[petsc-users] Reusing search directions

[Barry Smith]

On Mar 28, 2014, at 11:56 AM, De Groof, Vincent Frans Maria <Vincent.De-Groof at uibk.ac.at> wrote:

> Hi all,
> 
> 
> for the reanalysis of a set of problems (think sensitivity study, reliability, ... ), I'd like to reuse (some) of the search directions of previous problems I solved. The idea is to orthogonolize the new search directions also to a set of user-defined vectors. This idea is not new, and while browsing through the mailing lists, there have been a few discussions on related topics. 
> 
> 1* Can I access/store the old search directions of an iterative solve. Maybe through KSP?

   GMRES keeps its most recent Krylov space basis around so you can access that. See src/ksp/ksp/impls/gmres

> 2* Is there an augmented conjugated gradient implementation available? I've read some discussions on whether or not to implement this since a multigrid approach can reach the same objective. But I don't know if it was implemented or not. 

    No.

   Barry

> 
> 
> 
> thanks,
> Vincent