[petsc-dev] Lagging the numerical factorization for one of the fields of fieldsplit preconditioner

[Barry Smith]

On Mar 15, 2011, at 6:25 PM, Jed Brown wrote:

> On Wed, Mar 16, 2011 at 00:13, Barry Smith <bsmith at mcs.anl.gov> wrote:
> You cannot just check the row and column separately. You need to check them as a pair.  If row < m and column < n mark A as changed.
> 
> But you don't know what "A" is yet so all you can mark is some particular pair (row,col) as changed.

  Of course I know what A is. Because I am only doing this tracking after a MatGetSubmatrix() has been called which provide the information about what A is.

   Barry

> And you can't store cheaply in the form (row_range, col_range) because you don't know when to extend an existing range or when to create a new range. Maybe there is something clever that works in a lot of common cases, but I don't think there is a cheap automatic way to do this. I think you would have to declare index sets for submatrices up-front.