[petsc-dev] Question about PETSc and pardiso

[Cea, Stephen M]

Hi Matt,

Thanks!  I am not sure it is really needed,  not sure if you have seen anything to indicates a big improvement for the different reordering inside PETSc.   I just wanted to make sure I was understanding the results I was seeing.

Sorry if I submitted to the wrong email.

Thanks
steve

From: Matthew Knepley <knepley at gmail.com>
Sent: Thursday, July 13, 2023 10:20 AM
To: Cea, Stephen M <stephen.m.cea at intel.com>
Cc: petsc-dev at mcs.anl.gov
Subject: Re: [petsc-dev] Question about PETSc and pardiso

On Thu, Jul 13, 2023 at 12:43 PM Cea, Stephen M <stephen.m.cea at intel.com<mailto:stephen.m.cea at intel.com>> wrote:
Hi,

            I had a questions about PETSc and pardiso integration.    Can you use any of the built into PETSc matrix reordering routines with pardiso or just the built in options in pardiso?

Currently, you can only use the built-in reorderings.

However, it would be simple to extend. We already support using a user-supplied permutation

  -mat_mkl_pardiso_5

but there is no interface for setting the "perm" array in the Pardiso struct. We could add an interface
to take the IS from MatGetOrdering() and stick it into this array. Note that Pardiso only supports symmetric
reorderings.

  Thanks,

     Matt

Thanks!
steve


--
What most experimenters take for granted before they begin their experiments is infinitely more interesting than any results to which their experiments lead.
-- Norbert Wiener

https://www.cse.buffalo.edu/~knepley/<http://www.cse.buffalo.edu/~knepley/>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20230713/e9affee3/attachment-0001.html>