HP MLIB User's Guide Vol. 2 7th Ed.

Chapter 15 Sparse Eigenvalues and Eigenvectors 1007
What you need to know to use these subprograms
This sparse matrix format, known as the column pointer, row index
representation, is illustrated in Figure 15-2.
Figure 15-2 Column Pointer, Row Index Sparse Matrix Representation
There are three ways to communicate the coefficient matrix or matrices to the
package. One is a totally general form, which allows the user to store the
matrices outside the package in whatever form is most convenient. The other
two ways require that the user store the matrices in a form similar to the
internal format or at least with all entries in each column contiguous in
memory. Any of these three can be used. However, the most general form
carries additional overhead in computer time.
Description of sparse eigenvalue problems
Sparsity in the matrix does not guarantee sparsity in the matrix of
eigenvectors. Indeed, it is rare that any eigenvectors of a sparse matrix are
sparse. This implies that computing all the eigenvalues and eigenvectors of a
sparse matrix is an enormous computation requiring storage of a large dense
matrix. The common requirement for sparse eigenanalysis is to compute a
subset, usually small, of the eigenvalues and vectors. This section describes
how to specify the appropriate subsets to the sparse eigenanalysis package.
colstr
=
rowind
=
mxvalu
=
1 4 7 9 11 12 13
34235354556
11 31 41 22 32 52 33 53 44 54 55 66
1
_