numeric-linalg
Educational material on the SciPy implementation of numerical linear algebra algorithms
Name | Size | Mode | |
.. | |||
lapack/SRC/sbdsdc.f | 16783B | -rw-r--r-- |
001 002 003 004 005 006 007 008 009 010 011 012 013 014 015 016 017 018 019 020 021 022 023 024 025 026 027 028 029 030 031 032 033 034 035 036 037 038 039 040 041 042 043 044 045 046 047 048 049 050 051 052 053 054 055 056 057 058 059 060 061 062 063 064 065 066 067 068 069 070 071 072 073 074 075 076 077 078 079 080 081 082 083 084 085 086 087 088 089 090 091 092 093 094 095 096 097 098 099 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520
*> \brief \b SBDSDC * * =========== DOCUMENTATION =========== * * Online html documentation available at * http://www.netlib.org/lapack/explore-html/ * *> \htmlonly *> Download SBDSDC + dependencies *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/sbdsdc.f"> *> [TGZ]</a> *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/sbdsdc.f"> *> [ZIP]</a> *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/sbdsdc.f"> *> [TXT]</a> *> \endhtmlonly * * Definition: * =========== * * SUBROUTINE SBDSDC( UPLO, COMPQ, N, D, E, U, LDU, VT, LDVT, Q, IQ, * WORK, IWORK, INFO ) * * .. Scalar Arguments .. * CHARACTER COMPQ, UPLO * INTEGER INFO, LDU, LDVT, N * .. * .. Array Arguments .. * INTEGER IQ( * ), IWORK( * ) * REAL D( * ), E( * ), Q( * ), U( LDU, * ), * $ VT( LDVT, * ), WORK( * ) * .. * * *> \par Purpose: * ============= *> *> \verbatim *> *> SBDSDC computes the singular value decomposition (SVD) of a real *> N-by-N (upper or lower) bidiagonal matrix B: B = U * S * VT, *> using a divide and conquer method, where S is a diagonal matrix *> with non-negative diagonal elements (the singular values of B), and *> U and VT are orthogonal matrices of left and right singular vectors, *> respectively. SBDSDC can be used to compute all singular values, *> and optionally, singular vectors or singular vectors in compact form. *> *> The code currently calls SLASDQ if singular values only are desired. *> However, it can be slightly modified to compute singular values *> using the divide and conquer method. *> \endverbatim * * Arguments: * ========== * *> \param[in] UPLO *> \verbatim *> UPLO is CHARACTER*1 *> = 'U': B is upper bidiagonal. *> = 'L': B is lower bidiagonal. *> \endverbatim *> *> \param[in] COMPQ *> \verbatim *> COMPQ is CHARACTER*1 *> Specifies whether singular vectors are to be computed *> as follows: *> = 'N': Compute singular values only; *> = 'P': Compute singular values and compute singular *> vectors in compact form; *> = 'I': Compute singular values and singular vectors. *> \endverbatim *> *> \param[in] N *> \verbatim *> N is INTEGER *> The order of the matrix B. N >= 0. *> \endverbatim *> *> \param[in,out] D *> \verbatim *> D is REAL array, dimension (N) *> On entry, the n diagonal elements of the bidiagonal matrix B. *> On exit, if INFO=0, the singular values of B. *> \endverbatim *> *> \param[in,out] E *> \verbatim *> E is REAL array, dimension (N-1) *> On entry, the elements of E contain the offdiagonal *> elements of the bidiagonal matrix whose SVD is desired. *> On exit, E has been destroyed. *> \endverbatim *> *> \param[out] U *> \verbatim *> U is REAL array, dimension (LDU,N) *> If COMPQ = 'I', then: *> On exit, if INFO = 0, U contains the left singular