[Trilinos-Users] Is it possible to implement blockLanczos through anasazi?
aklinvex at gmail.com
Thu Oct 27 00:17:06 EDT 2016
They are certainly very similar. Here is a paper about BKS:
http://www-users.cs.umn.edu/~saad/PDF/umsi-2004-215.pdf I have also copied
Heidi Thornquist and Rich Lehoucq on this message, both of whom have done a
lot of work with Arnoldi/Lanczos/BKS.
If you absolutely need Lanczos and BKS just won't do, you can certainly
implement Lanczos yourself...but I wouldn't worry about doing so in the
Anasazi interface. You can use the Trilinos datatypes and routines to
implement Lanczos without trying to implement a new Anasazi solver
manager. If you have any specific questions about how to do this, I'd be
happy to answer them.
I don't know which Lanczos library you're trying to evaluate, but I believe
there's a Lanczos implementation in SLEPc if that helps:
On Wed, Oct 26, 2016 at 7:39 PM, SungHwan Choi <sunghwanchoi91 at gmail.com>
> Dear all,
> I want to use block lanczos method in anasazi interface. As far as I
> understood, the blockKrylovSchur method is almost same to the block lanczos
> method. Are they different? If so, what's the different points between
> them? Are there any simple ways to implement BlockLanczos method within
> anasazi interface? Since my final goal is to evaluate the performance of
> another Lanczos Library. Of coure, I can compare other methods (e.g.
> LOBPCG, BlockKrylovSchur), however, if comparison result for the same
> method exist, it could be much easier to discuss the results.
> Sunghwan Choi
> Trilinos-Users mailing list
> Trilinos-Users at trilinos.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Trilinos-Users