[petsc-users] mumps freezes for bigger problems

[Dominik Szczerba]

> Is one mesh just a refined version of the other, or is the shape different?
> Factorization time should be roughly quadratic in problem size for 3D
> problems.

Yes, that's simply a mesh refinement.

>
> If the increased fill caused the system to run out of physical memory and
> start swapping, then it would not be surprising for it to take days.

It does not swap, it consume memory that I can handle, it just does
nothing, or gets stuck somewhere. This is a toy problem, solved by
PARDISO in no time.

Dominik