Annotation of rpl/lapack/lapack/dgels.f, revision 1.10

1.9       bertrand    1: *> \brief <b> DGELS solves overdetermined or underdetermined systems for GE matrices</b>
                      2: *
                      3: *  =========== DOCUMENTATION ===========
                      4: *
                      5: * Online html documentation available at 
                      6: *            http://www.netlib.org/lapack/explore-html/ 
                      7: *
                      8: *> \htmlonly
                      9: *> Download DGELS + dependencies 
                     10: *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/dgels.f"> 
                     11: *> [TGZ]</a> 
                     12: *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/dgels.f"> 
                     13: *> [ZIP]</a> 
                     14: *> <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/dgels.f"> 
                     15: *> [TXT]</a>
                     16: *> \endhtmlonly 
                     17: *
                     18: *  Definition:
                     19: *  ===========
                     20: *
                     21: *       SUBROUTINE DGELS( TRANS, M, N, NRHS, A, LDA, B, LDB, WORK, LWORK,
                     22: *                         INFO )
                     23: * 
                     24: *       .. Scalar Arguments ..
                     25: *       CHARACTER          TRANS
                     26: *       INTEGER            INFO, LDA, LDB, LWORK, M, N, NRHS
                     27: *       ..
                     28: *       .. Array Arguments ..
                     29: *       DOUBLE PRECISION   A( LDA, * ), B( LDB, * ), WORK( * )
                     30: *       ..
                     31: *  
                     32: *
                     33: *> \par Purpose:
                     34: *  =============
                     35: *>
                     36: *> \verbatim
                     37: *>
                     38: *> DGELS solves overdetermined or underdetermined real linear systems
                     39: *> involving an M-by-N matrix A, or its transpose, using a QR or LQ
                     40: *> factorization of A.  It is assumed that A has full rank.
                     41: *>
                     42: *> The following options are provided:
                     43: *>
                     44: *> 1. If TRANS = 'N' and m >= n:  find the least squares solution of
                     45: *>    an overdetermined system, i.e., solve the least squares problem
                     46: *>                 minimize || B - A*X ||.
                     47: *>
                     48: *> 2. If TRANS = 'N' and m < n:  find the minimum norm solution of
                     49: *>    an underdetermined system A * X = B.
                     50: *>
                     51: *> 3. If TRANS = 'T' and m >= n:  find the minimum norm solution of
                     52: *>    an undetermined system A**T * X = B.
                     53: *>
                     54: *> 4. If TRANS = 'T' and m < n:  find the least squares solution of
                     55: *>    an overdetermined system, i.e., solve the least squares problem
                     56: *>                 minimize || B - A**T * X ||.
                     57: *>
                     58: *> Several right hand side vectors b and solution vectors x can be
                     59: *> handled in a single call; they are stored as the columns of the
                     60: *> M-by-NRHS right hand side matrix B and the N-by-NRHS solution
                     61: *> matrix X.
                     62: *> \endverbatim
                     63: *
                     64: *  Arguments:
                     65: *  ==========
                     66: *
                     67: *> \param[in] TRANS
                     68: *> \verbatim
                     69: *>          TRANS is CHARACTER*1
                     70: *>          = 'N': the linear system involves A;
                     71: *>          = 'T': the linear system involves A**T.
                     72: *> \endverbatim
                     73: *>
                     74: *> \param[in] M
                     75: *> \verbatim
                     76: *>          M is INTEGER
                     77: *>          The number of rows of the matrix A.  M >= 0.
                     78: *> \endverbatim
                     79: *>
                     80: *> \param[in] N
                     81: *> \verbatim
                     82: *>          N is INTEGER
                     83: *>          The number of columns of the matrix A.  N >= 0.
                     84: *> \endverbatim
                     85: *>
                     86: *> \param[in] NRHS
                     87: *> \verbatim
                     88: *>          NRHS is INTEGER
                     89: *>          The number of right hand sides, i.e., the number of
                     90: *>          columns of the matrices B and X. NRHS >=0.
