1DGELSY(1)             LAPACK driver routine (version 3.1)            DGELSY(1)
2
3
4

NAME

6       DGELSY - the minimum-norm solution to a real linear least squares prob‐
7       lem
8

SYNOPSIS

10       SUBROUTINE DGELSY( M, N, NRHS, A, LDA, B, LDB, JPVT, RCOND, RANK, WORK,
11                          LWORK, INFO )
12
13           INTEGER        INFO, LDA, LDB, LWORK, M, N, NRHS, RANK
14
15           DOUBLE         PRECISION RCOND
16
17           INTEGER        JPVT( * )
18
19           DOUBLE         PRECISION A( LDA, * ), B( LDB, * ), WORK( * )
20

PURPOSE

22       DGELSY  computes  the  minimum-norm  solution  to  a  real linear least
23       squares problem:
24           minimize || A * X - B ||
25       using a complete orthogonal factorization of A.  A is an M-by-N  matrix
26       which may be rank-deficient.
27
28       Several right hand side vectors b and solution vectors x can be handled
29       in a single call; they are stored as the columns of the M-by-NRHS right
30       hand side matrix B and the N-by-NRHS solution matrix X.
31
32       The routine first computes a QR factorization with column pivoting:
33           A * P = Q * [ R11 R12 ]
34                       [  0  R22 ]
35       with  R11 defined as the largest leading submatrix whose estimated con‐
36       dition number is less than 1/RCOND.  The order of  R11,  RANK,  is  the
37       effective rank of A.
38
39       Then,  R22  is  considered  to be negligible, and R12 is annihilated by
40       orthogonal transformations from the right,  arriving  at  the  complete
41       orthogonal factorization:
42          A * P = Q * [ T11 0 ] * Z
43                      [  0  0 ]
44       The minimum-norm solution is then
45          X = P * Z' [ inv(T11)*Q1'*B ]
46                     [        0       ]
47       where Q1 consists of the first RANK columns of Q.
48
49       This routine is basically identical to the original xGELSX except three
50       differences:
51         o The call to the subroutine xGEQPF has been substituted by the
52           the call to the subroutine xGEQP3. This subroutine is a Blas-3
53           version of the QR factorization with column pivoting.
54         o Matrix B (the right hand side) is updated with Blas-3.
55         o The permutation of matrix B (the right hand side) is faster and
56           more simple.
57
58

ARGUMENTS

60       M       (input) INTEGER
61               The number of rows of the matrix A.  M >= 0.
62
63       N       (input) INTEGER
64               The number of columns of the matrix A.  N >= 0.
65
66       NRHS    (input) INTEGER
67               The number of right hand sides, i.e., the number of columns  of
68               matrices B and X. NRHS >= 0.
69
70       A       (input/output) DOUBLE PRECISION array, dimension (LDA,N)
71               On entry, the M-by-N matrix A.  On exit, A has been overwritten
72               by details of its complete orthogonal factorization.
73
74       LDA     (input) INTEGER
75               The leading dimension of the array A.  LDA >= max(1,M).
76
77       B       (input/output) DOUBLE PRECISION array, dimension (LDB,NRHS)
78               On entry, the M-by-NRHS right hand side matrix B.  On exit, the
79               N-by-NRHS solution matrix X.
80
81       LDB     (input) INTEGER
82               The leading dimension of the array B. LDB >= max(1,M,N).
83
84       JPVT    (input/output) INTEGER array, dimension (N)
85               On  entry,  if JPVT(i) .ne. 0, the i-th column of A is permuted
86               to the front of AP, otherwise column i is a  free  column.   On
87               exit,  if  JPVT(i) = k, then the i-th column of AP was the k-th
88               column of A.
89
90       RCOND   (input) DOUBLE PRECISION
91               RCOND is used to determine the effective rank of  A,  which  is
92               defined  as  the order of the largest leading triangular subma‐
93               trix R11 in the QR factorization  with  pivoting  of  A,  whose
94               estimated condition number < 1/RCOND.
95
96       RANK    (output) INTEGER
97               The  effective rank of A, i.e., the order of the submatrix R11.
98               This is the same as the order of the submatrix T11 in the  com‐
99               plete orthogonal factorization of A.
100
101       WORK       (workspace/output)   DOUBLE   PRECISION   array,   dimension
102       (MAX(1,LWORK))
103               On exit, if INFO = 0, WORK(1) returns the optimal LWORK.
104
105       LWORK   (input) INTEGER
106               The dimension  of  the  array  WORK.   The  unblocked  strategy
107               requires  that: LWORK >= MAX( MN+3*N+1, 2*MN+NRHS ), where MN =
108               min( M, N ).  The block algorithm requires that: LWORK >=  MAX(
109               MN+2*N+NB*(N+1),  2*MN+NB*NRHS ), where NB is an upper bound on
110               the blocksize returned  by  ILAENV  for  the  routines  DGEQP3,
111               DTZRZF, STZRQF, DORMQR, and DORMRZ.
112
113               If  LWORK  = -1, then a workspace query is assumed; the routine
114               only calculates the optimal size of  the  WORK  array,  returns
115               this  value  as the first entry of the WORK array, and no error
116               message related to LWORK is issued by XERBLA.
117
118       INFO    (output) INTEGER
119               = 0: successful exit
120               < 0: If INFO = -i, the i-th argument had an illegal value.
121

FURTHER DETAILS

123       Based on contributions by
124         A. Petitet, Computer Science Dept., Univ. of Tenn., Knoxville, USA
125         E. Quintana-Orti, Depto. de Informatica, Universidad Jaime I, Spain
126         G. Quintana-Orti, Depto. de Informatica, Universidad Jaime I, Spain
127
128
129
130
131 LAPACK driver routine (version 3.N1o)vember 2006                       DGELSY(1)
Impressum