[petsc-users] Performance of mumps vs. Intel Pardiso

Hong hzhang at mcs.anl.gov
Wed Jun 29 13:17:30 CDT 2016


Faraz:

> I reviewed the code for how the matrix is being solved in Intel Paridso.
> It seems the matrix is getting reordered and sorted before calling Pardiso.
> I am not sure the reason for this. I know I need to ask the original
> developer of the application, but they are no longer responding.
>
> Is there something I am misunderstanding about how matrix's need to be
> re-ordered? I am assuming the metis or scotch algorithm does the reordering
> for me when using Mumps? Or am I supposed to reorder/sort it myself before
> calling Mumps?
>

Mumps uses parmetis or scotch for parallel symbolic factorization. For
sequential symbolic factorization, it has several matrix orderings, which
you can experiment with the option '-mat_mumps_icntl_7 <>'.
I doubt any of these ordering would match the performance of Pardiso.

Again, how large is your matrix? How do you run Pardiso in parallel? Can
you use Pardiso on large matrices as mumps?

Hong
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20160629/40914f93/attachment.html>


More information about the petsc-users mailing list