LAPACK: SRC/cgels.f Source File

Go to the documentation of this file.
00001       SUBROUTINE CGELS( TRANS, M, N, NRHS, A, LDA, B, LDB, WORK, LWORK,
00002      $                  INFO )
00003 *
00004 *  -- LAPACK driver routine (version 3.2) --
00005 *  -- LAPACK is a software package provided by Univ. of Tennessee,    --
00006 *  -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
00007 *     November 2006
00008 *
00009 *     .. Scalar Arguments ..
00010       CHARACTER          TRANS
00011       INTEGER            INFO, LDA, LDB, LWORK, M, N, NRHS
00012 *     ..
00013 *     .. Array Arguments ..
00014       COMPLEX            A( LDA, * ), B( LDB, * ), WORK( * )
00015 *     ..
00016 *
00017 *  Purpose
00018 *  =======
00019 *
00020 *  CGELS solves overdetermined or underdetermined complex linear systems
00021 *  involving an M-by-N matrix A, or its conjugate-transpose, using a QR
00022 *  or LQ factorization of A.  It is assumed that A has full rank.
00023 *
00024 *  The following options are provided:
00025 *
00026 *  1. If TRANS = 'N' and m >= n:  find the least squares solution of
00027 *     an overdetermined system, i.e., solve the least squares problem
00028 *                  minimize || B - A*X ||.
00029 *
00030 *  2. If TRANS = 'N' and m < n:  find the minimum norm solution of
00031 *     an underdetermined system A * X = B.
00032 *
00033 *  3. If TRANS = 'C' and m >= n:  find the minimum norm solution of
00034 *     an undetermined system A**H * X = B.
00035 *
00036 *  4. If TRANS = 'C' and m < n:  find the least squares solution of
00037 *     an overdetermined system, i.e., solve the least squares problem
00038 *                  minimize || B - A**H * X ||.
00039 *
00040 *  Several right hand side vectors b and solution vectors x can be
00041 *  handled in a single call; they are stored as the columns of the
00042 *  M-by-NRHS right hand side matrix B and the N-by-NRHS solution
00043 *  matrix X.
00044 *
00045 *  Arguments
00046 *  =========
00047 *
00048 *  TRANS   (input) CHARACTER*1
00049 *          = 'N': the linear system involves A;
00050 *          = 'C': the linear system involves A**H.
00051 *
00052 *  M       (input) INTEGER
00053 *          The number of rows of the matrix A.  M >= 0.
00054 *
00055 *  N       (input) INTEGER
00056 *          The number of columns of the matrix A.  N >= 0.
00057 *
00058 *  NRHS    (input) INTEGER
00059 *          The number of right hand sides, i.e., the number of
00060 *          columns of the matrices B and X. NRHS >= 0.
00061 *
00062 *  A       (input/output) COMPLEX array, dimension (LDA,N)
00063 *          On entry, the M-by-N matrix A.
00064 *            if M >= N, A is overwritten by details of its QR
00065 *                       factorization as returned by CGEQRF;
00066 *            if M <  N, A is overwritten by details of its LQ
00067 *                       factorization as returned by CGELQF.
00068 *
00069 *  LDA     (input) INTEGER
00070 *          The leading dimension of the array A.  LDA >= max(1,M).
00071 *
00072 *  B       (input/output) COMPLEX array, dimension (LDB,NRHS)
00073 *          On entry, the matrix B of right hand side vectors, stored
00074 *          columnwise; B is M-by-NRHS if TRANS = 'N', or N-by-NRHS
00075 *          if TRANS = 'C'.
00076 *          On exit, if INFO = 0, B is overwritten by the solution
00077 *          vectors, stored columnwise:
00078 *          if TRANS = 'N' and m >= n, rows 1 to n of B contain the least
00079 *          squares solution vectors; the residual sum of squares for the
00080 *          solution in each column is given by the sum of squares of the
00081 *          modulus of elements N+1 to M in that column;
00082 *          if TRANS = 'N' and m < n, rows 1 to N of B contain the
00083 *          minimum norm solution vectors;
00084 *          if TRANS = 'C' and m >= n, rows 1 to M of B contain the
00085 *          minimum norm solution vectors;
00086 *          if TRANS = 'C' and m < n, rows 1 to M of B contain the
00087 *          least squares solution vectors; the residual sum of squares
00088 *          for the solution in each column is given by the sum of
00089 *          squares of the modulus of elements M+1 to N in that column.
00090 *
00091 *  LDB     (input) INTEGER
00092 *          The leading dimension of the array B. LDB >= MAX(1,M,N).
00093 *
00094 *  WORK    (workspace/output) COMPLEX array, dimension (MAX(1,LWORK))
00095 *          On exit, if INFO = 0, WORK(1) returns the optimal LWORK.
