An external matrix library should be used in fem_1d_poisson_solver.hpp

[EmilyBourne]

An external matrix library should be used in fem_1d_poisson_solver.hpp. Currently we are still using the SLL matrices in:
https://github.com/gyselax/gyselalibxx/tree/main/src/matrix_tools

If this method is ported to GPU then maybe kokkos-kernels can be used, however this may be tricky in the periodic case where the matrix is not banded.

This issue was originally created by @blegouix at https://gitlab.maisondelasimulation.fr/gysela-developpers/gyselalibxx/-/issues/293

[EmilyBourne]

cc: @tpadioleau @gdgirard who were part of this discussion on GitLab. The last comments mentioned a possible Schur's complement in an upstream library. Is this still pertinent?

[gdgirard]

Sorry, I wasn't in the discussion. Maybe Abdelhadi ?

[EmilyBourne]

Sorry, I wasn't in the discussion. Maybe Abdelhadi ?

Abdelhadi does not appear in the discussion on GitLab. I cc-ed in the people who had commented on the issue.

My last question is related to @tpadioleau 's comment here which was addressed to @gdgirard

[tpadioleau]

Haha it was never mentionned in the end 😄