Next: Postprocessing and Accuracy Checking
Up: Modify the Problem Dependent
Previous: Modify the Problem Dependent
The following variables are also set in all drivers.
Their usage is described in Chapter 2.
In most cases, they do not need to be changed.<<354>>
- [lworkl] The size of the work array workl used by dsaupd.
Must be set to at least ncv*(ncv+8).
- [tol] The convergence criterion. The default setting
is machine precision. However, the value of tol
should be set to control the desired accuracy.
Typically, the smaller
this value the more work is required to satisfy
the stopping criteria. However, setting this value
too large may cause eigenvalues to be missed when
there are multiple or clustered eigenvalues.
- [ido] The reverse communication flag.
Must be set to 0 before entering dsaupd.
- [bmat] Designates whether a standard (bmat = 'I') or
generalized eigenvalue (bmat = 'G') problem
- [iparam(1)] The shifting strategy used during the implicitly restarted
portion of an IRLM. Unless the user has an expert
understanding of IRLM, an exact shifting strategy
selected by setting iparam(1) = 1 should be used.
- [iparam(3)] Maximum number of IRLM iterations allowed.
- [iparam(7)] Indicates the algorithmic mode used with ARPACK.
- [rvec] Indicates whether eigenvectors are desired.
If the eigenvectors are of to be computed,
then one must set rvec = .true. and otherwise
set it to .false. .
Next: Postprocessing and Accuracy Checking
Up: Modify the Problem Dependent
Previous: Modify the Problem Dependent
Chao Yang
11/7/1997