Purpose
To calculate a QR factorization of the first block column and
apply the orthogonal transformations (from the left) also to the
second block column of a structured matrix, as follows
_
[ R 0 ] [ R C ]
Q' * [ ] = [ ]
[ A B ] [ 0 D ]
_
where R and R are upper triangular. The matrix A can be full or
upper trapezoidal/triangular. The problem structure is exploited.
This computation is useful, for instance, in combined measurement
and time update of one iteration of the Kalman filter (square
root information filter).
Specification
SUBROUTINE MB04KD( UPLO, N, M, P, R, LDR, A, LDA, B, LDB, C, LDC,
$ TAU, DWORK )
C .. Scalar Arguments ..
CHARACTER UPLO
INTEGER LDA, LDB, LDC, LDR, M, N, P
C .. Array Arguments ..
DOUBLE PRECISION A(LDA,*), B(LDB,*), C(LDC,*), DWORK(*),
$ R(LDR,*), TAU(*)
Arguments
Mode Parameters
UPLO CHARACTER*1
Indicates if the matrix A is or not triangular as follows:
= 'U': Matrix A is upper trapezoidal/triangular;
= 'F': Matrix A is full.
Input/Output Parameters
N (input) INTEGER _
The order of the matrices R and R. N >= 0.
M (input) INTEGER
The number of columns of the matrices B, C and D. M >= 0.
P (input) INTEGER
The number of rows of the matrices A, B and D. P >= 0.
R (input/output) DOUBLE PRECISION array, dimension (LDR,N)
On entry, the leading N-by-N upper triangular part of this
array must contain the upper triangular matrix R.
On exit, the leading N-by-N upper triangular part of this
_
array contains the upper triangular matrix R.
The strict lower triangular part of this array is not
referenced.
LDR INTEGER
The leading dimension of array R. LDR >= MAX(1,N).
A (input/output) DOUBLE PRECISION array, dimension (LDA,N)
On entry, if UPLO = 'F', the leading P-by-N part of this
array must contain the matrix A. If UPLO = 'U', the
leading MIN(P,N)-by-N part of this array must contain the
upper trapezoidal (upper triangular if P >= N) matrix A,
and the elements below the diagonal are not referenced.
On exit, the leading P-by-N part (upper trapezoidal or
triangular, if UPLO = 'U') of this array contains the
trailing components (the vectors v, see Method) of the
elementary reflectors used in the factorization.
LDA INTEGER
The leading dimension of array A. LDA >= MAX(1,P).
B (input/output) DOUBLE PRECISION array, dimension (LDB,M)
On entry, the leading P-by-M part of this array must
contain the matrix B.
On exit, the leading P-by-M part of this array contains
the computed matrix D.
LDB INTEGER
The leading dimension of array B. LDB >= MAX(1,P).
C (output) DOUBLE PRECISION array, dimension (LDC,M)
The leading N-by-M part of this array contains the
computed matrix C.
LDC INTEGER
The leading dimension of array C. LDC >= MAX(1,N).
TAU (output) DOUBLE PRECISION array, dimension (N)
The scalar factors of the elementary reflectors used.
Workspace
DWORK DOUBLE PRECISION array, dimension (N)Method
The routine uses N Householder transformations exploiting the zero
pattern of the block matrix. A Householder matrix has the form
( 1 ),
H = I - tau *u *u', u = ( v )
i i i i i ( i)
where v is a P-vector, if UPLO = 'F', or an min(i,P)-vector, if
i
UPLO = 'U'. The components of v are stored in the i-th column
i
of A, and tau is stored in TAU(i).
i
Numerical Aspects
The algorithm is backward stable.Further Comments
NoneExample
Program Text
NoneProgram Data
NoneProgram Results
None
Click here to get a compressed (gzip) tar file containing the source code of the routine, the example program, data, documentation, and related files.
Return to index