From: Manav B. <manav@u.washington.edu> - 2005-07-22 20:27:35
|
Thanks for your response Ben, It is for a serial case. I am using MPI, but with a single processor. Yes, I have checked the assembled, and stored matrices, and both are identical. This happens only for a multiple dof case. Thanks Manav On Jul 22, 2005, at 12:52 PM, KIRK, BENJAMIN (JSC-EG) (NASA) wrote: > Is this problem for a serial or parallel case? > > The DOF numbering will be different in parallel when running on M > processors > that when running on N processors. > > Can you assemble the matrix *and* read an old one from disk and > make sure > the difference is 0? > > -Ben > > -----Original Message----- > From: lib...@li... > [mailto:lib...@li...] On Behalf Of > Manav Bhatia > Sent: Friday, July 22, 2005 12:56 PM > To: lib...@li... > Subject: [Libmesh-users] petsc issues > > > Hi > I am using libmesh with Petsc, and have implemented a feature > where you can write the system matrix to disk via PetscBinaryViewer > and then read it in later and then solve. > > I have tested this with a heat transfer problem that has one dof > per node, and it works fine. > > However, when I do this for a problem with multiple dofs, like a > structural problem, it is giving me all sorts of wierd results. This > is wierd since just before the solve, I print the matrix to check if > it was read alright, and there does not seem to be any problem with > that. > > The procedure I follow is > -- to write > -- create petsc binary viewer > -- use ma view to write to the viewer > -- destroy petsc binary viewer > > -- to read > -- create petsc binary viewer > -- destroy the existing matrix with matrix destroy > -- read in the matrix with mat load > -- destroy petsc binary viewer > > After that, I use the solve function in the petsclinearsolver class. > > Any hints would be appreciated. > > Thanks > Manav > > > > ------------------------------------------------------- > SF.Net email is sponsored by: Discover Easy Linux Migration > Strategies from > IBM. Find simple to follow Roadmaps, straightforward articles, > informative > Webcasts and more! Get everything you need to get up to speed, fast. > http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click > _______________________________________________ > Libmesh-users mailing list > Lib...@li... > https://lists.sourceforge.net/lists/listinfo/libmesh-users > |