                     91: *> \endverbatim
                     92: *>
                     93: *> \param[in,out] A
                     94: *> \verbatim
                     95: *>          A is DOUBLE PRECISION array, dimension (LDA,N)
                     96: *>          On entry, the M-by-N matrix A.
                     97: *>          On exit,
                     98: *>            if M >= N, A is overwritten by details of its QR
                     99: *>                       factorization as returned by DGEQRF;
                    100: *>            if M <  N, A is overwritten by details of its LQ
                    101: *>                       factorization as returned by DGELQF.
                    102: *> \endverbatim
                    103: *>
                    104: *> \param[in] LDA
                    105: *> \verbatim
                    106: *>          LDA is INTEGER
                    107: *>          The leading dimension of the array A.  LDA >= max(1,M).
                    108: *> \endverbatim
                    109: *>
                    110: *> \param[in,out] B
                    111: *> \verbatim
                    112: *>          B is DOUBLE PRECISION array, dimension (LDB,NRHS)
                    113: *>          On entry, the matrix B of right hand side vectors, stored
                    114: *>          columnwise; B is M-by-NRHS if TRANS = 'N', or N-by-NRHS
                    115: *>          if TRANS = 'T'.
                    116: *>          On exit, if INFO = 0, B is overwritten by the solution
                    117: *>          vectors, stored columnwise:
                    118: *>          if TRANS = 'N' and m >= n, rows 1 to n of B contain the least
                    119: *>          squares solution vectors; the residual sum of squares for the
                    120: *>          solution in each column is given by the sum of squares of
                    121: *>          elements N+1 to M in that column;
                    122: *>          if TRANS = 'N' and m < n, rows 1 to N of B contain the
                    123: *>          minimum norm solution vectors;
                    124: *>          if TRANS = 'T' and m >= n, rows 1 to M of B contain the
                    125: *>          minimum norm solution vectors;
                    126: *>          if TRANS = 'T' and m < n, rows 1 to M of B contain the
                    127: *>          least squares solution vectors; the residual sum of squares
                    128: *>          for the solution in each column is given by the sum of
                    129: *>          squares of elements M+1 to N in that column.
                    130: *> \endverbatim
                    131: *>
                    132: *> \param[in] LDB
                    133: *> \verbatim
                    134: *>          LDB is INTEGER
                    135: *>          The leading dimension of the array B. LDB >= MAX(1,M,N).
                    136: *> \endverbatim
                    137: *>
                    138: *> \param[out] WORK
                    139: *> \verbatim
                    140: *>          WORK is DOUBLE PRECISION array, dimension (MAX(1,LWORK))
                    141: *>          On exit, if INFO = 0, WORK(1) returns the optimal LWORK.
                    142: *> \endverbatim
                    143: *>
                    144: *> \param[in] LWORK
                    145: *> \verbatim
                    146: *>          LWORK is INTEGER
                    147: *>          The dimension of the array WORK.
                    148: *>          LWORK >= max( 1, MN + max( MN, NRHS ) ).
                    149: *>          For optimal performance,
                    150: *>          LWORK >= max( 1, MN + max( MN, NRHS )*NB ).
                    151: *>          where MN = min(M,N) and NB is the optimum block size.
                    152: *>
                    153: *>          If LWORK = -1, then a workspace query is assumed; the routine
                    154: *>          only calculates the optimal size of the WORK array, returns
                    155: *>          this value as the first entry of the WORK array, and no error
                    156: *>          message related to LWORK is issued by XERBLA.
                    157: *> \endverbatim
                    158: *>
                    159: *> \param[out] INFO
                    160: *> \verbatim
                    161: *>          INFO is INTEGER
                    162: *>          = 0:  successful exit
                    163: *>          < 0:  if INFO = -i, the i-th argument had an illegal value
                    164: *>          > 0:  if INFO =  i, the i-th diagonal element of the
                    165: *>                triangular factor of A is zero, so that A does not have
                    166: *>                full rank; the least squares solution could not be
                    167: *>                computed.
