From: Wolfgang Bangerth Date: Sat, 20 Jan 2018 00:59:55 +0000 (-0700) Subject: Address what 'sequential' means in the context of DoFHandlers. X-Git-Tag: v9.0.0-rc1~541^2 X-Git-Url: https://gitweb.dealii.org/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=refs%2Fpull%2F5766%2Fhead;p=dealii.git Address what 'sequential' means in the context of DoFHandlers. --- diff --git a/include/deal.II/dofs/dof_handler.h b/include/deal.II/dofs/dof_handler.h index 41010865bd..f1297fc263 100644 --- a/include/deal.II/dofs/dof_handler.h +++ b/include/deal.II/dofs/dof_handler.h @@ -850,9 +850,15 @@ public: /** * Return the number of degrees of freedom that belong to this process. * - * If this is a sequential job, then the result equals that produced by - * n_dofs(). On the other hand, if we are operating on a - * parallel::distributed::Triangulation, then it includes only the degrees + * If this is a sequential DoFHandler, then the result equals that produced by + * n_dofs(). (Here, "sequential" means that either + * the whole program does not use MPI, or that it uses MPI + * but only uses a single MPI process, or that there are multiple MPI + * processes but the Triangulation on which this DoFHandler builds + * works only on one MPI process.) + * On the other hand, if we are operating on a + * parallel::distributed::Triangulation or parallel::shared::Triangulation, + * then it includes only the degrees * of freedom that the current processor owns. Note that in this case this * does not include all degrees of freedom that have been distributed on the * current processor's image of the mesh: in particular, some of the degrees @@ -880,8 +886,13 @@ public: * you are only interested in the number of elements each processor owns * then n_locally_owned_dofs_per_processor() is a better choice. * - * If this is a sequential job, then the vector has a single element that + * If this is a sequential DoFHandler, then the vector has a single element that * equals the IndexSet representing the entire range [0,n_dofs()]. + * (Here, "sequential" means that either + * the whole program does not use MPI, or that it uses MPI + * but only uses a single MPI process, or that there are multiple MPI + * processes but the Triangulation on which this DoFHandler builds + * works only on one MPI process.) */ const std::vector & locally_owned_dofs_per_processor () const; @@ -895,8 +906,12 @@ public: * Each element of the vector returned by this function equals the number of * elements of the corresponding sets returned by global_dof_indices(). * - * If this is a sequential job, then the vector has a single element equal - * to n_dofs(). + * If this is a sequential DoFHandler, then the vector has a single element equal + * to n_dofs(). (Here, "sequential" means that either + * the whole program does not use MPI, or that it uses MPI + * but only uses a single MPI process, or that there are multiple MPI + * processes but the Triangulation on which this DoFHandler builds + * works only on one MPI process.) */ const std::vector & n_locally_owned_dofs_per_processor () const; @@ -905,8 +920,13 @@ public: * Return a vector that stores the locally owned DoFs of each processor on * the given level @p level. * - * If this is a sequential job, then the vector has a single element that + * If this is a sequential DoFHandler, then the vector has a single element that * equals the IndexSet representing the entire range [0,n_dofs()]. + * (Here, "sequential" means that either + * the whole program does not use MPI, or that it uses MPI + * but only uses a single MPI process, or that there are multiple MPI + * processes but the Triangulation on which this DoFHandler builds + * works only on one MPI process.) */ const std::vector & locally_owned_mg_dofs_per_processor (const unsigned int level) const; diff --git a/include/deal.II/dofs/number_cache.h b/include/deal.II/dofs/number_cache.h index 86de75e4a0..8fd8a6972d 100644 --- a/include/deal.II/dofs/number_cache.h +++ b/include/deal.II/dofs/number_cache.h @@ -54,7 +54,11 @@ namespace internal /** * Create a NumberCache object that corresponds to a sequential * DoFHandler object in which a single processor stores all - * degrees of freedom. + * degrees of freedom. (Here, "sequential" means that either + * the whole program does not use MPI, or that it uses MPI + * but only uses a single MPI process, or that there are multiple MPI + * processes but the Triangulation on which this DoFHandler builds + * works only on one MPI process.) */ NumberCache (const types::global_dof_index n_global_dofs); @@ -106,28 +110,44 @@ namespace internal /** * Number of dofs owned by this MPI process. If this is a sequential - * computation, then this equals n_global_dofs. + * computation, then