June 2012 - petsc-dev - mailman3.cels.anl.gov

Re: [petsc-dev] Parallel Gauss - Seidel
by Barry Smith 11 Jun '12

11 Jun '12

Currently we have MatGetRowIJ((Mat mat,PetscInt shift,PetscBool symmetric,PetscBool inodecompressed,PetscInt *n,PetscInt *ia[],PetscInt* ja[],PetscBool *done) which does the compression based on inode information. We could add another flag that indicated we wanted compression based on the block size for AIJ matrices. Barry On Jun 11, 2012, at 11:01 AM, Mark F. Adams wrote: > As far as how to proceed, now that we are insisting that the block size is set early in matrix creation is there a way to get a vertex graph from a blocked matrix? The MIS code has to manually collapse a blocked matrix, which is ugly. Perhaps I could move this code into MATAIJ at least. > > There is not hurry to do this so I'd like to look at doing this better than MIS. We can talk about this when I'm in town Wed and Thurs (I will also be coming in July). Maybe we could add a getGraph method to Mat ... > > Should I work in SOR? I could add a struct* to PC_SOR that would hold all my junk and a setup method would have to be called. > > Mark

1 0

Re: [petsc-dev] still not communicating to google well
by Jed Brown 11 Jun '12

11 Jun '12

On Mon, Jun 11, 2012 at 11:00 AM, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > Can we somehow imbed referral information into what we distribute so > that when someone dumps them on a website the referral information gives us > credit? That would require them to crawl out of 2006 and upgrade from petsc-2.3.2p7. http://w3.pppl.gov/m3d/petsc-dev/include/petscversion.h.html

1 0

Re: [petsc-dev] Parallel Gauss - Seidel
by Mark F. Adams 11 Jun '12

11 Jun '12

As far as how to proceed, now that we are insisting that the block size is set early in matrix creation is there a way to get a vertex graph from a blocked matrix? The MIS code has to manually collapse a blocked matrix, which is ugly. Perhaps I could move this code into MATAIJ at least. There is not hurry to do this so I'd like to look at doing this better than MIS. We can talk about this when I'm in town Wed and Thurs (I will also be coming in July). Maybe we could add a getGraph method to Mat ... Should I work in SOR? I could add a struct* to PC_SOR that would hold all my junk and a setup method would have to be called. Mark

1 0

Re: [petsc-dev] still not communicating to google well
by Barry Smith 11 Jun '12

11 Jun '12

On Jun 11, 2012, at 10:59 AM, Jed Brown wrote: > Perhaps PPPL has some high-traffic pages linking to their version, and we have very few linking to our version. I don't think there is anything we can do about this short of having them take down their page or insert a referral to our copy. Can we somehow imbed referral information into what we distribute so that when someone dumps them on a website the referral information gives us credit? Barry > > On Mon, Jun 11, 2012 at 8:47 AM, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > > How come petscviewerfilesetname to google goes to: > > • PetscViewerFileSetName > w3.pppl.gov/m3d/petsc-dev/docs/.../PetscViewerFileSetName.html > > Synopsis. #include "petsc.h" PetscErrorCode PETSC_DLLEXPORTPetscViewerFileSetName(PetscViewer viewer,const char name[]) Collective on PetscViewer ... > • #include "src/dm/mesh/meshimpl.h" /*I "petscmesh.h" I*/ #include ... > w3.pppl.gov/m3d/petsc-dev/src/dm/mesh/mesh.c > ... ierr = PetscExceptionTry1(PetscViewerFileSetName(viewer, filename), PETSC_ERR_FILE_OPEN); if (PetscExceptionValue(ierr)) { /* this means that a caller... > > > • PetscViewerFileSetName > www.mcs.anl.gov/petsc/petsc.../docs/.../PetscViewerFileSetName.html > Synopsis. #include "petscsys.h" PetscErrorCode PetscViewerFileSetName( PetscViewer viewer,const char name[]) Collective on PetscViewer ... > > and MCS is not the first listed? > > Barry > >

