AssertDimension (partition_blocks[partition], task_info.n_blocks) when calling MatrixFree::reinit() for finest level in GMG

[Denis Davydov]

Dear all,

I am hitting  an AssertDimension (partition_blocks[partition], task_info.n_blocks); when initializing MatrixFree object for the finest level of the mesh.

This happens at 2nd h-adaptive step with 3 a-priori global refinements. 

The part of the code during setup of each level is rather standard GMG with Laplace operator:

        typename MatrixFree<dim,LevelNumberType>::AdditionalData mg_additional_data;

        mg_additional_data.tasks_parallel_scheme = MatrixFree<dim,LevelNumberType>::AdditionalData::partition_color;

        mg_additional_data.level_mg_handler = level;

        mg_additional_data.mapping_update_flags = update_gradients | update_JxW_values;

        ConstraintMatrix level_constraints;

        IndexSet relevant_dofs;

        DoFTools::extract_locally_relevant_level_dofs(dof_handler, level,

                                                      relevant_dofs);

        level_constraints.reinit(relevant_dofs);

        level_constraints.add_lines(mg_constrained_dofs.get_boundary_indices(level));

        level_constraints.close();

        std::shared_ptr<MatrixFree<dim,LevelNumberType>> mg_level_data(new MatrixFree<dim,LevelNumberType>());

        mg_level_data->reinit (mapping, dof_handler, level_constraints, quadrature_formula,

                               mg_additional_data);

backtrace from this reinit is: 

 matrix_free.h:1668

   1665     quad_hp.emplace_back (quad[q]);

   1666   internal_reinit(StaticMappingQ1<dim>::mapping, dof_handler,constraint,

   1667                   locally_owned_set, quad_hp, additional_data);

-> 1668 }

 matrix_free.templates.h:166

   163        // constraint_pool_data. It also reorders the way cells are gone through

   164        // (to separate cells with overlap to other processors from others

   165        // without).

-> 166        initialize_indices (constraint, locally_owned_set);

 matrix_free.templates.h:689

   686          (size_info, task_info, renumbering, irregular_cells,

   687           dof_handlers.active_dof_handler == DoFHandlers::hp);

   688        else

-> 689          dof_info[0].make_thread_graph_partition_color

   690          (size_info, task_info, renumbering, irregular_cells,

   691           dof_handlers.active_dof_handler == DoFHandlers::hp);

dof_info.templates.h:1011

   1008                 break;

   1009               }

   1010         }

-> 1011       AssertDimension (partition_blocks[partition], task_info.n_blocks);

Any ideas?

Regards,

Denis.

[Martin Kronbichler]

Hi Denis,

This is a bug somewhere in the initialization of the blocks. Could you
please send me the code to reproduce this behavior?

Best,
Martin

[Denis Davydov]

Just to confirm my sanity, I can reproduce the error by loading in the p::d::Tria and setting up GMG objects.

Of course, that does not really help anyone to try this as the refinement history in p::d::Tria::save() is written as a binary file... :-(

[Denis Davydov]

p.s. missed attachments in the last post.

[Daniel Arndt]

Denis,

this is solved by https://github.com/dealii/dealii/pull/4376, isn't it?

[Denis Davydov]

Correct!

On 14 May 2017, at 10:50, Daniel Arndt <daniel...@iwr.uni-heidelberg.de> wrote:

Denis,

this is solved by https://github.com/dealii/dealii/pull/4376, isn't it?

--
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/zSC33HFN9AA/unsubscribe.
To unsubscribe from this group and all its topics, send an email to dealii+un...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.