00096 *
00097 *  LWORK   (input) INTEGER
00098 *          The dimension of the array WORK.
00099 *          LWORK >= max( 1, MN + max( MN, NRHS ) ).
00100 *          For optimal performance,
00101 *          LWORK >= max( 1, MN + max( MN, NRHS )*NB ).
00102 *          where MN = min(M,N) and NB is the optimum block size.
00103 *
00104 *          If LWORK = -1, then a workspace query is assumed; the routine
00105 *          only calculates the optimal size of the WORK array, returns
00106 *          this value as the first entry of the WORK array, and no error
00107 *          message related to LWORK is issued by XERBLA.
00108 *
00109 *  INFO    (output) INTEGER
00110 *          = 0:  successful exit
00111 *          < 0:  if INFO = -i, the i-th argument had an illegal value
00112 *          > 0:  if INFO =  i, the i-th diagonal element of the
00113 *                triangular factor of A is zero, so that A does not have
00114 *                full rank; the least squares solution could not be
00115 *                computed.
00116 *
00117 *  =====================================================================
00118 *
00119 *     .. Parameters ..
00120       REAL               ZERO, ONE
00121       PARAMETER          ( ZERO = 0.0E+0, ONE = 1.0E+0 )
00122       COMPLEX            CZERO
00123       PARAMETER          ( CZERO = ( 0.0E+0, 0.0E+0 ) )
00124 *     ..
00125 *     .. Local Scalars ..
00126       LOGICAL            LQUERY, TPSD
00127       INTEGER            BROW, I, IASCL, IBSCL, J, MN, NB, SCLLEN, WSIZE
00128       REAL               ANRM, BIGNUM, BNRM, SMLNUM
00129 *     ..
00130 *     .. Local Arrays ..
00131       REAL               RWORK( 1 )
00132 *     ..
00133 *     .. External Functions ..
00134       LOGICAL            LSAME
00135       INTEGER            ILAENV
00136       REAL               CLANGE, SLAMCH
00137       EXTERNAL           LSAME, ILAENV, CLANGE, SLAMCH
00138 *     ..
00139 *     .. External Subroutines ..
00140       EXTERNAL           CGELQF, CGEQRF, CLASCL, CLASET, CTRTRS, CUNMLQ,
00141      $                   CUNMQR, SLABAD, XERBLA
00142 *     ..
00143 *     .. Intrinsic Functions ..
00144       INTRINSIC          MAX, MIN, REAL
00145 *     ..
00146 *     .. Executable Statements ..
00147 *
00148 *     Test the input arguments.
00149 *
00150       INFO = 0
00151       MN = MIN( M, N )
00152       LQUERY = ( LWORK.EQ.-1 )
00153       IF( .NOT.( LSAME( TRANS, 'N' ) .OR. LSAME( TRANS, 'C' ) ) ) THEN
00154          INFO = -1
00155       ELSE IF( M.LT.0 ) THEN
00156          INFO = -2
00157       ELSE IF( N.LT.0 ) THEN
00158          INFO = -3
00159       ELSE IF( NRHS.LT.0 ) THEN
00160          INFO = -4
00161       ELSE IF( LDA.LT.MAX( 1, M ) ) THEN
00162          INFO = -6
00163       ELSE IF( LDB.LT.MAX( 1, M, N ) ) THEN
00164          INFO = -8
00165       ELSE IF( LWORK.LT.MAX( 1, MN+MAX( MN, NRHS ) ) .AND.
00166      $   .NOT.LQUERY ) THEN
00167          INFO = -10
00168       END IF
00169 *
00170 *     Figure out optimal block size
00171 *
00172       IF( INFO.EQ.0 .OR. INFO.EQ.-10 ) THEN
00173 *
00174          TPSD = .TRUE.
00175          IF( LSAME( TRANS, 'N' ) )
00176      $      TPSD = .FALSE.
