V1T, LDV1T, V2T, LDV2T, 0, 00299 $ 0, 0, 0, 0, 0, 0, 0, RWORK, -1, CHILDINFO ) 00300 LBBCSDWORKOPT = INT( RWORK(1) ) 00301 LBBCSDWORKMIN = LBBCSDWORKOPT 00302 LRWORKOPT = IBBCSD + LBBCSDWORKOPT - 1 00303 LRWORKMIN = IBBCSD + LBBCSDWORKMIN - 1 00304 RWORK(1) = LRWORKOPT 00305 * 00306 * Complex workspace 00307 * 00308 ITAUP1 = 2 00309 ITAUP2 = ITAUP1 + MAX( 1, P ) 00310 ITAUQ1 = ITAUP2 + MAX( 1, M - P ) 00311 ITAUQ2 = ITAUQ1 + MAX( 1, Q ) 00312 IORGQR = ITAUQ2 + MAX( 1, M - Q ) 00313 CALL CUNGQR( M-Q, M-Q, M-Q, 0, MAX(1,M-Q), 0, WORK, -1, 00314 $ CHILDINFO ) 00315 LORGQRWORKOPT = INT( WORK(1) ) 00316 LORGQRWORKMIN = MAX( 1, M - Q ) 00317 IORGLQ = ITAUQ2 + MAX( 1, M - Q ) 00318 CALL CUNGLQ( M-Q, M-Q, M-Q, 0, MAX(1,M-Q), 0, WORK, -1, 00319 $ CHILDINFO ) 00320 LORGLQWORKOPT = INT( WORK(1) ) 00321 LORGLQWORKMIN = MAX( 1, M - Q ) 00322 IORBDB = ITAUQ2 + MAX( 1, M - Q ) 00323 CALL CUNBDB( TRANS, SIGNS, M, P, Q, X11, LDX11, X12, LDX12, 00324 $ X21, LDX21, X22, LDX22, 0, 0, 0, 0, 0, 0, WORK, 00325 $ -1, CHILDINFO ) 00326 LORBDBWORKOPT = INT( WORK(1) ) 00327 LORBDBWORKMIN = LORBDBWORKOPT 00328 LWORKOPT = MAX( IORGQR + LORGQRWORKOPT, IORGLQ + LORGLQWORKOPT, 00329 $ IORBDB + LORBDBWORKOPT ) - 1 00330 LWORKMIN = MAX( IORGQR + LORGQRWORKMIN, IORGLQ + LORGLQWORKMIN, 00331 $ IORBDB + LORBDBWORKMIN ) - 1 00332 WORK(1) = MAX(LWORKOPT,LWORKMIN) 00333 * 00334 IF( LWORK .LT. LWORKMIN 00335 $ .AND. .NOT. ( LQUERY .OR. LRQUERY ) ) THEN 00336 INFO = -22 00337 ELSE IF( LRWORK .LT. LRWORKMIN 00338 $ .AND. .NOT. ( LQUERY .OR. LRQUERY ) ) THEN 00339 INFO = -24 00340 ELSE 00341 LORGQRWORK = LWORK - IORGQR + 1 00342 LORGLQWORK = LWORK - IORGLQ + 1 00343 LORBDBWORK = LWORK - IORBDB + 1 00344 LBBCSDWORK = LRWORK - IBBCSD + 1 00345 END IF 00346 END IF 00347 * 00348 * Abort if any illegal arguments 00349 * 00350 IF( INFO .NE. 0 ) THEN 00351 CALL XERBLA( 'CUNCSD', -INFO ) 00352 RETURN 00353 ELSE IF( LQUERY .OR. LRQUERY ) THEN 00354 RETURN 00355 END IF 00356 * 00357 * Transform to bidiagonal block form 00358 * 00359 CALL CUNBDB( TRANS, SIGNS, M, P, Q, X11, LDX11, X12, LDX12, X21, 00360 $ LDX21, X22, LDX22, THETA, RWORK(IPHI), WORK(ITAUP1), 00361 $ WORK(ITAUP2), WORK(ITAUQ1), WORK(ITAUQ2), 00362 $ WORK(IORBDB), LORBDBWORK, CHILDINFO ) 00363 * 00364 * Accumulate Householder reflectors 00365 * 00366 IF( COLMAJOR ) THEN 00367 IF( WANTU1 .AND. P .GT. 0 ) THEN 00368 CALL CLACPY( 'L', P, Q, X11, LDX11, U1, LDU1 ) 00369 CALL CUNGQR( P, P, Q, U1, LDU1, WORK(ITAUP1), WORK(IORGQR), 00370 $ LORGQRWORK, INFO) 00371 END IF 00372 IF( WANTU2 .AND. M-P .GT. 0 ) THEN 00373 CALL CLACPY( 'L', M-P, Q, X21, LDX21, U2, LDU2 ) 00374 CALL CUNGQR( M-P, M-P, Q, U2, LDU2, WORK(ITAUP2), 00375 $ WORK(IORGQR), LORGQRWORK, INFO ) 00376 END IF 00377 IF( WANTV1T .AND. Q .GT. 0 ) THEN 00378 CALL CLACPY( 'U', Q-1, Q-1, X11(1,2), LDX11, V1T(2,2), 00379 $ LDV1T ) 00380 V1T(1, 1) = ONE 00381 DO J = 2, Q 00382 V1T(1,J) = ZERO 00383 V1T(J,1) = ZERO 