From: John P. <jwp...@gm...> - 2009-06-15 16:07:49
|
On Mon, Jun 15, 2009 at 11:00 AM, Brent Kraczek<bkr...@ic...> wrote: > Roy, > > I am working toward having a patch to do this. > > But there is something which I don't understand here. It appears to me > that SparseMatrix::vector_mult doesn't call the appropriate PETSc > routine. It's not clear to me why SparseMatrix::vector_mult or at least > SparseMatrix::vector_mult_add hasn't been implemented as a pure-virtual > member function of SparseMatrix like virtually every other member > function of this class has been. Is this a mistake, or was it something > intended to be implemented later? Or am I just mis-reading the code? > (note: I am working from svn revision 3397) Looks like it calls vector_mult_add() which calls add_vector() from the NumericVector class. This has a specialized PETSc implementation in petsc_vector.C -- John |