00177 *
00178          IF( M.GE.N ) THEN
00179             NB = ILAENV( 1, 'CGEQRF', ' ', M, N, -1, -1 )
00180             IF( TPSD ) THEN
00181                NB = MAX( NB, ILAENV( 1, 'CUNMQR', 'LN', M, NRHS, N,
00182      $              -1 ) )
00183             ELSE
00184                NB = MAX( NB, ILAENV( 1, 'CUNMQR', 'LC', M, NRHS, N,
00185      $              -1 ) )
00186             END IF
00187          ELSE
00188             NB = ILAENV( 1, 'CGELQF', ' ', M, N, -1, -1 )
00189             IF( TPSD ) THEN
00190                NB = MAX( NB, ILAENV( 1, 'CUNMLQ', 'LC', N, NRHS, M,
00191      $              -1 ) )
00192             ELSE
00193                NB = MAX( NB, ILAENV( 1, 'CUNMLQ', 'LN', N, NRHS, M,
00194      $              -1 ) )
00195             END IF
00196          END IF
00197 *
00198          WSIZE = MAX( 1, MN + MAX( MN, NRHS )*NB )
00199          WORK( 1 ) = REAL( WSIZE )
00200 *
00201       END IF
00202 *
00203       IF( INFO.NE.0 ) THEN
00204          CALL XERBLA( 'CGELS ', -INFO )
00205          RETURN
00206       ELSE IF( LQUERY ) THEN
00207          RETURN
00208       END IF
00209 *
00210 *     Quick return if possible
00211 *
00212       IF( MIN( M, N, NRHS ).EQ.0 ) THEN
00213          CALL CLASET( 'Full', MAX( M, N ), NRHS, CZERO, CZERO, B, LDB )
00214          RETURN
00215       END IF
00216 *
00217 *     Get machine parameters
00218 *
00219       SMLNUM = SLAMCH( 'S' ) / SLAMCH( 'P' )
00220       BIGNUM = ONE / SMLNUM
00221       CALL SLABAD( SMLNUM, BIGNUM )
00222 *
00223 *     Scale A, B if max element outside range [SMLNUM,BIGNUM]
00224 *
00225       ANRM = CLANGE( 'M', M, N, A, LDA, RWORK )
00226       IASCL = 0
00227       IF( ANRM.GT.ZERO .AND. ANRM.LT.SMLNUM ) THEN
00228 *
00229 *        Scale matrix norm up to SMLNUM
00230 *
00231          CALL CLASCL( 'G', 0, 0, ANRM, SMLNUM, M, N, A, LDA, INFO )
00232          IASCL = 1
00233       ELSE IF( ANRM.GT.BIGNUM ) THEN
00234 *
00235 *        Scale matrix norm down to BIGNUM
00236 *
00237          CALL CLASCL( 'G', 0, 0, ANRM, BIGNUM, M, N, A, LDA, INFO )
00238          IASCL = 2
00239       ELSE IF( ANRM.EQ.ZERO ) THEN
00240 *
00241 *        Matrix all zero. Return zero solution.
00242 *
00243          CALL CLASET( 'F', MAX( M, N ), NRHS, CZERO, CZERO, B, LDB )
00244          GO TO 50
00245       END IF
00246 *
00247       BROW = M
00248       IF( TPSD )
00249      $   BROW = N
00250       BNRM = CLANGE( 'M', BROW, NRHS, B, LDB, RWORK )
00251       IBSCL = 0
00252       IF( BNRM.GT.ZERO .AND. BNRM.LT.SMLNUM ) THEN
00253 *
00254 *        Scale matrix norm up to SMLNUM
00255 *
00256          CALL CLASCL( 'G', 0, 0, BNRM, SMLNUM, BROW, NRHS, B, LDB,
00257      $                INFO )
00258          IBSCL = 1
00259       ELSE IF( BNRM.GT.BIGNUM ) THEN
00260 *
00261 *        Scale matrix norm down to BIGNUM
00262 *
00263          CALL CLASCL( 'G', 0, 0, BNRM, BIGNUM, BROW, NRHS, B, LDB,
00264      $                INFO )
00265          IBSCL = 2
00266       END IF
00267 *
00268       IF( M.GE.N ) THEN
00269 *
00270 *        compute QR factorization of A
00271 *
00272          CALL CGEQRF( M, N, A, LDA, WORK( 1 ), WORK( MN+1 ), LWORK-MN,
00273      $                INFO )
00274 *
00275 *        workspace at least N, optimally N*NB
00276 *
00277          IF( .NOT.TPSD ) THEN
00278 *
00279 *           Least-Squares Problem min || A * X - B ||
00280 *
00281 *           B(1:M,1:NRHS) := Q' * B(1:M,1:NRHS)
00282 *
00283             CALL CUNMQR( 'Left', 'Conjugate transpose', M, NRHS, N, A,
00284      $                   LDA, WORK( 1 ), B, LDB, WORK( MN+1 ), LWORK-MN,
00285      $                   INFO )
00286 *
00287 *           workspace at least NRHS, optimally NRHS*NB
00288 *
00289 *           B(1:N,1:NRHS) := inv(R) * B(1:N,1:NRHS)
00290 *
00291             CALL CTRTRS( 'Upper', 'No transpose', 'Non-unit', N, NRHS,
00292      $                   A, LDA, B, LDB, INFO )
00293 *
00294             IF( INFO.GT.0 ) THEN
00295                RETURN
00296             END IF
00297 *
00298             SCLLEN = N
00299 *
00300          ELSE
00301 *
00302 *           Overdetermined system of equations A' * X = B
00303 *
00304 *           B(1:N,1:NRHS) := inv(R') * B(1:N,1:NRHS)
00305 *
00306             CALL CTRTRS( 'Upper', 'Conjugate transpose','Non-unit',
00307      $                   N, NRHS, A, LDA, B, LDB, INFO )
00308 *
00309             IF( INFO.GT.0 ) THEN
00310                RETURN
00311             END IF
00312 *
00313 *           B(N+1:M,1:NRHS) = ZERO
00314 *
00315             DO 20 J = 1, NRHS
00316                DO 10 I = N + 1, M
00317                   B( I, J ) = CZERO
00318    10          CONTINUE
00319    20       CONTINUE
00320 *
00321 *           B(1:M,1:NRHS) := Q(1:N,:) * B(1:N,1:NRHS)
00322 *
00323             CALL CUNMQR( 'Left', 'No transpose', M, NRHS, N, A, LDA,
00324      $                   WORK( 1 ), B, LDB, WORK( MN+1 ), LWORK-MN,
00325      $                   INFO )
00326 *
00327 *           workspace at least NRHS, optimally NRHS*NB
00328 *
00329             SCLLEN = M
00330 *
00331          END IF
00332 *
00333       ELSE
00334 *
00335 *        Compute LQ factorization of A
00336 *
00337          CALL CGELQF( M, N, A, LDA, WORK( 1 ), WORK( MN+1 ), LWORK-MN,
00338      $                INFO )
00339 *
00340 *        workspace at least M, optimally M*NB.