vectors *> of the bidiagonal matrix. *> For other values of COMPQ, U is not referenced. *> \endverbatim *> *> \param[in] LDU *> \verbatim *> LDU is INTEGER *> The leading dimension of the array U. LDU >= 1. *> If singular vectors are desired, then LDU >= max( 1, N ). *> \endverbatim *> *> \param[out] VT *> \verbatim *> VT is REAL array, dimension (LDVT,N) *> If COMPQ = 'I', then: *> On exit, if INFO = 0, VT**T contains the right singular *> vectors of the bidiagonal matrix. *> For other values of COMPQ, VT is not referenced. *> \endverbatim *> *> \param[in] LDVT *> \verbatim *> LDVT is INTEGER *> The leading dimension of the array VT. LDVT >= 1. *> If singular vectors are desired, then LDVT >= max( 1, N ). *> \endverbatim *> *> \param[out] Q *> \verbatim *> Q is REAL array, dimension (LDQ) *> If COMPQ = 'P', then: *> On exit, if INFO = 0, Q and IQ contain the left *> and right singular vectors in a compact form, *> requiring O(N log N) space instead of 2*N**2. *> In particular, Q contains all the REAL data in *> LDQ >= N*(11 + 2*SMLSIZ + 8*INT(LOG_2(N/(SMLSIZ+1)))) *> words of memory, where SMLSIZ is returned by ILAENV and *> is equal to the maximum size of the subproblems at the *> bottom of the computation tree (usually about 25). *> For other values of COMPQ, Q is not referenced. *> \endverbatim *> *> \param[out] IQ *> \verbatim *> IQ is INTEGER array, dimension (LDIQ) *> If COMPQ = 'P', then: *> On exit, if INFO = 0, Q and IQ contain the left *> and right singular vectors in a compact form, *> requiring O(N log N) space instead of 2*N**2. *> In particular, IQ contains all INTEGER data in *> LDIQ >= N*(3 + 3*INT(LOG_2(N/(SMLSIZ+1)))) *> words of memory, where SMLSIZ is returned by ILAENV and *> is equal to the maximum size of the subproblems at the *> bottom of the computation tree (usually about 25). *> For other values of COMPQ, IQ is not referenced. *> \endverbatim *> *> \param[out] WORK *> \verbatim *> WORK is REAL array, dimension (MAX(1,LWORK)) *> If COMPQ = 'N' then LWORK >= (4 * N). *> If COMPQ = 'P' then LWORK >= (6 * N). *> If COMPQ = 'I' then LWORK >= (3 * N**2 + 4 * N). *> \endverbatim *> *> \param[out] IWORK *> \verbatim *> IWORK is INTEGER array, dimension (8*N) *> \endverbatim *> *> \param[out] INFO *> \verbatim *> INFO is INTEGER *> = 0: successful exit. *> < 0: if INFO = -i, the i-th argument had an illegal value. *> > 0: The algorithm failed to compute a singular value. *> The update process of divide and conquer failed. *> \endverbatim * * Authors: * ======== * *> \author Univ. of Tennessee *> \author Univ. of California Berkeley *> \author Univ. of Colorado Denver *> \author NAG Ltd. * *> \ingroup bdsdc * *> \par Contributors: * ================== *> *> Ming Gu and Huan Ren, Computer Science Division, University of *> California at Berkeley, USA *> * ===================================================================== SUBROUTINE SBDSDC( UPLO, COMPQ, N, D, E, U, LDU, VT, LDVT, Q, $ IQ, $ WORK, IWORK, INFO ) * * -- LAPACK computational routine -- * -- LAPACK is a software package provided by Univ. of Tennessee, -- * -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- * * .. Scalar Arguments .. CHARACTER COMPQ, UPLO INTEGER INFO, LDU, LDVT, N * .. * .. Array Arguments .. INTEGER IQ( * ), IWORK( * ) REAL D( * ), E( * ), Q( * ), U( LDU, * ), $ VT( LDVT, * ), WORK( * ) * .. * * ===================================================================== * Changed dimension statement in comment describing E from (N) to * (N-1). Sven, 17 Feb 05. * ===================================================================== * * .. Parameters .. REAL ZERO, ONE, TWO PARAMETER ( ZERO = 0.0E+0, ONE = 1.0E+0, TWO = 2.0E+0 ) * .. * .. Local Scalars .. INTEGER DIFL, DIFR, GIVCOL, GIVNUM, GIVPTR, I, IC, $ ICOMPQ, IERR, II, IS, IU, IUPLO, IVT, J, K, KK, $ MLVL, NM1, NSIZE, PERM, POLES, QSTART, SMLSIZ, $ SMLSZP, SQRE, START, WSTART, Z REAL CS, EPS, ORGNRM, P, R, SN * .. * .. External Functions .. LOGICAL LSAME INTEGER ILAENV REAL SLAMCH, SLANST EXTERNAL SLAMCH, SLANST, ILAENV, LSAME * .. * .. External Subroutines .. EXTERNAL SCOPY, SLARTG, SLASCL, SLASD0, SLASDA, $ SLASDQ, $ SLASET, SLASR, SSWAP, XERBLA * .. * .. Intrinsic Functions .. INTRINSIC REAL, ABS, INT, LOG, SIGN * .. * .. Executable Statements .. * * Test the input parameters. * INFO = 0 * IUPLO = 0 IF( LSAME( UPLO, 'U' ) ) $ IUPLO = 1 IF( LSAME( UPLO, 'L' ) ) $ IUPLO = 2 IF( LSAME( COMPQ, 'N' ) ) THEN ICOMPQ = 0 ELSE IF( LSAME( COMPQ, 'P' ) ) THEN ICOMPQ = 1 ELSE IF( LSAME( COMPQ, 'I' ) ) THEN ICOMPQ = 2 ELSE ICOMPQ = -1 END IF IF( IUPLO.EQ.0 ) THEN INFO = -1 ELSE IF( ICOMPQ.LT.0 ) THEN INFO = -2 ELSE IF( N.LT.0 ) THEN INFO = -3 ELSE IF( ( LDU.LT.1 ) .OR. ( ( ICOMPQ.EQ.2 ) .AND. ( LDU.LT. $ N ) ) ) THEN INFO = -7 ELSE IF( ( LDVT.LT.1 ) .OR. ( ( ICOMPQ.EQ.2 ) .AND. ( LDVT.LT. $ N ) ) ) THEN INFO = -9 END IF IF( INFO.NE.0 ) THEN CALL XERBLA( 'SBDSDC', -INFO ) RETURN END IF * * Quick return if possible * IF( N.EQ.0 ) $ RETURN SMLSIZ = ILAENV( 9, 'SBDSDC', ' ', 0, 0, 0, 0 ) IF( N.EQ.1 ) THEN IF( ICOMPQ.EQ.1 ) THEN Q( 1 ) = SIGN( ONE, D( 1 ) ) Q( 1+SMLSIZ*N ) = ONE ELSE IF( ICOMPQ.EQ.2 ) THEN U( 1, 1 ) = SIGN( ONE, D( 1 ) ) VT( 1, 1 ) = ONE END IF D( 1 ) = ABS( D( 1 ) ) RETURN END IF NM1 = N - 1 * * If matrix lower bidiagonal, rotate to be upper bidiagonal * by applying Givens rotations on the left * WSTART = 1 QSTART = 3 IF( ICOMPQ.EQ.1 ) THEN CALL SCOPY( N, D, 1, Q( 1 ), 1 ) CALL SCOPY( N-1, E, 1, Q( N+1 ), 1 ) END IF IF( IUPLO.EQ.2 ) THEN QSTART = 5 IF( ICOMPQ .EQ. 2 ) WSTART = 2*N - 1 DO 10 I = 1, N - 1 CALL SLARTG( D( I ), E( I ), CS, SN, R ) D( I ) = R E( I ) = SN*D( I+1 ) D( I+1 ) = CS*D( I+1 ) IF( ICOMPQ.EQ.1 ) THEN Q( I+2*N ) = CS Q( I+3*N ) = SN ELSE IF( ICOMPQ.EQ.2 ) THEN WORK( I ) = CS WORK( NM1+I ) = -SN END IF 10 CONTINUE END IF * * If ICOMPQ = 0, use SLASDQ to compute the singular values. * IF( ICOMPQ.EQ.0 ) THEN * Ignore WSTART, instead using WORK( 1 ), since the two vectors * for CS and -SN above are added only if ICOMPQ == 2, * and adding them exceeds documented WORK size of 4*n. CALL SLASDQ( 'U', 0, N, 0, 0, 0, D, E, VT, LDVT, U, LDU, U, $ LDU, WORK( 1 ), INFO ) GO TO 40 END IF * * If N is smaller than the minimum divide size SMLSIZ, then solve * the problem with another solver. * IF( N.LE.SMLSIZ ) THEN IF( ICOMPQ.EQ.2 ) THEN CALL SLASET( 'A', N, N, ZERO, ONE, U, LDU ) CALL SLASET( 'A', N, N, ZERO, ONE, VT, LDVT ) CALL SLASDQ( 'U', 0, N, N, N, 0, D, E, VT, LDVT, U, LDU, $ U, $ LDU, WORK( WSTART ), INFO ) ELSE IF( ICOMPQ.EQ.1 ) THEN IU = 1 IVT = IU + N CALL SLASET( 'A', N, N, ZERO, ONE, $ Q( IU+( QSTART-1 )*N ), $ N ) CALL SLASET( 'A', N, N, ZERO, ONE, $ Q( IVT+( QSTART-1 )*N ), $ N ) CALL SLASDQ( 