                    168: *> \endverbatim
                    169: *
                    170: *  Authors:
                    171: *  ========
                    172: *
                    173: *> \author Univ. of Tennessee 
                    174: *> \author Univ. of California Berkeley 
                    175: *> \author Univ. of Colorado Denver 
                    176: *> \author NAG Ltd. 
                    177: *
                    178: *> \date November 2011
                    179: *
                    180: *> \ingroup doubleGEsolve
                    181: *
                    182: *  =====================================================================
1.1       bertrand  183:       SUBROUTINE DGELS( TRANS, M, N, NRHS, A, LDA, B, LDB, WORK, LWORK,
                    184:      $                  INFO )
                    185: *
1.9       bertrand  186: *  -- LAPACK driver routine (version 3.4.0) --
1.1       bertrand  187: *  -- LAPACK is a software package provided by Univ. of Tennessee,    --
                    188: *  -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
1.9       bertrand  189: *     November 2011
1.1       bertrand  190: *
                    191: *     .. Scalar Arguments ..
                    192:       CHARACTER          TRANS
                    193:       INTEGER            INFO, LDA, LDB, LWORK, M, N, NRHS
                    194: *     ..
                    195: *     .. Array Arguments ..
                    196:       DOUBLE PRECISION   A( LDA, * ), B( LDB, * ), WORK( * )
                    197: *     ..
                    198: *
                    199: *  =====================================================================
                    200: *
                    201: *     .. Parameters ..
                    202:       DOUBLE PRECISION   ZERO, ONE
                    203:       PARAMETER          ( ZERO = 0.0D0, ONE = 1.0D0 )
                    204: *     ..
                    205: *     .. Local Scalars ..
                    206:       LOGICAL            LQUERY, TPSD
                    207:       INTEGER            BROW, I, IASCL, IBSCL, J, MN, NB, SCLLEN, WSIZE
                    208:       DOUBLE PRECISION   ANRM, BIGNUM, BNRM, SMLNUM
                    209: *     ..
                    210: *     .. Local Arrays ..
                    211:       DOUBLE PRECISION   RWORK( 1 )
                    212: *     ..
                    213: *     .. External Functions ..
                    214:       LOGICAL            LSAME
                    215:       INTEGER            ILAENV
                    216:       DOUBLE PRECISION   DLAMCH, DLANGE
                    217:       EXTERNAL           LSAME, ILAENV, DLABAD, DLAMCH, DLANGE
                    218: *     ..
                    219: *     .. External Subroutines ..
                    220:       EXTERNAL           DGELQF, DGEQRF, DLASCL, DLASET, DORMLQ, DORMQR,
                    221:      $                   DTRTRS, XERBLA
                    222: *     ..
                    223: *     .. Intrinsic Functions ..
                    224:       INTRINSIC          DBLE, MAX, MIN
                    225: *     ..
                    226: *     .. Executable Statements ..
                    227: *
                    228: *     Test the input arguments.
                    229: *
                    230:       INFO = 0
                    231:       MN = MIN( M, N )
                    232:       LQUERY = ( LWORK.EQ.-1 )
                    233:       IF( .NOT.( LSAME( TRANS, 'N' ) .OR. LSAME( TRANS, 'T' ) ) ) THEN
                    234:          INFO = -1
                    235:       ELSE IF( M.LT.0 ) THEN
                    236:          INFO = -2
                    237:       ELSE IF( N.LT.0 ) THEN
                    238:          INFO = -3
                    239:       ELSE IF( NRHS.LT.0 ) THEN
                    240:          INFO = -4
                    241:       ELSE IF( LDA.LT.MAX( 1, M ) ) THEN
                    242:          INFO = -6
                    243:       ELSE IF( LDB.LT.MAX( 1, M, N ) ) THEN
                    244:          INFO = -8
                    245:       ELSE IF( LWORK.LT.MAX( 1, MN+MAX( MN, NRHS ) ) .AND. .NOT.LQUERY )
                    246:      $          THEN
                    247:          INFO = -10
                    248:       END IF
                    249: *
                    250: *     Figure out optimal block size
                    251: *
                    252:       IF( INFO.EQ.0 .OR. INFO.EQ.-10 ) THEN
                    253: *
                    254:          TPSD = .TRUE.
