Thanks! They indeed are for the overlapping division. I just wanted to confirm it to debug my code, which I now think is giving me the right size of the subdomain matrices.
On Thu, Jun 4, 2015 at 2:11 PM, Matthew Knepley <[email protected]> wrote: > On Thu, Jun 4, 2015 at 3:47 PM, Amneet Bhalla <[email protected]> > wrote: > >> >> Hi Folks, >> >> I have a basic question regarding matrix size as created by subdomains in >> PCASM. >> >> In the PCASMSetLocalSubdomains(PC pc,PetscInt n,IS is[],IS is_local[]) >> routine, I am defining is[] to be the local+overlapping DOFs and is_local >> [] to be the nonoverlapping local DOFs only. Now >> if I grab the subksp's from PCASM and print out the matrices, would the >> matrix size correspond to size of IS in is[] or in is_local[] dofs? Put in >> another words, do the subdomain matrices are defined for just local DOFs or >> with local+overlapping DOFs? >> > > The matrices should definitely be for the overlapping division. If they > are not, something is wrong. If they were > the non-overlapping distribution, we would have block Jacobi. > > Thanks, > > Matt > > >> Thanks, >> -- >> --Amneet >> >> >> >> > > > -- > What most experimenters take for granted before they begin their > experiments is infinitely more interesting than any results to which their > experiments lead. > -- Norbert Wiener > -- --Amneet
