[petsc-users] Replace schur solver in fieldsplit by a matrix multiplication

[Jed Brown]

On Fri, Oct 26, 2012 at 2:23 AM, Thomas Witkowski <
thomas.witkowski at tu-dresden.de> wrote:

> Just to add what I have now is to set the KSP type to PREONLY and the
> preconditioner is a PCSHELL which makes this matrix vector multiplication.
> But is there is simpler way?
>

The PCSHELL could be a PCMAT (or PCSPAI), but I think this is the
consistent approach. We do the same PREONLY approach to do an approximate
solve on subdomains of ASM, for example.


>
> Thomas
>
> Am 26.10.2012 11:20, schrieb Thomas Witkowski:
>
>  Is there a simple way to replace the Schur ksp in fieldsplit
>> preconditioner by a matrix multiplication? So, instead of solving with some
>> Schur operator approximation I just want to make a matrix-vector
>> multiplication, where the matrix is a sparse approximation of the inverse
>> of the Schur complement. Thanks for any suggestions!
>>
>> Thomas
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20121026/27330e39/attachment-0001.html>