![]() |
Reference documentation for deal.II version 8.1.0
|
#include <partitioner.h>
Public Member Functions | |
Partitioner () | |
Partitioner (const unsigned int size) | |
Partitioner (const IndexSet &locally_owned_indices, const IndexSet &ghost_indices_in, const MPI_Comm communicator_in) | |
Partitioner (const IndexSet &locally_owned_indices, const MPI_Comm communicator_in) | |
void | set_owned_indices (const IndexSet &locally_owned_indices) |
void | set_ghost_indices (const IndexSet &ghost_indices) |
types::global_dof_index | size () const |
unsigned int | local_size () const |
const IndexSet & | locally_owned_range () const |
std::pair< types::global_dof_index, types::global_dof_index > | local_range () const |
bool | in_local_range (const types::global_dof_index global_index) const |
unsigned int | global_to_local (const types::global_dof_index global_index) const |
types::global_dof_index | local_to_global (const unsigned int local_index) const |
bool | is_ghost_entry (const types::global_dof_index global_index) const |
const IndexSet & | ghost_indices () const |
unsigned int | n_ghost_indices () const |
const std::vector< std::pair< unsigned int, types::global_dof_index > > & | ghost_targets () const |
const std::vector< std::pair< types::global_dof_index, types::global_dof_index > > & | import_indices () const |
unsigned int | n_import_indices () const |
const std::vector< std::pair< unsigned int, types::global_dof_index > > & | import_targets () const |
bool | is_compatible (const Partitioner &part) const |
unsigned int | this_mpi_process () const |
unsigned int | n_mpi_processes () const |
const MPI_Comm & | get_communicator () const |
std::size_t | memory_consumption () const |
DeclException2 (ExcIndexNotPresent, types::global_dof_index, unsigned int,<< "Global index "<< arg1<< " neither owned nor ghost on proc "<< arg2) | |
Private Attributes | |
const types::global_dof_index | global_size |
IndexSet | locally_owned_range_data |
std::pair< types::global_dof_index, types::global_dof_index > | local_range_data |
IndexSet | ghost_indices_data |
unsigned int | n_ghost_indices_data |
std::vector< std::pair< unsigned int, types::global_dof_index > > | ghost_targets_data |
std::vector< std::pair< types::global_dof_index, types::global_dof_index > > | import_indices_data |
unsigned int | n_import_indices_data |
std::vector< std::pair< unsigned int, types::global_dof_index > > | import_targets_data |
unsigned int | my_pid |
unsigned int | n_procs |
const MPI_Comm | communicator |
This class defines a model for the partitioning of a vector (or, in fact, any linear data structure) among processors using MPI.
The partitioner stores the global vector size and the locally owned range as a half-open interval [lower
, upper
). Furthermore, it includes a structure for the point-to-point communication patterns. It allows the inclusion of ghost indices (i.e. indices that a current processor needs to have access to, but are owned by another process) through an IndexSet. In addition, it also stores the other processors' ghost indices belonging to the current processor, which are the indices where other processors might require information from. In a sense, these import indices form the dual of the ghost indices. This information is gathered once when constructing the partitioner, which obviates subsequent global communication steps when exchanging data.
The partitioner includes a mechanism for converting global to local and local to global indices. Internally, this class stores vector elements using the convention as follows: The local range is associated with local indices [0,local_size
), and ghost indices are stored consecutively in [local_size
, local_size
+ n_ghost_indices
). The ghost indices are sorted according to their global index.
Definition at line 66 of file partitioner.h.
Utilities::MPI::Partitioner::Partitioner | ( | ) |
Empty Constructor.
Utilities::MPI::Partitioner::Partitioner | ( | const unsigned int | size | ) |
Constructor with size argument. Creates an MPI_COMM_SELF structure where there is no real parallel layout.
Utilities::MPI::Partitioner::Partitioner | ( | const IndexSet & | locally_owned_indices, |
const IndexSet & | ghost_indices_in, | ||
const MPI_Comm | communicator_in | ||
) |
Constructor with index set arguments. This constructor creates a distributed layout based on a given communicators, an IndexSet describing the locally owned range and another one for describing ghost indices that are owned by other processors, but we need to have read or write access to.
Utilities::MPI::Partitioner::Partitioner | ( | const IndexSet & | locally_owned_indices, |
const MPI_Comm | communicator_in | ||
) |
Constructor with one index set argument. This constructor creates a distributed layout based on a given communicator, and an IndexSet describing the locally owned range. It allows to set the ghost indices at a later time. Apart from this, it is similar to the other constructor with two index sets.
void Utilities::MPI::Partitioner::set_owned_indices | ( | const IndexSet & | locally_owned_indices | ) |
Sets the locally owned indices. Used in the constructor.
void Utilities::MPI::Partitioner::set_ghost_indices | ( | const IndexSet & | ghost_indices | ) |
Allows to set the ghost indices after the constructor has been called.
types::global_dof_index Utilities::MPI::Partitioner::size | ( | ) | const |
Returns the global size.
unsigned int Utilities::MPI::Partitioner::local_size | ( | ) | const |
Returns the local size, i.e. local_range().second minus local_range().first.
const IndexSet& Utilities::MPI::Partitioner::locally_owned_range | ( | ) | const |
Returns an IndexSet representation of the local range. This class only supports contiguous local ranges, so the IndexSet actually only consists of one single range of data, and is equivalent to the result of local_range().
std::pair<types::global_dof_index,types::global_dof_index> Utilities::MPI::Partitioner::local_range | ( | ) | const |
Returns the local range. The returned pair consists of the index of the first element and the index of the element one past the last locally owned one.
bool Utilities::MPI::Partitioner::in_local_range | ( | const types::global_dof_index | global_index | ) | const |
Returns true if the given global index is in the local range of this processor.
unsigned int Utilities::MPI::Partitioner::global_to_local | ( | const types::global_dof_index | global_index | ) | const |
Returns the local index corresponding to the given global index. If the given global index is neither locally owned nor a ghost, an exception is thrown.