                    255:          IF( LSAME( TRANS, 'N' ) )
                    256:      $      TPSD = .FALSE.
                    257: *
                    258:          IF( M.GE.N ) THEN
                    259:             NB = ILAENV( 1, 'DGEQRF', ' ', M, N, -1, -1 )
                    260:             IF( TPSD ) THEN
                    261:                NB = MAX( NB, ILAENV( 1, 'DORMQR', 'LN', M, NRHS, N,
                    262:      $              -1 ) )
                    263:             ELSE
                    264:                NB = MAX( NB, ILAENV( 1, 'DORMQR', 'LT', M, NRHS, N,
                    265:      $              -1 ) )
                    266:             END IF
                    267:          ELSE
                    268:             NB = ILAENV( 1, 'DGELQF', ' ', M, N, -1, -1 )
                    269:             IF( TPSD ) THEN
                    270:                NB = MAX( NB, ILAENV( 1, 'DORMLQ', 'LT', N, NRHS, M,
                    271:      $              -1 ) )
                    272:             ELSE
                    273:                NB = MAX( NB, ILAENV( 1, 'DORMLQ', 'LN', N, NRHS, M,
                    274:      $              -1 ) )
                    275:             END IF
                    276:          END IF
                    277: *
                    278:          WSIZE = MAX( 1, MN+MAX( MN, NRHS )*NB )
                    279:          WORK( 1 ) = DBLE( WSIZE )
                    280: *
                    281:       END IF
                    282: *
                    283:       IF( INFO.NE.0 ) THEN
                    284:          CALL XERBLA( 'DGELS ', -INFO )
                    285:          RETURN
                    286:       ELSE IF( LQUERY ) THEN
                    287:          RETURN
                    288:       END IF
                    289: *
                    290: *     Quick return if possible
                    291: *
                    292:       IF( MIN( M, N, NRHS ).EQ.0 ) THEN
                    293:          CALL DLASET( 'Full', MAX( M, N ), NRHS, ZERO, ZERO, B, LDB )
                    294:          RETURN
                    295:       END IF
                    296: *
                    297: *     Get machine parameters
                    298: *
                    299:       SMLNUM = DLAMCH( 'S' ) / DLAMCH( 'P' )
                    300:       BIGNUM = ONE / SMLNUM
                    301:       CALL DLABAD( SMLNUM, BIGNUM )
                    302: *
                    303: *     Scale A, B if max element outside range [SMLNUM,BIGNUM]
                    304: *
                    305:       ANRM = DLANGE( 'M', M, N, A, LDA, RWORK )
                    306:       IASCL = 0
                    307:       IF( ANRM.GT.ZERO .AND. ANRM.LT.SMLNUM ) THEN
                    308: *
                    309: *        Scale matrix norm up to SMLNUM
                    310: *
                    311:          CALL DLASCL( 'G', 0, 0, ANRM, SMLNUM, M, N, A, LDA, INFO )
                    312:          IASCL = 1
                    313:       ELSE IF( ANRM.GT.BIGNUM ) THEN
                    314: *
                    315: *        Scale matrix norm down to BIGNUM
                    316: *
                    317:          CALL DLASCL( 'G', 0, 0, ANRM, BIGNUM, M, N, A, LDA, INFO )
                    318:          IASCL = 2
                    319:       ELSE IF( ANRM.EQ.ZERO ) THEN
                    320: *
                    321: *        Matrix all zero. Return zero solution.
