[petsc-dev] Schur complement computation support from MKL_PARDISO

[Stefano Zampini]

Is in next for testing. I had to rewrite part of the code for MKL_Pardiso
and Lisandro solved some segfaults.

https://bitbucket.org/petsc/petsc/commits/10b9b08673a199fd5bc1888206ed8f5ac057f924

2015-10-08 20:58 GMT+03:00 Barry Smith <bsmith at mcs.anl.gov>:

>
> > On Oct 8, 2015, at 3:24 AM, Stefano Zampini <stefano.zampini at gmail.com>
> wrote:
> >
> > As I already did for MUMPS, I'm going to interface PETSc with the Schur
> complement feature of MKL.
> >
> > My plans are to add APIs for Schur complement handling to the MatFactor
> class instead of providing specific APIs for MKL and MUMPS; basically, I
> will change the MatMumps namespace to MatFactor one in the Schur complement
> calls; in case the factor does not support Schur complements, it will throw
> an error. You can take a look at src/mat/examples/tests/ex192.c
> >
> > Is it ok for you?
>
>   Sounds reasonable to me.
>
>
> >
> > --
> > Stefano
>
>


-- 
Stefano
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20151014/ccab31b0/attachment.html>