f08jgc computes all the eigenvalues and, optionally, all the eigenvectors of a real symmetric positive definite tridiagonal matrix, or of a real symmetric positive definite matrix which has been reduced to tridiagonal form.
The function may be called by the names: f08jgc, nag_lapackeig_dpteqr or nag_dpteqr.
3Description
f08jgc computes all the eigenvalues and, optionally, all the eigenvectors of a real symmetric positive definite tridiagonal matrix .
In other words, it can compute the spectral factorization of as
where is a diagonal matrix whose diagonal elements are the eigenvalues , and is the orthogonal matrix whose columns are the eigenvectors . Thus
The function may also be used to compute all the eigenvalues and eigenvectors of a real symmetric positive definite matrix which has been reduced to tridiagonal form :
In this case, the matrix must be formed explicitly and passed to f08jgc, which must be called with . The functions which must be called to perform the reduction to tridiagonal form and form are:
f08jgc first factorizes as where is unit lower bidiagonal and is diagonal. It forms the bidiagonal matrix , and then calls f08mec to compute the singular values of which are the same as the eigenvalues of . The method used by the function allows high relative accuracy to be achieved in the small eigenvalues of . The eigenvectors are normalized so that , but are determined only to within a factor .
4References
Barlow J and Demmel J W (1990) Computing accurate eigensystems of scaled diagonally dominant matrices SIAM J. Numer. Anal.27 762–791
5Arguments
1: – Nag_OrderTypeInput
On entry: the order argument specifies the two-dimensional storage scheme being used, i.e., row-major ordering or column-major ordering. C language defined storage is specified by . See Section 3.1.3 in the Introduction to the NAG Library CL Interface for a more detailed explanation of the use of this argument.
Constraint:
or .
2: – Nag_ComputeZTypeInput
On entry: indicates whether the eigenvectors are to be computed.
Only the eigenvalues are computed (and the array z is not referenced).
The eigenvalues and eigenvectors of are computed (and the array z must contain the matrix on entry).
The eigenvalues and eigenvectors of are computed (and the array z is initialized by the function).
Constraint:
, or .
3: – IntegerInput
On entry: , the order of the matrix .
Constraint:
.
4: – doubleInput/Output
Note: the dimension, dim, of the array d
must be at least
.
On entry: the diagonal elements of the tridiagonal matrix .
On exit: the eigenvalues in descending order, unless NE_CONVERGENCE or NE_POS_DEF, in which case d is overwritten.
5: – doubleInput/Output
Note: the dimension, dim, of the array e
must be at least
.
On entry: the off-diagonal elements of the tridiagonal matrix .
On exit: if or , the required orthonormal eigenvectors stored as columns of ; the th column corresponds to the th eigenvalue, where , unless NE_CONVERGENCE or NE_POS_DEF.
On entry: the stride separating row or column elements (depending on the value of order) in the array z.
Constraints:
if or , ;
if , .
8: – NagError *Input/Output
The NAG error argument (see Section 7 in the Introduction to the NAG Library CL Interface).
6Error Indicators and Warnings
NE_ALLOC_FAIL
Dynamic memory allocation failed.
See Section 3.1.2 in the Introduction to the NAG Library CL Interface for further information.
NE_BAD_PARAM
On entry, argument had an illegal value.
NE_CONVERGENCE
The algorithm to compute the singular values of the Cholesky factor failed to converge; off-diagonal elements did not converge to zero.
NE_ENUM_INT_2
On entry, , and .
Constraint: if or , ;
if , .
NE_INT
On entry, .
Constraint: .
On entry, . Constraint: .
NE_INTERNAL_ERROR
An internal error has occurred in this function. Check the function call and any array sizes. If the call is correct then please contact NAG for assistance.
See Section 7.5 in the Introduction to the NAG Library CL Interface for further information.
NE_NO_LICENCE
Your licence key may have expired or may not have been installed correctly.
See Section 8 in the Introduction to the NAG Library CL Interface for further information.
NE_POS_DEF
The leading minor of order is not positive definite and the Cholesky factorization of could not be completed. Hence itself is not positive definite.
7Accuracy
The eigenvalues and eigenvectors of are computed to high relative accuracy which means that if they vary widely in magnitude, then any small eigenvalues (and corresponding eigenvectors) will be computed more accurately than, for example, with the standard method. However, the reduction to tridiagonal form (prior to calling the function) may exclude the possibility of obtaining high relative accuracy in the small eigenvalues of the original matrix if its eigenvalues vary widely in magnitude.
To be more precise, let be the tridiagonal matrix defined by , where is diagonal with , and for all . If is an exact eigenvalue of and is the corresponding computed value, then
where is a modestly increasing function of , is the machine precision, and is the condition number of with respect to inversion defined by: .
If is the corresponding exact eigenvector of , and is the corresponding computed eigenvector, then the angle between them is bounded as follows:
where is the relative gap between and the other eigenvalues, defined by
8Parallelism and Performance
Background information to multithreading can be found in the Multithreading documentation.
f08jgc is threaded by NAG for parallel execution in multithreaded implementations of the NAG Library.
f08jgc 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 function. Please also consult the Users' Note for your implementation for any additional implementation-specific information.
9Further Comments
The total number of floating-point operations is typically about if and about if or , but depends on how rapidly the algorithm converges. When , the operations are all performed in scalar mode; the additional operations to compute the eigenvectors when or can be vectorized and on some machines may be performed much faster.