00341 *
00342          IF( .NOT.TPSD ) THEN
00343 *
00344 *           underdetermined system of equations A * X = B
00345 *
00346 *           B(1:M,1:NRHS) := inv(L) * B(1:M,1:NRHS)
00347 *
00348             CALL CTRTRS( 'Lower', 'No transpose', 'Non-unit', M, NRHS,
00349      $                   A, LDA, B, LDB, INFO )
00350 *
00351             IF( INFO.GT.0 ) THEN
00352                RETURN
00353             END IF
00354 *
00355 *           B(M+1:N,1:NRHS) = 0
00356 *
00357             DO 40 J = 1, NRHS
00358                DO 30 I = M + 1, N
00359                   B( I, J ) = CZERO
00360    30          CONTINUE
00361    40       CONTINUE
00362 *
00363 *           B(1:N,1:NRHS) := Q(1:N,:)' * B(1:M,1:NRHS)
00364 *
00365             CALL CUNMLQ( 'Left', 'Conjugate transpose', N, NRHS, M, A,
00366      $                   LDA, WORK( 1 ), B, LDB, WORK( MN+1 ), LWORK-MN,
00367      $                   INFO )
00368 *
00369 *           workspace at least NRHS, optimally NRHS*NB
00370 *
00371             SCLLEN = N
00372 *
00373          ELSE
00374 *
00375 *           overdetermined system min || A' * X - B ||
00376 *
00377 *           B(1:N,1:NRHS) := Q * B(1:N,1:NRHS)
00378 *
00379             CALL CUNMLQ( 'Left', 'No transpose', N, NRHS, M, A, LDA,
00380      $                   WORK( 1 ), B, LDB, WORK( MN+1 ), LWORK-MN,
00381      $                   INFO )
00382 *
00383 *           workspace at least NRHS, optimally NRHS*NB
00384 *
00385 *           B(1:M,1:NRHS) := inv(L') * B(1:M,1:NRHS)
00386 *
00387             CALL CTRTRS( 'Lower', 'Conjugate transpose', 'Non-unit',
00388      $                   M, NRHS, A, LDA, B, LDB, INFO )
00389 *
00390             IF( INFO.GT.0 ) THEN
00391                RETURN
00392             END IF
00393 *
00394             SCLLEN = M
00395 *
00396          END IF
00397 *
00398       END IF
00399 *
00400 *     Undo scaling
00401 *
00402       IF( IASCL.EQ.1 ) THEN
00403          CALL CLASCL( 'G', 0, 0, ANRM, SMLNUM, SCLLEN, NRHS, B, LDB,
00404      $                INFO )
00405       ELSE IF( IASCL.EQ.2 ) THEN
00406          CALL CLASCL( 'G', 0, 0, ANRM, BIGNUM, SCLLEN, NRHS, B, LDB,
00407      $                INFO )
00408       END IF
00409       IF( IBSCL.EQ.1 ) THEN
00410          CALL CLASCL( 'G', 0, 0, SMLNUM, BNRM, SCLLEN, NRHS, B, LDB,
00411      $                INFO )
00412       ELSE IF( IBSCL.EQ.2 ) THEN
00413          CALL CLASCL( 'G', 0, 0, BIGNUM, BNRM, SCLLEN, NRHS, B, LDB,
00414      $                INFO )
00415       END IF
00416 *
00417    50 CONTINUE
00418       WORK( 1 ) = REAL( WSIZE )
00419 *
00420       RETURN
00421 *
00422 *     End of CGELS
00423 *
00424       END