Hi Victor,

if you decide to use a direct solver and you go into the direction Prof.
Banghert hinted, i.e. you decide to use PARDISO: You can access it
indirectly to through the PETSC interface of deal.ii. You have to set
your -ksp_type to none and your -pc_type to lu  and then you can choose
Pardiso with  -pc_factor_mat_solver_type mkl_pardiso

See https://petsc.org/main/manualpages/Mat/MATSOLVERMKL_PARDISO/ for details.

Keep in mind that you need to instruct PARDISO how many cores to use. On
many clusters you can do this with exporting the variable OMP_NUM_THREADS

If you do not like to use a shared memory PARDISO, you can also use a
distributed memory variant called cluster PARDISO
https://petsc.org/release/manualpages/Mat/MATSOLVERMKL_CPARDISO/

A sidenote: If you like to use these you to need to compile PETSC such
that the solvers are known to it.

Best,
Nils


Victor Eijkhout <[email protected]> writes:

> 1. I thought multifrontal solvers are designed for parallelism. But anyway, 
> since this is external to Deal, it wouldn't use your TaskFlow
> threads.
> 2. So BlockJacobi doesn't use parallel tasks either? I thought I was seeing 
> thread activity in my run.
>
> Ok, so how do I get a parallel preconditioner of higher quality than Jacobi?
> Or since this is a time-stepping method, any parallel solver.
>
> Should I use the Trilinos or PETSc wrappers and implicitly use MPI, even 
> though I'm on a single node?

-- 
Nils Schween

Phone: +49 6221 516 557
Mail: [email protected]
PGP-Key: 4DD3DCC0532EE96DB0C1F8B5368DBFA14CB81849

Max Planck Institute for Nuclear Physics
Astrophysical Plasma Theory (APT)
Saupfercheckweg 1, D-69117 Heidelberg
https://www.mpi-hd.mpg.de/mpi/en/research/scientific-divisions-and-groups/independent-research-groups/apt

-- 
The deal.II project is located at http://www.dealii.org/
For mailing list/forum options, see 
https://groups.google.com/d/forum/dealii?hl=en
--- 
You received this message because you are subscribed to the Google Groups 
"deal.II User Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion visit 
https://groups.google.com/d/msgid/dealii/uo28q9hw3pi.fsf%40mpi-hd.mpg.de.

Reply via email to