[petsc-dev] preallocation

Barry Smith bsmith at mcs.anl.gov
Tue Sep 20 12:04:39 CDT 2011


On Sep 20, 2011, at 11:14 AM, Jed Brown wrote:

> On Tue, Sep 20, 2011 at 18:06, Matthew Knepley <knepley at gmail.com> wrote:
> I disagree with this. I think the correct reduction in complexity is for the DM to call all relevant versions.
> 
> Many people do not use the DM. I'm working with an unstructured FV code right now that has never heard of DM. It's just a matter of keeping the purely algebraic as simple as possible.

  Why not write a wrapper preallocator that then calls all the known ones, for example

    MatPreallocateJed(.....)
    {
        ierr = MatPreallocateSeqAIJ(....)
        ierr = MatPreallocate(MPIAIJ(....)

    }

    I think that is a much cleaner, easier thing to maintain then trying to get MatPreallocateMPIBAIJ() to deal with all the other ones too.

   Barry





More information about the petsc-dev mailing list