00384 END DO 00385 CALL CUNGLQ( Q-1, Q-1, Q-1, V1T(2,2), LDV1T, WORK(ITAUQ1), 00386 $ WORK(IORGLQ), LORGLQWORK, INFO ) 00387 END IF 00388 IF( WANTV2T .AND. M-Q .GT. 0 ) THEN 00389 CALL CLACPY( 'U', P, M-Q, X12, LDX12, V2T, LDV2T ) 00390 CALL CLACPY( 'U', M-P-Q, M-P-Q, X22(Q+1,P+1), LDX22, 00391 $ V2T(P+1,P+1), LDV2T ) 00392 CALL CUNGLQ( M-Q, M-Q, M-Q, V2T, LDV2T, WORK(ITAUQ2), 00393 $ WORK(IORGLQ), LORGLQWORK, INFO ) 00394 END IF 00395 ELSE 00396 IF( WANTU1 .AND. P .GT. 0 ) THEN 00397 CALL CLACPY( 'U', Q, P, X11, LDX11, U1, LDU1 ) 00398 CALL CUNGLQ( P, P, Q, U1, LDU1, WORK(ITAUP1), WORK(IORGLQ), 00399 $ LORGLQWORK, INFO) 00400 END IF 00401 IF( WANTU2 .AND. M-P .GT. 0 ) THEN 00402 CALL CLACPY( 'U', Q, M-P, X21, LDX21, U2, LDU2 ) 00403 CALL CUNGLQ( M-P, M-P, Q, U2, LDU2, WORK(ITAUP2), 00404 $ WORK(IORGLQ), LORGLQWORK, INFO ) 00405 END IF 00406 IF( WANTV1T .AND. Q .GT. 0 ) THEN 00407 CALL CLACPY( 'L', Q-1, Q-1, X11(2,1), LDX11, V1T(2,2), 00408 $ LDV1T ) 00409 V1T(1, 1) = ONE 00410 DO J = 2, Q 00411 V1T(1,J) = ZERO 00412 V1T(J,1) = ZERO 00413 END DO 00414 CALL CUNGQR( Q-1, Q-1, Q-1, V1T(2,2), LDV1T, WORK(ITAUQ1), 00415 $ WORK(IORGQR), LORGQRWORK, INFO ) 00416 END IF 00417 IF( WANTV2T .AND. M-Q .GT. 0 ) THEN 00418 CALL CLACPY( 'L', M-Q, P, X12, LDX12, V2T, LDV2T ) 00419 CALL CLACPY( 'L', M-P-Q, M-P-Q, X22(P+1,Q+1), LDX22, 00420 $ V2T(P+1,P+1), LDV2T ) 00421 CALL CUNGQR( M-Q, M-Q, M-Q, V2T, LDV2T, WORK(ITAUQ2), 00422 $ WORK(IORGQR), LORGQRWORK, INFO ) 00423 END IF 00424 END IF 00425 * 00426 * Compute the CSD of the matrix in bidiagonal-block form 00427 * 00428 CALL CBBCSD( JOBU1, JOBU2, JOBV1T, JOBV2T, TRANS, M, P, Q, THETA, 00429 $ RWORK(IPHI), U1, LDU1, U2, LDU2, V1T, LDV1T, V2T, 00430 $ LDV2T, RWORK(IB11D), RWORK(IB11E), RWORK(IB12D), 00431 $ RWORK(IB12E), RWORK(IB21D), RWORK(IB21E), 00432 $ RWORK(IB22D), RWORK(IB22E), RWORK(IBBCSD), 00433 $ LBBCSDWORK, INFO ) 00434 * 00435 * Permute rows and columns to place identity submatrices in top- 00436 * left corner of (1,1)-block and/or bottom-right corner of (1,2)- 00437 * block and/or bottom-right corner of (2,1)-block and/or top-left 00438 * corner of (2,2)-block 00439 * 00440 IF( Q .GT. 0 .AND. WANTU2 ) THEN 00441 DO I = 1, Q 00442 IWORK(I) = M - P - Q + I 00443 END DO 00444 DO I = Q + 1, M - P 00445 IWORK(I) = I - Q 00446 END DO 00447 IF( COLMAJOR ) THEN 00448 CALL CLAPMT( .FALSE., M-P, M-P, U2, LDU2, IWORK ) 00449 ELSE 00450 CALL CLAPMR( .FALSE., M-P, M-P, U2, LDU2, IWORK ) 00451 END IF 00452 END IF 00453 IF( M .GT. 0 .AND. WANTV2T ) THEN 00454 DO I = 1, P 00455 IWORK(I) = M - P - Q + I 00456 END DO 00457 DO I = P + 1, M - Q 00458 IWORK(I) = I - P 00459 END DO 00460 IF( .NOT. COLMAJOR ) THEN 00461 CALL CLAPMT( .FALSE., M-Q, M-Q, V2T, LDV2T, IWORK ) 00462 ELSE 00463 CALL CLAPMR( .FALSE., M-Q, M-Q, V2T, LDV2T, IWORK ) 00464 END IF 00465 END IF 00466 * 00467 RETURN 00468 * 00469 * End CUNCSD 00470 * 00471 END 00472
1.7.3 