this equals n_global_dofs. (Here, "sequential" means that either + * the whole program does not use MPI, or that it uses MPI + * but only uses a single MPI process, or that there are multiple MPI + * processes but the Triangulation on which this DoFHandler builds + * works only on one MPI process.) */ types::global_dof_index n_locally_owned_dofs; /** * An index set denoting the set of locally owned dofs. If this is a * sequential computation, then it contains the entire range - * [0,n_global_dofs). + * [0,n_global_dofs). (Here, "sequential" means that either + * the whole program does not use MPI, or that it uses MPI + * but only uses a single MPI process, or that there are multiple MPI + * processes but the Triangulation on which this DoFHandler builds + * works only on one MPI process.) */ IndexSet locally_owned_dofs; /** * The number of dofs owned by each of the various MPI processes. If - * this is a sequential job, then the vector contains a single element - * equal to n_global_dofs. + * this is a sequential computation, then the vector contains a single element + * equal to n_global_dofs. (Here, "sequential" means that either + * the whole program does not use MPI, or that it uses MPI + * but only uses a single MPI process, or that there are multiple MPI + * processes but the Triangulation on which this DoFHandler builds + * works only on one MPI process.) */ std::vector n_locally_owned_dofs_per_processor; /** * The dofs owned by each of the various MPI processes. If this is a - * sequential job, then the vector has a single element equal to - * locally_owned_dofs. + * sequential DoFHandler, then the vector has a single element equal to + * locally_owned_dofs. (Here, "sequential" means that either + * the whole program does not use MPI, or that it uses MPI + * but only uses a single MPI process, or that there are multiple MPI + * processes but the Triangulation on which this DoFHandler builds + * works only on one MPI process.) */ std::vector locally_owned_dofs_per_processor; diff --git a/include/deal.II/hp/dof_handler.h b/include/deal.II/hp/dof_handler.h index a54afc278e..e5ccb23246 100644 --- a/include/deal.II/hp/dof_handler.h +++ b/include/deal.II/hp/dof_handler.h @@ -627,9 +627,14 @@ namespace hp /** * Return the number of degrees of freedom that belong to this process. * - * If this is a sequential job, then the result equals that produced by - * n_dofs(). On the other hand, if we are operating on a - * parallel::distributed::Triangulation, then it includes only the degrees + * If this is a sequential DoFHandler, then the result equals that produced by + * n_dofs(). (Here, "sequential" means that either + * the whole program does not use MPI, or that it uses MPI + * but only uses a single MPI process, or that there are multiple MPI + * processes but the Triangulation on which this DoFHandler builds + * works only on one MPI process.) On the other hand, if we are operating on a + * parallel::distributed::Triangulation or parallel::shared::Triangulation, + * then it includes only the degrees * of freedom that the current processor owns. Note that in this case this * does not include all degrees of freedom that have been distributed on * the current processor's image of the mesh: in particular, some of the @@ -651,8 +656,13 @@ namespace hp * If you are only interested in the number of elements each processor * owns then n_dofs_per_processor() is a better choice. * - * If this is a sequential job, then the vector has a single element that + * If this is a sequential DoFHandler, then the vector has a single element that * equals the IndexSet representing the entire range [0,n_dofs()]. + * (Here, "sequential" means that either + * the whole program does not use MPI, or that it uses MPI + * but only uses a single MPI process, or that there are multiple MPI + * processes but the Triangulation on which this DoFHandler builds + * works only on one MPI process.) */ const std::vector & locally_owned_dofs_per_processor () const; @@ -666,8 +676,12 @@ namespace hp * Each element of the vector returned by this function equals the number * of elements of the corresponding sets returned by global_dof_indices(). * - * If this is a sequential job, then the vector has a single element equal - * to n_dofs(). + * If this is a sequential DoFHandler, then the vector has a single element equal + * to n_dofs(). (Here, "sequential" means that either + * the whole program does not use MPI, or that it uses MPI + * but only uses a single MPI process, or that there are multiple MPI + * processes but the Triangulation on which this DoFHandler builds + * works only on one MPI process.) */ const std::vector & n_locally_owned_dofs_per_processor () const;