MUMPS solver with Trilinos

[Rahul Gopalan Ramachandran]

Hello everyone,

I am not able to get Trilinos working with MUMPS direct solver, so I am reaching out for some guidance. 

The Os is MacOs (Apple Silicon) and I am able to install mumps solver using Candi. By default candi installs version 5.4.0.  However, Trilinos with Amesos throws error while compiling. I am suspecting a conflict of scalapack to be the issue here. I am currently using MPICH as my MPI compiler and I think I will need to change it to OpenMPI to solve the issue with scalapack. Also interestingly, Trilinos with Amesos2 do not show any issue. So before I change to OpenMPI and reinstall everything, I thought to clarify a couple of  things.

1. Does mumps has to be any specific version to compile with  amesos. Trilinos documentation available on net suggest 4.7.3. However, this is fairly old version.

2. Dealii has no compatibility with Amesos2, right?

Thank you,

Rahul

[Bruno Turcksin]

Rahul,

On Sunday, October 16, 2022 at 11:42:50 AM UTC-4 grrah...@gmail.com wrote:

1. Does mumps has to be any specific version to compile with  amesos. Trilinos documentation available on net suggest 4.7.3. However, this is fairly old version.

I don't know the answer to this question but amesos is in maintenance mode so it's very possible that it only works with old version of mumps

2. Dealii has no compatibility with Amesos2, right?

That's right but you can extract the underlying Trilinos matrices/vectors and use them with Amesos2 yourself. That's what we do in the Trilinos wrappers.

Best,

Bruno

[Rahul Gopalan Ramachandran]

Hello Bruno, 

Thank you for the reply. Your suggestion for Amesos2 might be the best option. 

Thanks again,

Rahul

--
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 a topic in the Google Groups "deal.II User Group" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/dealii/PczQlYM41Mc/unsubscribe.
To unsubscribe from this group and all its topics, send an email to dealii+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/dealii/b185caa7-d2cd-4243-b09f-7a2989560aa1n%40googlegroups.com.

[Praveen C]

I have compiled and used mumps from trilinos, using spack, on M1 machines.

trilinos@13.0.1

mumps@5.4.1

best

praveen

[Rahul Gopalan Ramachandran]

Hello Praveen,

Excellent! Thank you for the input. May I ask if you had success with Open MPI or MPICH?

Regards,

Rahul

-- 
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 a topic in the Google Groups "deal.II User Group" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/dealii/PczQlYM41Mc/unsubscribe.
To unsubscribe from this group and all its topics, send an email to dealii+un...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/dealii/B5C0BD90-5314-4684-A819-0AE5E8A33825%40gmail.com.

[Praveen C]

openmpi is what I have now, I remember having compilation problems with mpich, but I have not tried it recently on M1 machines, so dont know if it works or not. 

Best

praveen

[Rahul Gopalan Ramachandran]

Thank you, most likely that is the issue!

Have a good day,

Rahul

--
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 a topic in the Google Groups "deal.II User Group" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/dealii/PczQlYM41Mc/unsubscribe.
To unsubscribe from this group and all its topics, send an email to dealii+un...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/dealii/131A438D-A6C8-45D1-98C3-B8931667E7D9%40gmail.com.