                    322: *
                    323:          CALL DLASET( 'F', MAX( M, N ), NRHS, ZERO, ZERO, B, LDB )
                    324:          GO TO 50
                    325:       END IF
                    326: *
                    327:       BROW = M
                    328:       IF( TPSD )
                    329:      $   BROW = N
                    330:       BNRM = DLANGE( 'M', BROW, NRHS, B, LDB, RWORK )
                    331:       IBSCL = 0
                    332:       IF( BNRM.GT.ZERO .AND. BNRM.LT.SMLNUM ) THEN
                    333: *
                    334: *        Scale matrix norm up to SMLNUM
                    335: *
                    336:          CALL DLASCL( 'G', 0, 0, BNRM, SMLNUM, BROW, NRHS, B, LDB,
                    337:      $                INFO )
                    338:          IBSCL = 1
                    339:       ELSE IF( BNRM.GT.BIGNUM ) THEN
                    340: *
                    341: *        Scale matrix norm down to BIGNUM
                    342: *
                    343:          CALL DLASCL( 'G', 0, 0, BNRM, BIGNUM, BROW, NRHS, B, LDB,
                    344:      $                INFO )
                    345:          IBSCL = 2
                    346:       END IF
                    347: *
                    348:       IF( M.GE.N ) THEN
                    349: *
                    350: *        compute QR factorization of A
                    351: *
                    352:          CALL DGEQRF( M, N, A, LDA, WORK( 1 ), WORK( MN+1 ), LWORK-MN,
                    353:      $                INFO )
                    354: *
                    355: *        workspace at least N, optimally N*NB
                    356: *
                    357:          IF( .NOT.TPSD ) THEN
                    358: *
                    359: *           Least-Squares Problem min || A * X - B ||
                    360: *
1.8       bertrand  361: *           B(1:M,1:NRHS) := Q**T * B(1:M,1:NRHS)
1.1       bertrand  362: *
                    363:             CALL DORMQR( 'Left', 'Transpose', M, NRHS, N, A, LDA,
                    364:      $                   WORK( 1 ), B, LDB, WORK( MN+1 ), LWORK-MN,
                    365:      $                   INFO )
                    366: *
                    367: *           workspace at least NRHS, optimally NRHS*NB
                    368: *
                    369: *           B(1:N,1:NRHS) := inv(R) * B(1:N,1:NRHS)
                    370: *
                    371:             CALL DTRTRS( 'Upper', 'No transpose', 'Non-unit', N, NRHS,
                    372:      $                   A, LDA, B, LDB, INFO )
                    373: *
                    374:             IF( INFO.GT.0 ) THEN
                    375:                RETURN
                    376:             END IF
                    377: *
                    378:             SCLLEN = N
                    379: *
                    380:          ELSE
                    381: *
1.8       bertrand  382: *           Overdetermined system of equations A**T * X = B
1.1       bertrand  383: *
1.8       bertrand  384: *           B(1:N,1:NRHS) := inv(R**T) * B(1:N,1:NRHS)
1.1       bertrand  385: *
                    386:             CALL DTRTRS( 'Upper', 'Transpose', 'Non-unit', N, NRHS,
                    387:      $                   A, LDA, B, LDB, INFO )
                    388: *
                    389:             IF( INFO.GT.0 ) THEN
                    390:                RETURN
                    391:             END IF
                    392: *
                    393: *           B(N+1:M,1:NRHS) = ZERO
                    394: *
                    395:             DO 20 J = 1, NRHS
                    396:                DO 10 I = N + 1, M
                    397:                   B( I, J ) = ZERO
                    398:    10          CONTINUE
                    399:    20       CONTINUE
                    400: *
                    401: *           B(1:M,1:NRHS) := Q(1:N,:) * B(1:N,1:NRHS)
                    402: *
                    403:             CALL DORMQR( 'Left', 'No transpose', M, NRHS, N, A, LDA,
                    404:      $                   WORK( 1 ), B, LDB, WORK( MN+1 ), LWORK-MN,
                    405:      $                   INFO )
                    406: *
                    407: *           workspace at least NRHS, optimally NRHS*NB
                    408: *
                    409:             SCLLEN = M
                    410: *
                    411:          END IF
                    412: *
                    413:       ELSE
                    414: *
                    415: *        Compute LQ factorization of A
                    416: *
                    417:          CALL DGELQF( M, N, A, LDA, WORK( 1 ), WORK( MN+1 ), LWORK-MN,
                    418:      $                INFO )
                    419: *
                    420: *        workspace at least M, optimally M*NB.
