Integer, Intent (In)	::	n, ldu, ldvt
Integer, Intent (Inout)	::	iq(*)
Integer, Intent (Out)	::	iwork(8*n), info
Real (Kind=nag_wp), Intent (Inout)	::	d(), e(), u(ldu,), vt(ldvt,), q(), work()
Character (1), Intent (In)	::	uplo, compq

C Header Interface

#include nagmk26.h

void	f08mdf_ ( const char uplo, const char compq, const Integer n, double d[], double e[], double u[], const Integer ldu, double vt[], const Integer ldvt, double q[], Integer iq[], double work[], Integer iwork[], Integer info, const Charlen length_uplo, const Charlen length_compq)

The routine may be called by its LAPACK name dbdsdc.

3

Description

f08mdf (dbdsdc) computes the singular value decomposition (SVD) of the (upper or lower) bidiagonal matrix

B

B = U S V^{T},

where

S

is a diagonal matrix with non-negative diagonal elements

s_{i i} = s_{i}

, such that

s_{1} \geq s_{2} \geq \dots \geq s_{n} \geq 0,

and

U

and

V

are orthogonal matrices. The diagonal elements of

S

are the singular values of

B

and the columns of

U

and

V

are respectively the corresponding left and right singular vectors of

B

When only singular values are required the routine uses the

Q R

algorithm, but when singular vectors are required a divide and conquer method is used. The singular values can optionally be returned in compact form, although currently no routine is available to apply

U

V

when stored in compact form.

4

References

Anderson E, Bai Z, Bischof C, Blackford S, Demmel J, Dongarra J J, Du Croz J J, Greenbaum A, Hammarling S, McKenney A and Sorensen D (1999) LAPACK Users' Guide (3rd Edition) SIAM, Philadelphia http://www.netlib.org/lapack/lug

Golub G H and Van Loan C F (1996) Matrix Computations (3rd Edition) Johns Hopkins University Press, Baltimore

5

Arguments

1: $uplo$ – Character(1)Input

On entry: indicates whether

B

is upper or lower bidiagonal.

$uplo ='U'$: $B$ is upper bidiagonal.
$uplo ='L'$: $B$ is lower bidiagonal.

Constraint:

uplo ='U'

'L'

2: $compq$ – Character(1)Input

On entry: specifies whether singular vectors are to be computed.

$compq ='N'$: Compute singular values only.
$compq ='P'$: Compute singular values and compute singular vectors in compact form.
$compq ='I'$: Compute singular values and singular vectors.

Constraint:

compq ='N'

'P'

'I'

3: $n$ – IntegerInput

On entry:

n

, the order of the matrix

B

Constraint:

n \geq 0

4: $d (*)$ – Real (Kind=nag_wp) arrayInput/Output

Note: the dimension of the array d must be at least

\max (1, n)

On entry: the

n

diagonal elements of the bidiagonal matrix

B

On exit: if

info = 0

, the singular values of

B

5: $e (*)$ – Real (Kind=nag_wp) arrayInput/Output

Note: the dimension of the array e must be at least

\max (1, n - 1)

On entry: the

(n - 1)

off-diagonal elements of the bidiagonal matrix

B

On exit: the contents of e are destroyed.

6: $u (ldu *)$ – Real (Kind=nag_wp) arrayOutput

Note: the second dimension of the array u must be at least

\max (1, n)

compq ='I'

, and at least

1

otherwise.

On exit: if

compq ='I'

, then if

info = 0

, u contains the left singular vectors of the bidiagonal matrix

B

compq \neq'I'

, u is not referenced.

7: $ldu$ – IntegerInput

On entry: the first dimension of the array u as declared in the (sub)program from which f08mdf (dbdsdc) is called.

Constraints:

if $compq ='I'$ , $ldu \geq \max (1, n)$ ;
otherwise $ldu \geq 1$ .

8: $vt (ldvt *)$ – Real (Kind=nag_wp) arrayOutput

Note: the second dimension of the array vt must be at least

\max (1, n)

compq ='I'

, and at least

1

otherwise.

On exit: if

compq ='I'

, then if

info = 0

, the rows of vt contain the right singular vectors of the bidiagonal matrix

B

compq \neq'I'

, vt is not referenced.

9: $ldvt$ – IntegerInput

On entry: the first dimension of the array vt as declared in the (sub)program from which f08mdf (dbdsdc) is called.