'U', 0, N, N, N, 0, D, E, $ Q( IVT+( QSTART-1 )*N ), N, $ Q( IU+( QSTART-1 )*N ), N, $ Q( IU+( QSTART-1 )*N ), N, WORK( WSTART ), $ INFO ) END IF GO TO 40 END IF * IF( ICOMPQ.EQ.2 ) THEN CALL SLASET( 'A', N, N, ZERO, ONE, U, LDU ) CALL SLASET( 'A', N, N, ZERO, ONE, VT, LDVT ) END IF * * Scale. * ORGNRM = SLANST( 'M', N, D, E ) IF( ORGNRM.EQ.ZERO ) $ RETURN CALL SLASCL( 'G', 0, 0, ORGNRM, ONE, N, 1, D, N, IERR ) CALL SLASCL( 'G', 0, 0, ORGNRM, ONE, NM1, 1, E, NM1, IERR ) * EPS = SLAMCH( 'Epsilon' ) * MLVL = INT( LOG( REAL( N ) / REAL( SMLSIZ+1 ) ) / LOG( TWO ) ) + 1 SMLSZP = SMLSIZ + 1 * IF( ICOMPQ.EQ.1 ) THEN IU = 1 IVT = 1 + SMLSIZ DIFL = IVT + SMLSZP DIFR = DIFL + MLVL Z = DIFR + MLVL*2 IC = Z + MLVL IS = IC + 1 POLES = IS + 1 GIVNUM = POLES + 2*MLVL * K = 1 GIVPTR = 2 PERM = 3 GIVCOL = PERM + MLVL END IF * DO 20 I = 1, N IF( ABS( D( I ) ).LT.EPS ) THEN D( I ) = SIGN( EPS, D( I ) ) END IF 20 CONTINUE * START = 1 SQRE = 0 * DO 30 I = 1, NM1 IF( ( ABS( E( I ) ).LT.EPS ) .OR. ( I.EQ.NM1 ) ) THEN * * Subproblem found. First determine its size and then * apply divide and conquer on it. * IF( I.LT.NM1 ) THEN * * A subproblem with E(I) small for I < NM1. * NSIZE = I - START + 1 ELSE IF( ABS( E( I ) ).GE.EPS ) THEN * * A subproblem with E(NM1) not too small but I = NM1. * NSIZE = N - START + 1 ELSE * * A subproblem with E(NM1) small. This implies an * 1-by-1 subproblem at D(N). Solve this 1-by-1 problem * first. * NSIZE = I - START + 1 IF( ICOMPQ.EQ.2 ) THEN U( N, N ) = SIGN( ONE, D( N ) ) VT( N, N ) = ONE ELSE IF( ICOMPQ.EQ.1 ) THEN Q( N+( QSTART-1 )*N ) = SIGN( ONE, D( N ) ) Q( N+( SMLSIZ+QSTART-1 )*N ) = ONE END IF D( N ) = ABS( D( N ) ) END IF IF( ICOMPQ.EQ.2 ) THEN CALL SLASD0( NSIZE, SQRE, D( START ), E( START ), $ U( START, START ), LDU, VT( START, START ), $ LDVT, SMLSIZ, IWORK, WORK( WSTART ), INFO ) ELSE CALL SLASDA( ICOMPQ, SMLSIZ, NSIZE, SQRE, D( START ), $ E( START ), Q( START+( IU+QSTART-2 )*N ), N, $ Q( START+( IVT+QSTART-2 )*N ), $ IQ( START+K*N ), Q( START+( DIFL+QSTART-2 )* $ N ), Q( START+( DIFR+QSTART-2 )*N ), $ Q( START+( Z+QSTART-2 )*N ), $ Q( START+( POLES+QSTART-2 )*N ), $ IQ( START+GIVPTR*N ), IQ( START+GIVCOL*N ), $ N, IQ( START+PERM*N ), $ Q( START+( GIVNUM+QSTART-2 )*N ), $ Q( START+( IC+QSTART-2 )*N ), $ Q( START+( IS+QSTART-2 )*N ), $ WORK( WSTART ), IWORK, INFO ) END IF IF( INFO.NE.0 ) THEN RETURN END IF START = I + 1 END IF 30 CONTINUE * * Unscale * CALL SLASCL( 'G', 0, 0, ONE, ORGNRM, N, 1, D, N, IERR ) 40 CONTINUE * * Use Selection Sort to minimize swaps of singular vectors * DO 60 II = 2, N I = II - 1 KK = I P = D( I ) DO 50 J = II, N IF( D( J ).GT.P ) THEN KK = J P = D( J ) END IF 50 CONTINUE IF( KK.NE.I ) THEN D( KK ) = D( I ) D( I ) = P IF( ICOMPQ.EQ.1 ) THEN IQ( I ) = KK ELSE IF( ICOMPQ.EQ.2 ) THEN CALL SSWAP( N, U( 1, I ), 1, U( 1, KK ), 1 ) CALL SSWAP( N, VT( I, 1 ), LDVT, VT( KK, 1 ), LDVT ) END IF ELSE IF( ICOMPQ.EQ.1 ) THEN IQ( I ) = I END IF 60 CONTINUE * * If ICOMPQ = 1, use IQ(N,1) as the indicator for UPLO * IF( ICOMPQ.EQ.1 ) THEN IF( IUPLO.EQ.1 ) THEN IQ( N ) = 1 ELSE IQ( N ) = 0 END IF END IF * * If B is lower bidiagonal, update U by those Givens rotations * which rotated B to be upper bidiagonal * IF( ( IUPLO.EQ.2 ) .AND. ( ICOMPQ.EQ.2 ) ) $ CALL SLASR( 'L', 'V', 'B', N, N, WORK( 1 ), WORK( N ), U, $ LDU ) * RETURN * * End of SBDSDC * END