Note that the local index for locally owned indices is between 0 and local_size()-1, and the local index for ghosts is between local_size() and local_size()+n_ghost_indices()-1.
types::global_dof_index Utilities::MPI::Partitioner::local_to_global | ( | const unsigned int | local_index | ) | const |
Returns the global index corresponding to the given local index.
Note that the local index for locally owned indices is between 0 and local_size()-1, and the local index for ghosts is between local_size() and local_size()+n_ghost_indices()-1.
bool Utilities::MPI::Partitioner::is_ghost_entry | ( | const types::global_dof_index | global_index | ) | const |
Returns whether the given global index is a ghost index on the present processor. Returns false for indices that are owned locally and for indices not present at all.
const IndexSet& Utilities::MPI::Partitioner::ghost_indices | ( | ) | const |
Returns an IndexSet representation of all ghost indices.
unsigned int Utilities::MPI::Partitioner::n_ghost_indices | ( | ) | const |
Returns the number of ghost indices. Same as ghost_indices().n_elements(), but cached for simpler access.
const std::vector<std::pair<unsigned int, types::global_dof_index> >& Utilities::MPI::Partitioner::ghost_targets | ( | ) | const |
Returns a list of processors (first entry) and the number of degrees of freedom for the individual processor on the ghost elements present (second entry).
const std::vector<std::pair<types::global_dof_index, types::global_dof_index> >& Utilities::MPI::Partitioner::import_indices | ( | ) | const |
The set of (local) indices that we are importing during compress(), i.e., others' ghosts that belong to the local range. Similar structure as in an IndexSet, but tailored to be iterated over, and some indices may be duplicates.
unsigned int Utilities::MPI::Partitioner::n_import_indices | ( | ) | const |
Number of import indices, i.e., indices that are ghosts on other processors and we will receive data from.
const std::vector<std::pair<unsigned int, types::global_dof_index> >& Utilities::MPI::Partitioner::import_targets | ( | ) | const |
Returns a list of processors (first entry) and the number of degrees of freedom for all the processors that data is obtained from (second entry), i.e., locally owned indices that are ghosts on other processors.
bool Utilities::MPI::Partitioner::is_compatible | ( | const Partitioner & | part | ) | const |
Checks whether the given partitioner is compatible with the partitioner used for this vector. Two partitioners are compatible if the have the same local size and the same ghost indices. They do not necessarily need to be the same data field. This is a local operation only, i.e., if only some processors decide that the partitioning is not compatible, only these processors will return false
, whereas the other processors will return true
.
unsigned int Utilities::MPI::Partitioner::this_mpi_process | ( | ) | const |
Returns the MPI ID of the calling processor. Cached to have simple access.
unsigned int Utilities::MPI::Partitioner::n_mpi_processes | ( | ) | const |
Returns the total number of MPI processor participating in the given partitioner. Cached to have simple access.
const MPI_Comm& Utilities::MPI::Partitioner::get_communicator | ( | ) | const |
Returns the MPI communicator underlying the partitioner object.
std::size_t Utilities::MPI::Partitioner::memory_consumption | ( | ) | const |
Computes the memory consumption of this structure.
Utilities::MPI::Partitioner::DeclException2 | ( | ExcIndexNotPresent | , |
types::global_dof_index | , | ||
unsigned | int, | ||
<< "Global index "<< arg1<< " neither owned nor ghost on proc "<< | arg2 | ||
) |
Exception
|
private |
The global size of the vector over all processors
Definition at line 301 of file partitioner.h.
|
private |
The range of the vector that is stored locally.
Definition at line 307 of file partitioner.h.
|
private |
The range of the vector that is stored locally. Extracted from locally_owned_range for performance reasons.
Definition at line 314 of file partitioner.h.
|
private |
The set of indices to which we need to have read access but that are not locally owned
Definition at line 320 of file partitioner.h.
|
private |
Caches the number of ghost indices. It would be expensive to use ghost_indices.n_elements()
to compute this.
Definition at line 327 of file partitioner.h.
|
private |
Contains information which processors my ghost indices belong to and how many those indices are
Definition at line 334 of file partitioner.h.
|
private |
The set of (local) indices that we are importing during compress(), i.e., others' ghosts that belong to the local range. Similar structure as in an IndexSet, but tailored to be iterated over, and some indices may be duplicates.
Definition at line 344 of file partitioner.h.
|
private |
Caches the number of ghost indices. It would be expensive to compute it by iterating over the import indices and accumulate them.
Definition at line 352 of file partitioner.h.
|
private |
The set of processors and length of data field which send us their ghost data
Definition at line 358 of file partitioner.h.
|
private |
The ID of the current processor in the MPI network
Definition at line 364 of file partitioner.h.
|
private |
The total number of processors active in the problem
Definition at line 370 of file partitioner.h.
|
private |
The MPI communicator involved in the problem
Definition at line 376 of file partitioner.h.