Sir, Just recalled I asked this before. So I put what I have tried here.
在 2015年10月1日星期四 UTC-5下午10:58:40，Wolfgang Bangerth写道： > > On 10/01/2015 10:52 PM, Weixiong Zheng wrote: > > I tried to understand what you said and looked at step-9. So what you > suggest is: > > > > std::vector<SmartPointer<PETScWrappers::SparseMatrix> > > > vector_of_pointer_to_sys_matrix, and manipulate each system matrix as > what > > they are usually operated. Is this right? > > Raw pointers work. I used std::vector<PETScWrappers::MPI::SparseMatrix*> and std::vector<PETScWrappers::MPI::Vector*>. As it is for steady state calculations for multiple directions in particle transport, I will use std::vector<std_cxx11::unique_ptr<PETScxxx_preconditioners> > to contain preconditioners for different directions since I have to solve the same equation with different RHS for a lot of times. I was more thinking of > std::vector<std_cxx11::shared_ptr<PETScWrappers::SparseMatrix> > > or > std::vector<std_cxx11::unique_ptr<PETScWrappers::SparseMatrix> > > > > > BTW, is there any possibility that deal.II group implements a serial > version > > of AMG preconditioner so for time-stress people with strange problems > can > > easily use one day in the future? > > I don't see that as likely to be happening. AMG implementations are very > complex, many thousands or tens of thousands of lines of code. (Maybe at > the > lower end of this for sequential implementations.) I also wouldn't know of > anyone within the project having the expertise to do this. > > > Thanks for the discussions, as always. It would of course be nice to have, but ultimately we have to realize that > as > a collective of project contributors, we have a set of skills and > expertise > that covers finite element methods in all of its facets, parallel > computing, > geometric multigrid, software design, and maybe project management. But > none > of us are real experts in linear solvers or algebraic multigrid, to the > best > of my knowledge, and so we need to rely on importing these algorithms from > some of the (excellent) implementations that are available elsewhere. > > That said, there is no doubt that I'd love to be surprised by someone > implementing and contributing a working AMG ;-) > > Best > W. > > Best, Weixiong > -- > ------------------------------------------------------------------------ > Wolfgang Bangerth email: bang...@math.tamu.edu > <javascript:> > www: http://www.math.tamu.edu/~bangerth/ > > -- The deal.II project is located at http://www.dealii.org/ For mailing list/forum options, see https://groups.google.com/d/forum/dealii?hl=en --- You received this message because you are subscribed to the Google Groups "deal.II User Group" group. To unsubscribe from this group and stop receiving emails from it, send an email to dealii+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