                    421: *
                    422:          IF( .NOT.TPSD ) THEN
                    423: *
                    424: *           underdetermined system of equations A * X = B
                    425: *
                    426: *           B(1:M,1:NRHS) := inv(L) * B(1:M,1:NRHS)
                    427: *
                    428:             CALL DTRTRS( 'Lower', 'No transpose', 'Non-unit', M, NRHS,
                    429:      $                   A, LDA, B, LDB, INFO )
                    430: *
                    431:             IF( INFO.GT.0 ) THEN
                    432:                RETURN
                    433:             END IF
                    434: *
                    435: *           B(M+1:N,1:NRHS) = 0
                    436: *
                    437:             DO 40 J = 1, NRHS
                    438:                DO 30 I = M + 1, N
                    439:                   B( I, J ) = ZERO
                    440:    30          CONTINUE
                    441:    40       CONTINUE
                    442: *
1.8       bertrand  443: *           B(1:N,1:NRHS) := Q(1:N,:)**T * B(1:M,1:NRHS)
1.1       bertrand  444: *
                    445:             CALL DORMLQ( 'Left', 'Transpose', N, NRHS, M, A, LDA,
                    446:      $                   WORK( 1 ), B, LDB, WORK( MN+1 ), LWORK-MN,
                    447:      $                   INFO )
                    448: *
                    449: *           workspace at least NRHS, optimally NRHS*NB
                    450: *
                    451:             SCLLEN = N
                    452: *
                    453:          ELSE
                    454: *
1.8       bertrand  455: *           overdetermined system min || A**T * X - B ||
1.1       bertrand  456: *
                    457: *           B(1:N,1:NRHS) := Q * B(1:N,1:NRHS)
                    458: *
                    459:             CALL DORMLQ( 'Left', 'No transpose', N, NRHS, M, A, LDA,
                    460:      $                   WORK( 1 ), B, LDB, WORK( MN+1 ), LWORK-MN,
                    461:      $                   INFO )
                    462: *
                    463: *           workspace at least NRHS, optimally NRHS*NB
                    464: *
1.8       bertrand  465: *           B(1:M,1:NRHS) := inv(L**T) * B(1:M,1:NRHS)
1.1       bertrand  466: *
                    467:             CALL DTRTRS( 'Lower', 'Transpose', 'Non-unit', M, NRHS,
                    468:      $                   A, LDA, B, LDB, INFO )
                    469: *
                    470:             IF( INFO.GT.0 ) THEN
                    471:                RETURN
                    472:             END IF
                    473: *
                    474:             SCLLEN = M
                    475: *
                    476:          END IF
                    477: *
                    478:       END IF
                    479: *
                    480: *     Undo scaling
                    481: *
                    482:       IF( IASCL.EQ.1 ) THEN
                    483:          CALL DLASCL( 'G', 0, 0, ANRM, SMLNUM, SCLLEN, NRHS, B, LDB,
                    484:      $                INFO )
                    485:       ELSE IF( IASCL.EQ.2 ) THEN
                    486:          CALL DLASCL( 'G', 0, 0, ANRM, BIGNUM, SCLLEN, NRHS, B, LDB,
                    487:      $                INFO )
                    488:       END IF
                    489:       IF( IBSCL.EQ.1 ) THEN
                    490:          CALL DLASCL( 'G', 0, 0, SMLNUM, BNRM, SCLLEN, NRHS, B, LDB,
                    491:      $                INFO )
                    492:       ELSE IF( IBSCL.EQ.2 ) THEN
                    493:          CALL DLASCL( 'G', 0, 0, BIGNUM, BNRM, SCLLEN, NRHS, B, LDB,
                    494:      $                INFO )
                    495:       END IF
                    496: *
                    497:    50 CONTINUE
                    498:       WORK( 1 ) = DBLE( WSIZE )
                    499: *
                    500:       RETURN
                    501: *
                    502: *     End of DGELS
                    503: *
                    504:       END

CVSweb interface <joel.bertrand@systella.fr>