Amesos: mumps solver gives segfault during SymbolicFactorization()

[swiesheier]

When solving a small linear system (~8000 degrees of freedom), I produce the following stack trace

dmumps_c(&(MDS)); is referred to as "initialization of MUMPS" . That said, I am not sure whether my observation is a bug or not. I can factorize my matrix using Superlu_Dist or any serial package such as UMFPACK. Interestingly, even smaller problems (~1000 degrees of freedom) can be factorized with MUMPS.

Any ideas?

[cgcgcg]

@trilinos/amesos

[cgcgcg]

@swiesheier Could you double check that the matrix is not singular?

EDIT NVM, the issue seems to be in some ordering call using metis.

[swiesheier]

@cgcgcg So likely a bug in metis/gmsh ?

Anything I can do to narrow down the problem? The gmsh library does not provide debug information I think.

Strange is that it works for small problems (~1,000 degrees of freedom).

[cgcgcg]

You could try to reproduce the bug with MUMPS directly to be sure.

[swiesheier]

What do you mean? Making the call dmumps_c(...) directly without using trilinos?

[cgcgcg]

Right. But let's see if someone from the @trilinos/amesos team has more insight.

[cgcgcg]

@ndellingwood @iyamazaki Do you have any idea what could be happening here?

[iyamazaki]

I agree with @cgcgcg. The issue may not be the Amesos MUMPS interface, but may be with your MUMPS to METIS interfacing. It may be good to run MUMPS' example/test with the same matrix.

[swiesheier]

@iyamazaki

MUMPS' example/test

Can you point me to that folder? They have no github repo if I am not mistaken.

[iyamazaki]

Under MUMPS, I see a directory called examples, @swiesheier .

[swiesheier]

I can only access the include directory on my installation. On their website, I also see no source code. Can you give me a link, @iyamazaki

[iyamazaki]

Under my MUMPS directory, I have:

$ ls MUMPS_5.3.3
ChangeLog  CREDITS  doc  examples  include  INSTALL  lib  libseq  LICENSE  Makefile  Makefile.inc  Make.inc  MATLAB  PORD  README  SCILAB  src  VERSION

[ndellingwood]

@ndellingwood @iyamazaki Do you have any idea what could be happening here?

I don't have any direct experience using MUMPS, thanks for the suggestions @iyamazaki !