Hello everyone, I am currently solving a stabilized form of the Navier-Stokes equations. Due to the stabilization terms, I can easily lump everything in a single matrix. Everything is done via the Trillinos Wrappers. I get relatively nice scaling up to a certain number of cells (say 1e6 or so), but then the resolution time greatly increases, especially in parallel. My feeling is that this is due to the decreased efficiency of the ILU preconditioner that I use. I would like to test the use of the AMG preconditioner of Trillinos. I have managed to set-up everything accordling, but I have an issue with the constant mode in the additional data.
https://www.dealii.org/8.5.0/doxygen/deal.II/structTrilinosWrappers_1_1PreconditionAMG_1_1AdditionalData.html#af4c9b8fcda773646bafc45a09b9c800f Which reads: *Specifies the constant modes (near null space) of the matrix. This parameter tells AMG whether we work on a scalar equation (where the near null space only consists of ones, and default value is OK) or on a vector-valued equation. For vector-valued equation problem with n_component, the provided constant_modes should fulfill the following requirements: * - * n_component.size() == n_component* - * n_component[*].size() == n_dof_local or n_component[*].size() == n_dof_global * - * n_component[ic][id] == "idth DoF is corresponding to component ic* My understanding is that since my problem has 4 DOF per node (if I am P1), I need to use specify the correct constant_modes. However, I have not really been able to figure out how. I know this may sound like a very simple question, but could someone guide me in how I should specify by constant modes if I have a nDim+1 component problem? I don't think I really understand this part of the documentation. Thank you for your understanding and help. -- 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]. For more options, visit https://groups.google.com/d/optout.
