[petsc-users] question about LU decomposition of a conjugated matrix

[Michael Povolotskyi]

On 11/22/2013 06:36 PM, Jed Brown wrote:
> Michael, your time zone is mistakenly set to UTC, though the time is
> actually set to CST, so your mails appear out of order.
>
> Michael Povolotskyi <mpovolot at purdue.edu> writes:
>
>> I'm actually interested in implementing it.
>> Could you, please, tell what has to be done in MatMatSolveTranspose()?
>>
>> By transpose I mean transpose and complex conjugation.
> Some time ago, Barry proposed adding a function like
> MatSetTransposeType(A,MAT_TRANSPOSE_HERMITIAN) that would cause
> MatMultTranspose(), MatSolveTranspose(), etc., to use Hermitian
> transpose.  The thought was that this was ultimately simpler to work
> since the problem formulation usually dictates whether a complex matrix
> uses a Hermitian transpose or not.  Does that sound reasonable to you?
>
>
> I can see two implementation strategies.  One is to go into
> MatLUFactor_SeqAIJ and set A->ops->transpose to a new function that
> explicitly transposes the factors.  The other is to copy MatMatSolve and
> modify to be MatMatSolveTranspose, along with an implementation
> MatMatSolveTranspose_SeqAIJ (and external packages if relevant).
Thank you, Jed,
I fixed the time zone, sorry for the inconvenience.

This sounds very reasonable for me and my colleagues to have 
MatSetTransposeType(A,MAT_TRANSPOSE_HERMITIAN).
Our problems always require Hermitian transpose but I assume there are 
problems where it is not the case.

Actually my goal is to solve the transposed problem with PARDISO 
package. It looks like it is the fastest option for the problems we 
usually solve.

By the way, it there a chance to have interface to Pardiso in the next 
PETSc release? My colleague Jose David Bermeo sent the modules for this 
some time ago for your approval.
Michael.