1SGELSY(1)             LAPACK driver routine (version 3.2)            SGELSY(1)
2
3
4

NAME

6       SGELSY  -  computes  the  minimum-norm  solution to a real linear least
7       squares problem
8

SYNOPSIS

10       SUBROUTINE SGELSY( 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           REAL           RCOND
16
17           INTEGER        JPVT( * )
18
19           REAL           A( LDA, * ), B( LDB, * ), WORK( * )
20

PURPOSE

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

ARGUMENTS

55       M       (input) INTEGER
56               The number of rows of the matrix A.  M >= 0.
57
58       N       (input) INTEGER
59               The number of columns of the matrix A.  N >= 0.
60
61       NRHS    (input) INTEGER
62               The number of right hand sides, i.e., the number of columns  of
63               matrices B and X. NRHS >= 0.
64
65       A       (input/output) REAL array, dimension (LDA,N)
66               On entry, the M-by-N matrix A.  On exit, A has been overwritten
67               by details of its complete orthogonal factorization.
68
69       LDA     (input) INTEGER
70               The leading dimension of the array A.  LDA >= max(1,M).
71
72       B       (input/output) REAL array, dimension (LDB,NRHS)
73               On entry, the M-by-NRHS right hand side matrix B.  On exit, the
74               N-by-NRHS solution matrix X.
75
76       LDB     (input) INTEGER
77               The leading dimension of the array B. LDB >= max(1,M,N).
78
79       JPVT    (input/output) INTEGER array, dimension (N)
80               On  entry,  if JPVT(i) .ne. 0, the i-th column of A is permuted
81               to the front of AP, otherwise column i is a  free  column.   On
82               exit,  if  JPVT(i) = k, then the i-th column of AP was the k-th
83               column of A.
84
85       RCOND   (input) REAL
86               RCOND is used to determine the effective rank of  A,  which  is
87               defined  as  the order of the largest leading triangular subma‐
88               trix R11 in the QR factorization  with  pivoting  of  A,  whose
89               estimated condition number < 1/RCOND.
90
91       RANK    (output) INTEGER
92               The  effective rank of A, i.e., the order of the submatrix R11.
93               This is the same as the order of the submatrix T11 in the  com‐
94               plete orthogonal factorization of A.
95
96       WORK    (workspace/output) REAL array, dimension (MAX(1,LWORK))
97               On exit, if INFO = 0, WORK(1) returns the optimal LWORK.
98
99       LWORK   (input) INTEGER
100               The  dimension  of  the  array  WORK.   The  unblocked strategy
101               requires that: LWORK >= MAX( MN+3*N+1, 2*MN+NRHS ), where MN  =
102               min(  M, N ).  The block algorithm requires that: LWORK >= MAX(
103               MN+2*N+NB*(N+1), 2*MN+NB*NRHS ), where NB is an upper bound  on
104               the  blocksize  returned  by  ILAENV  for  the routines SGEQP3,
105               STZRZF, STZRQF, SORMQR, and SORMRZ.  If  LWORK  =  -1,  then  a
106               workspace  query  is  assumed;  the routine only calculates the
107               optimal size of the WORK array, returns this value as the first
108               entry  of the WORK array, and no error message related to LWORK
109               is issued by XERBLA.
110
111       INFO    (output) INTEGER
112               = 0: successful exit
113               < 0: If INFO = -i, the i-th argument had an illegal value.
114

FURTHER DETAILS

116       Based on contributions by
117         A. Petitet, Computer Science Dept., Univ. of Tenn., Knoxville, USA
118         E. Quintana-Orti, Depto. de Informatica, Universidad Jaime I, Spain
119         G. Quintana-Orti, Depto. de Informatica, Universidad Jaime I, Spain
120
121
122
123 LAPACK driver routine (version 3.N2o)vember 2008                       SGELSY(1)
Impressum