1 0

Re: [petsc-dev] still not communicating to google well
by Jed Brown 11 Jun '12

11 Jun '12

Perhaps PPPL has some high-traffic pages linking to their version, and we have very few linking to our version. I don't think there is anything we can do about this short of having them take down their page or insert a referral to our copy. On Mon, Jun 11, 2012 at 8:47 AM, Barry Smith <bsmith(a)mcs.anl.gov> wrote: > > How come petscviewerfilesetname to google goes to: > > • PetscViewerFileSetName > w3.pppl.gov/m3d/petsc-dev/docs/.../PetscViewerFileSetName.html > > Synopsis. #include "petsc.h" PetscErrorCode > PETSC_DLLEXPORTPetscViewerFileSetName(PetscViewer viewer,const char name[]) > Collective on PetscViewer ... > • #include "src/dm/mesh/meshimpl.h" /*I "petscmesh.h" I*/ #include > ... > w3.pppl.gov/m3d/petsc-dev/src/dm/mesh/mesh.c > ... ierr = PetscExceptionTry1(PetscViewerFileSetName(viewer, filename), > PETSC_ERR_FILE_OPEN); if (PetscExceptionValue(ierr)) { /* this means that a > caller... > > > • PetscViewerFileSetName > www.mcs.anl.gov/petsc/petsc.../docs/.../PetscViewerFileSetName.html > Synopsis. #include "petscsys.h" PetscErrorCode PetscViewerFileSetName( > PetscViewer viewer,const char name[]) Collective on PetscViewer ... > > and MCS is not the first listed? > > Barry > >

1 0

Re: [petsc-dev] API changes in MatIS
by Jed Brown 11 Jun '12

11 Jun '12

On Mon, Jun 11, 2012 at 9:11 AM, Stefano Zampini <stefano.zampini(a)gmail.com>wrote: > thus the Mat should have a public flag indicating Assembly or not at > preselected vertices during MatAssembly? Part of the Mat configuration would be to select vertices for assembly. If you don't select any, then none would be assembled.

1 0

Re: [petsc-dev] still not communicating to google well
by Mark F. Adams 11 Jun '12

11 Jun '12

On Jun 11, 2012, at 9:47 AM, Barry Smith wrote: > > How come petscviewerfilesetname to google goes to: > > • PetscViewerFileSetName > w3.pppl.gov/m3d/petsc-dev/docs/.../PetscViewerFileSetName.html > > Synopsis. #include "petsc.h" PetscErrorCode PETSC_DLLEXPORTPetscViewerFileSetName(PetscViewer viewer,const char name[]) Collective on PetscViewer ... > • #include "src/dm/mesh/meshimpl.h" /*I "petscmesh.h" I*/ #include ... > w3.pppl.gov/m3d/petsc-dev/src/dm/mesh/mesh.c > ... ierr = PetscExceptionTry1(PetscViewerFileSetName(viewer, filename), PETSC_ERR_FILE_OPEN); if (PetscExceptionValue(ierr)) { /* this means that a caller... > > > • PetscViewerFileSetName > www.mcs.anl.gov/petsc/petsc.../docs/.../PetscViewerFileSetName.html > Synopsis. #include "petscsys.h" PetscErrorCode PetscViewerFileSetName( PetscViewer viewer,const char name[]) Collective on PetscViewer ... > > and MCS is not the first listed? When you figure that out tell me and you can use my Gulfstream whenever you like :) > > Barry > >

