From: Wolfgang Bangerth Date: Fri, 15 Jul 2022 02:52:32 +0000 (-0600) Subject: Clarify documentation of p::s::T. X-Git-Tag: v9.5.0-rc1~1089^2 X-Git-Url: https://gitweb.dealii.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=e03dd45c3a3c910fb464b2523e4b621b0ba3f9b8;p=dealii.git Clarify documentation of p::s::T. --- diff --git a/include/deal.II/distributed/shared_tria.h b/include/deal.II/distributed/shared_tria.h index 6512842390..f9d56f0839 100644 --- a/include/deal.II/distributed/shared_tria.h +++ b/include/deal.II/distributed/shared_tria.h @@ -69,9 +69,16 @@ namespace parallel * about cells owned by other processors with the exception of a single * layer of ghost cells around their own part of the domain. * - * Active cells need to be flagged for refinement or coarsening - * locally owned cells. The relevant information is gathered for - * ghost and artificial cells internally via MPI communication. + * Because every MPI process has a complete copy of the entire mesh + * as if it were stored on only this process, it needs to know for + * every active cell whether it is flagged for refinement or coarsening + * when doing mesh refinement via + * Triangulation::execute_coarsening_and_refinement(). In practice, + * each process only needs to set this information for its own + * "locally owned" cells; upon calling + * Triangulation::execute_coarsening_and_refinement(), the + * relevant information is then exchanged between processes + * internally, via MPI communication. * * The class is also useful in cases where compute time and memory * considerations dictate that the program needs to be run in parallel,