Yes, this requires mucking with the internals of the data structure. Not terribly hard but requires some thought (especially in parallel).
Barry I wish I had time to do this, but we don't unless more PETSc developers appear. On Fri, 29 Jun 2007, Toby D. Young wrote: > > > > > Ok, I understand. They want to zero out all the rows AND COLUMNS > > that associated with the zero boundary conditioned points. Since > > PETSc sparse matrices are stored by row it is difficult to zero > > certain columns. > > Yes, we are trying to zero out rows & columns. I think I understand the > source of the problem myself a lot better too now; thanks for that > comment Barry. > > > Perhaps we could augment the CSR storage format by having an > > additional list of locations for each column? This could be generated > > at MatAssembly time. > > This can not be implemented by a user can it, in fact this requires some > modification of the internals of PETSc matrices, am I right? > > Sadly, I am only a beginner in the PETSc world and therefore not "PETSc > competent" enough to try it. :-( > > Cheers, > Toby. > > >