1 0

Re: [petsc-dev] API changes in MatIS
by Stefano Zampini 11 Jun '12

11 Jun '12

2012/6/11 Jed Brown <jedbrown(a)mcs.anl.gov> > The problem is that you can't factor the subdomain problem directly > because it is singular. So you either copy it into a bigger matrix (with > Lagrange multipliers) or extract a submatrix (by dropping vertices). > In my experience, a singular subdomain matrix does not constitute a real problem if you use 0 as tolerance for zero pivots to avoid PETSc errors (zero pivot detection) at runtime ( I always use UMFPACK or SuperLU solvers; even the PETSc solver do the right job). For FETI and NN, if you design a coarse space in the right way, you can always solve for the local problems (at least for the problems I faced). The solution will have a nonzero component on the kernel of the local matrix, but this will be uneffective for the convergence of the global problem. If I assemble a matrix with selected vertices, I can factor it and solve > the constrained problems without needing to refactor. > > The case with no selected vertices could be a special case, and worked > with exactly like MatIS is now. > thus the Mat should have a public flag indicating Assembly or not at preselected vertices during MatAssembly? > On Jun 11, 2012 8:35 AM, "Stefano Zampini" <stefano.zampini(a)gmail.com> > wrote: > >> I missed your answer...my e-mail client hid your last email. >> >> >>> >>>> Do you mean assembling the matrix on preselected vertices during >>>> MatAssemblyBegin/End? Note that this will imply that standard >>>> Neumann-Neumann methods will not work (they need the unassembled matrix to >>>> solve for the local Schur complements). >>> >>> >>> I'm not too concerned about that since I consider the classic N-N and >>> original FETI methods to be rather special-purpose compared to the newer >>> generation. I would like to limit the number of copies of a matrix to >>> control peak memory usage. >>> >> >> Duplicate vertices don't concur so much to memory usage as faces and edge >> nodes do, so why limit your matrix class? In principle the new class should >> handle _every_ nonoverlapping method, either belonging to the old or the >> new generation. >> >> -- >> Stefano >> > -- Stefano

1 0

still not communicating to google well
by Barry Smith 11 Jun '12

11 Jun '12

How come petscviewerfilesetname to google goes to: • PetscViewerFileSetName w3.pppl.gov/m3d/petsc-dev/docs/.../PetscViewerFileSetName.html Synopsis. #include "petsc.h" PetscErrorCode PETSC_DLLEXPORTPetscViewerFileSetName(PetscViewer viewer,const char name[]) Collective on PetscViewer ... • #include "src/dm/mesh/meshimpl.h" /*I "petscmesh.h" I*/ #include ... w3.pppl.gov/m3d/petsc-dev/src/dm/mesh/mesh.c ... ierr = PetscExceptionTry1(PetscViewerFileSetName(viewer, filename), PETSC_ERR_FILE_OPEN); if (PetscExceptionValue(ierr)) { /* this means that a caller... • PetscViewerFileSetName www.mcs.anl.gov/petsc/petsc.../docs/.../PetscViewerFileSetName.html Synopsis. #include "petscsys.h" PetscErrorCode PetscViewerFileSetName( PetscViewer viewer,const char name[]) Collective on PetscViewer ... and MCS is not the first listed? Barry

1 0

Re: [petsc-dev] API changes in MatIS
by Jed Brown 11 Jun '12

11 Jun '12

The problem is that you can't factor the subdomain problem directly because it is singular. So you either copy it into a bigger matrix (with Lagrange multipliers) or extract a submatrix (by dropping vertices). If I assemble a matrix with selected vertices, I can factor it and solve the constrained problems without needing to refactor. The case with no selected vertices could be a special case, and worked with exactly like MatIS is now. On Jun 11, 2012 8:35 AM, "Stefano Zampini" <stefano.zampini(a)gmail.com> wrote: > I missed your answer...my e-mail client hid your last email. > > >> >>> Do you mean assembling the matrix on preselected vertices during >>> MatAssemblyBegin/End? Note that this will imply that standard >>> Neumann-Neumann methods will not work (they need the unassembled matrix to >>> solve for the local Schur complements). >> >> >> I'm not too concerned about that since I consider the classic N-N and >> original FETI methods to be rather special-purpose compared to the newer >> generation. I would like to limit the number of copies of a matrix to >> control peak memory usage. >> > > Duplicate vertices don't concur so much to memory usage as faces and edge > nodes do, so why limit your matrix class? In principle the new class should > handle _every_ nonoverlapping method, either belonging to the old or the > new generation. > > -- > Stefano >

1 0