Constraints:

if $compq ='I'$ , $ldvt \geq \max (1, n)$ ;
otherwise $ldvt \geq 1$ .

10: $q (*)$ – Real (Kind=nag_wp) arrayOutput

Note: the dimension of the array q must be at least

\max (1, n^{2} + 5 n, ldq)

On exit: if

compq ='P'

, then if

info = 0

, q and iq contain the left and right singular vectors in a compact form, requiring

O (n \log_{2} n)

space instead of

2 \times n^{2}

. In particular, q contains all the real data in the first

ldq = n \times (11 + 2 \times smlsiz + 8 \times int (\log_{2} (n / (smlsiz + 1))))

elements of q, where

smlsiz

is equal to the maximum size of the subproblems at the bottom of the computation tree (usually about

25

compq \neq'P'

, q is not referenced.

11: $iq (*)$ – Integer arrayOutput

Note: the dimension of the array iq must be at least

\max (1, ldiq)

On exit: if

compq ='P'

, then if

info = 0

, q and iq contain the left and right singular vectors in a compact form, requiring

O (n \log_{2} n)

space instead of

2 \times n^{2}

. In particular, iq contains all integer data in the first

ldiq = n \times (3 + 3 \times int (\log_{2} (n / (smlsiz + 1))))

elements of iq, where

smlsiz

is equal to the maximum size of the subproblems at the bottom of the computation tree (usually about

25

compq \neq'P'

, iq is not referenced.

12: $work (*)$ – Real (Kind=nag_wp) arrayWorkspace

Note: the dimension of the array work must be at least

\max (1, 6 \times n - 2)

compq ='N'

\max (1, 6 \times n)

compq ='P'

\max (1, 3 \times n^{2} + 4 \times n)

compq ='I'

, and at least

1

otherwise.

13: $iwork (8 \times n)$ – Integer arrayWorkspace

14: $info$ – IntegerOutput

On exit:

info = 0

unless the routine detects an error (see Section 6).

6

Error Indicators and Warnings

$info < 0$: If $info = - i$ , argument $i$ had an illegal value. An explanatory message is output, and execution of the program is terminated.

$info > 0$: The algorithm failed to compute a singular value. The update process of divide-and-conquer failed.

7

Accuracy

Each computed singular value of

B

is accurate to nearly full relative precision, no matter how tiny the singular value. The

i

th computed singular value,

{\hat{s}}_{i}

, satisfies the bound

|{\hat{s}}_{i} - s_{i}| \leq p (n) ε s_{i}

where

ε

is the machine precision and

p (n)

is a modest function of

n

For bounds on the computed singular values, see Section 4.9.1 of Anderson et al. (1999). See also f08flf (ddisna).

8

Parallelism and Performance

f08mdf (dbdsdc) is threaded by NAG for parallel execution in multithreaded implementations of the NAG Library.

f08mdf (dbdsdc) makes calls to BLAS and/or LAPACK routines, which may be threaded within the vendor library used by this implementation. Consult the documentation for the vendor library for further information.

Please consult the X06 Chapter Introduction for information on how to control and interrogate the OpenMP environment used within this routine. Please also consult the Users' Note for your implementation for any additional implementation-specific information.

9

Further Comments

If only singular values are required, the total number of floating-point operations is approximately proportional to

n^{2}

. When singular vectors are required the number of operations is bounded above by approximately the same number of operations as f08mef (dbdsqr), but for large matrices f08mdf (dbdsdc) is usually much faster.

There is no complex analogue of f08mdf (dbdsdc).

10

Example

This example computes the singular value decomposition of the upper bidiagonal matrix

B = (\begin{array}{r} 3.62 & 1.26 & 0 & 0 \\ 0 & - 2.41 & - 1.53 & 0 \\ 0 & 0 & 1.92 & 1.19 \\ 0 & 0 & 0 & - 1.43 \end{array}) .

NAG Library Routine Document

f08mdf (dbdsdc)

▸▿ Contents

1 Purpose

2 Specification

3 Description

4 References

5 Arguments

6 Error Indicators and Warnings

7 Accuracy

8 Parallelism and Performance

9 Further Comments

10 Example

10.1 Program Text

10.2 Program Data

10.3 Program Results

1

Purpose

2

Specification

3

Description

4

References

5

Arguments

6

Error Indicators and Warnings

7

Accuracy

8

Parallelism and Performance

9

Further Comments

10

Example

10.1

Program Text

10.2

Program Data

10.3

Program Results