Hello: Sorry for the longish email that follows, I tried to keep it as succinct as possible.
tl;dr Is there any way to alter EquationSystems::reinit() to not call refine/coarsen_elements()? These calls prevent me from properly reinitializing my new QoI class. Or, is there another way to use the element refinement flags to identify elements that were just refined/coarsened? @pbauman pointed me to a GitHub PR <https://github.com/libMesh/libmesh/pull/1080> discussion where it seems that these refine/coarsen calls in ES::reinit() are not really necessary anymore. ----------------------------- I am currently working on implementing a new QoI in GRINS, but I am having an issue with EquationSystems::reinit(). My new QoI is based on the GRINS::RayfireMesh class, which must be reinitialized after every refine/coarsening. To fully integrate this, I am adding hooks into GRINS::MultiphysicsSystem::reinit() that will call reinit() on my QoI class. I have a test case that follows this flow: assemble QoI, refine certain elements, reinit the EquationSystems (which would also reinit my QoI), assemble QoI again. The reinit() <https://github.com/grinsfem/grins/blob/master/src/qoi/src/rayfire_mesh.C#L160> for RayfireMesh looks for an INACTIVE element with JUST_REFINED children in order to refine its own internal 1D mesh. The problem arises in that MeshRefinement::refine_elements() ends up getting called *twice*, once in my test case and again in ES::reinit() ( https://github.com/libMesh/libmesh/blob/master/src/systems/equation_systems.C#L265 ) When refine_elements() is called the second time, it replaces all JUST_REFINED flags with DO_NOTHING, breaking my RayfireMesh reinit function. If I do not call refine_elements() in my test, and instead just flag the elements and let the ES::reinit() do the refinement, then this <https://github.com/libMesh/libmesh/blob/master/src/systems/equation_systems.C#L270> IF statement is skipped, and prolong_vectors() below it trips an assert. Perhaps I could alter how I am identifying the just refined elements? The RayfireMesh map stores the IDs of main mesh elements along the 1D rayfire. On reinit(), it looks for inactive elements, and then tries to determine if they were refined or coarsened by looking at the refinement flags of its children/parent. Would there be any other situation where an element would go from ACTIVE to INACTIVE with DO_NOTHING children? I'm not sure how robust of a check that would be. I would appreciate any thoughts/insights/suggestions for how to address this. Thank you, -- *~Tim Adowski*
------------------------------------------------------------------------------
_______________________________________________ Libmesh-devel mailing list Libmesh-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libmesh-devel
