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

NAME

6       ZGELSY  -  the  minimum-norm solution to a complex linear least squares
7       problem
8

SYNOPSIS

10       SUBROUTINE ZGELSY( M, N, NRHS, A, LDA, B, LDB, JPVT, RCOND, RANK, WORK,
11                          LWORK, RWORK, 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 RWORK( * )
20
21           COMPLEX*16     A( LDA, * ), B( LDB, * ), WORK( * )
22

PURPOSE

24       ZGELSY  computes  the  minimum-norm  solution to a complex linear least
25       squares problem:
26           minimize || A * X - B ||
27       using a complete orthogonal factorization of A.  A is an M-by-N  matrix
28       which may be rank-deficient.
29
30       Several right hand side vectors b and solution vectors x can be handled
31       in a single call; they are stored as the columns of the M-by-NRHS right
32       hand side matrix B and the N-by-NRHS solution matrix X.
33
34       The routine first computes a QR factorization with column pivoting:
35           A * P = Q * [ R11 R12 ]
36                       [  0  R22 ]
37       with  R11 defined as the largest leading submatrix whose estimated con‐
38       dition number is less than 1/RCOND.  The order of  R11,  RANK,  is  the
39       effective rank of A.
40
41       Then,  R22  is  considered  to be negligible, and R12 is annihilated by
42       unitary transformations  from  the  right,  arriving  at  the  complete
43       orthogonal factorization:
44          A * P = Q * [ T11 0 ] * Z
45                      [  0  0 ]
46       The minimum-norm solution is then
47          X = P * Z' [ inv(T11)*Q1'*B ]
48                     [        0       ]
49       where Q1 consists of the first RANK columns of Q.
50
51       This routine is basically identical to the original xGELSX except three
52       differences:
53         o The permutation of matrix B (the right hand side) is faster and
54           more simple.
55         o The call to the subroutine xGEQPF has been substituted by the
56           the call to the subroutine xGEQP3. This subroutine is a Blas-3
57           version of the QR factorization with column pivoting.
58         o Matrix B (the right hand side) is updated with Blas-3.
59
60

ARGUMENTS

62       M       (input) INTEGER
63               The number of rows of the matrix A.  M >= 0.
64
65       N       (input) INTEGER
66               The number of columns of the matrix A.  N >= 0.
67
68       NRHS    (input) INTEGER
69               The number of right hand sides, i.e., the number of columns  of
70               matrices B and X. NRHS >= 0.
71
72       A       (input/output) COMPLEX*16 array, dimension (LDA,N)
73               On entry, the M-by-N matrix A.  On exit, A has been overwritten
74               by details of its complete orthogonal factorization.
75
76       LDA     (input) INTEGER
77               The leading dimension of the array A.  LDA >= max(1,M).
78
79       B       (input/output) COMPLEX*16 array, dimension (LDB,NRHS)
80               On entry, the M-by-NRHS right hand side matrix B.  On exit, the
81               N-by-NRHS solution matrix X.
82
83       LDB     (input) INTEGER
84               The leading dimension of the array B. LDB >= max(1,M,N).
85
86       JPVT    (input/output) INTEGER array, dimension (N)
87               On  entry,  if JPVT(i) .ne. 0, the i-th column of A is permuted
88               to the front of AP, otherwise column i is a  free  column.   On
89               exit,  if JPVT(i) = k, then the i-th column of A*P was the k-th
90               column of A.
91
92       RCOND   (input) DOUBLE PRECISION
93               RCOND is used to determine the effective rank of  A,  which  is
94               defined  as  the order of the largest leading triangular subma‐
95               trix R11 in the QR factorization  with  pivoting  of  A,  whose
96               estimated condition number < 1/RCOND.
97
98       RANK    (output) INTEGER
99               The  effective rank of A, i.e., the order of the submatrix R11.
100               This is the same as the order of the submatrix T11 in the  com‐
101               plete orthogonal factorization of A.
102
103       WORK    (workspace/output) COMPLEX*16 array, dimension (MAX(1,LWORK))
104               On exit, if INFO = 0, WORK(1) returns the optimal LWORK.
105
106       LWORK   (input) INTEGER
107               The  dimension  of  the  array  WORK.   The  unblocked strategy
108               requires that: LWORK >= MN + MAX( 2*MN, N+1, MN+NRHS ) where MN
109               =  min(M,N).   The block algorithm requires that: LWORK >= MN +
110               MAX( 2*MN, NB*(N+1), MN+MN*NB, MN+NB*NRHS  )  where  NB  is  an
111               upper  bound  on  the blocksize returned by ILAENV for the rou‐
112               tines ZGEQP3, ZTZRZF, CTZRQF, ZUNMQR, and ZUNMRZ.
113
114               If LWORK = -1, then a workspace query is assumed;  the  routine
115               only  calculates  the  optimal  size of the WORK array, returns
116               this value as the first entry of the WORK array, and  no  error
117               message related to LWORK is issued by XERBLA.
118
119       RWORK   (workspace) DOUBLE PRECISION array, dimension (2*N)
120
121       INFO    (output) INTEGER
122               = 0: successful exit
123               < 0: if INFO = -i, the i-th argument had an illegal value
124

FURTHER DETAILS

126       Based on contributions by
127         A. Petitet, Computer Science Dept., Univ. of Tenn., Knoxville, USA
128         E. Quintana-Orti, Depto. de Informatica, Universidad Jaime I, Spain
129         G. Quintana-Orti, Depto. de Informatica, Universidad Jaime I, Spain
130
131
132
133
134 LAPACK driver routine (version 3.N1o)vember 2006                       ZGELSY(1)
Impressum