Tpetra parallel linear algebra
Version of the Day
|
Namespace Tpetra contains the class and methods constituting the Tpetra library. More...
Namespaces | |
Details | |
Namespace for Tpetra implementation details. | |
Experimental | |
Namespace for new Tpetra features that are not ready for public release, but are ready for evaluation by friendly expert users. | |
Ext | |
Namespace for external Tpetra functionality. | |
MatrixMatrix | |
Distributed sparse matrix-matrix multiply and add. | |
RTI | |
Namespace for Tpetra Reduction/Tranformation Interface. | |
Classes | |
class | CrsGraph |
A distributed graph accessed by rows (adjacency lists) and stored sparsely. More... | |
class | CrsMatrix |
Sparse matrix that presents a row-oriented interface that lets users read or modify entries. More... | |
class | CrsMatrixMultiplyOp |
A class for wrapping a CrsMatrix multiply in a Operator. More... | |
class | CrsMatrixSolveOp |
Wrap a CrsMatrix instance's triangular solve in an Operator. More... | |
class | CrsMatrixStruct |
Struct that holds views of the contents of a CrsMatrix. More... | |
class | DefaultPlatform |
Returns a default platform appropriate for the enviroment. More... | |
class | Directory |
Implement mapping from global ID to process ID and local ID. More... | |
class | DistObject |
Base class for distributed Tpetra objects that support data redistribution. More... | |
class | Distributor |
Sets up and executes a communication plan for a Tpetra DistObject. More... | |
class | Export |
Communication plan for data redistribution from a (possibly) multiply-owned to a uniquely-owned distribution. More... | |
class | Import |
Communication plan for data redistribution from a uniquely-owned to a (possibly) multiply-owned distribution. More... | |
class | ImportExportData |
Implementation detail of Import and Export. More... | |
class | Map |
Describes a parallel distribution of objects over processes. More... | |
class | MpiPlatform |
Implementation of the Platform concept for MPI-based platforms. More... | |
class | MpiPlatform< Tpetra::Details::DefaultTypes::node_type > |
MpiPlatform specialization for the default Node type. More... | |
class | MultiVector |
One or more distributed dense vectors. More... | |
class | MultiVectorFiller |
Adds nonlocal sum-into functionality to Tpetra::MultiVector. More... | |
class | Operator |
Abstract interface for operators (e.g., matrices and preconditioners). More... | |
class | Packable |
Abstract base class for objects that can be the source of an Import or Export operation, and that also know how to pack their data to send to the target object. More... | |
class | project1st |
Binary function that returns its first argument. More... | |
class | project2nd |
Binary function that returns its second argument. More... | |
class | RowGraph |
An abstract interface for graphs accessed by rows. More... | |
struct | RowInfo |
Allocation information for a locally owned row in a CrsGraph or CrsMatrix. More... | |
class | RowMatrix |
A read-only, row-oriented interface to a sparse matrix. More... | |
class | RowMatrixTransposer |
Construct and (optionally) redistribute the explicitly stored transpose of a CrsMatrix. More... | |
class | SerialPlatform |
Implementation of the Platform concept for non-MPI platforms. More... | |
class | SerialPlatform< Tpetra::Details::DefaultTypes::node_type > |
SerialPlatform specialization for the default Node type. More... | |
class | SrcDistObject |
Abstract base class for objects that can be the source of an Import or Export operation. More... | |
class | Vector |
A distributed dense vector. More... | |
Typedefs | |
typedef Teuchos_Ordinal | Array_size_type |
Size type for Teuchos Array objects. More... | |
typedef size_t | global_size_t |
Global size_t object. More... | |
Enumerations |
Functions | |
void | setCombineModeParameter (Teuchos::ParameterList &plist, const std::string ¶mName) |
Set CombineMode parameter in a Teuchos::ParameterList. More... | |
bool | isInitialized () |
Whether Tpetra is in an initialized state. More... | |
Teuchos::RCP< const Teuchos::Comm< int > > | getDefaultComm () |
Get Tpetra's default communicator. More... | |
void | initialize (int *argc, char ***argv) |
Initialize Tpetra. More... | |
void | initialize (int *argc, char ***argv, const Teuchos::RCP< const Teuchos::Comm< int > > &comm) |
Initialize Tpetra. More... | |
template<class LocalOrdinal , class GlobalOrdinal , class Node , const bool classic = Node::classic> | |
Teuchos::RCP< CrsGraph< LocalOrdinal, GlobalOrdinal, Node, classic > > | createCrsGraph (const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > &map, size_t maxNumEntriesPerRow=0, const Teuchos::RCP< Teuchos::ParameterList > ¶ms=Teuchos::null) |
Nonmember function to create an empty CrsGraph given a row Map and the max number of entries allowed locally per row. More... | |
template<class Scalar , class LocalOrdinal , class GlobalOrdinal , class Node , const bool classic = Node::classic> | |
Teuchos::RCP< CrsMatrix< Scalar, LocalOrdinal, GlobalOrdinal, Node, classic > > | createCrsMatrix (const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > &map, size_t maxNumEntriesPerRow=0, const Teuchos::RCP< Teuchos::ParameterList > ¶ms=Teuchos::null) |
Non-member function to create an empty CrsMatrix given a row map and a non-zero profile. More... | |
template<class CrsMatrixType > | |
Teuchos::RCP< CrsMatrixType > | importAndFillCompleteCrsMatrix (const Teuchos::RCP< const CrsMatrixType > &sourceMatrix, const Import< typename CrsMatrixType::local_ordinal_type, typename CrsMatrixType::global_ordinal_type, typename CrsMatrixType::node_type > &importer, const Teuchos::RCP< const Map< typename CrsMatrixType::local_ordinal_type, typename CrsMatrixType::global_ordinal_type, typename CrsMatrixType::node_type > > &domainMap=Teuchos::null, const Teuchos::RCP< const Map< typename CrsMatrixType::local_ordinal_type, typename CrsMatrixType::global_ordinal_type, typename CrsMatrixType::node_type > > &rangeMap=Teuchos::null, const Teuchos::RCP< Teuchos::ParameterList > ¶ms=Teuchos::null) |
Nonmember CrsMatrix constructor that fuses Import and fillComplete(). More... | |
template<class CrsMatrixType > | |
Teuchos::RCP< CrsMatrixType > | exportAndFillCompleteCrsMatrix (const Teuchos::RCP< const CrsMatrixType > &sourceMatrix, const Export< typename CrsMatrixType::local_ordinal_type, typename CrsMatrixType::global_ordinal_type, typename CrsMatrixType::node_type > &exporter, const Teuchos::RCP< const Map< typename CrsMatrixType::local_ordinal_type, typename CrsMatrixType::global_ordinal_type, typename CrsMatrixType::node_type > > &domainMap=Teuchos::null, const Teuchos::RCP< const Map< typename CrsMatrixType::local_ordinal_type, typename CrsMatrixType::global_ordinal_type, typename CrsMatrixType::node_type > > &rangeMap=Teuchos::null, const Teuchos::RCP< Teuchos::ParameterList > ¶ms=Teuchos::null) |
Nonmember CrsMatrix constructor that fuses Export and fillComplete(). More... | |
template<class OpScalar , class MatScalar , class LocalOrdinal , class GlobalOrdinal , class Node > | |
Teuchos::RCP< CrsMatrixMultiplyOp< OpScalar, MatScalar, LocalOrdinal, GlobalOrdinal, Node > > | createCrsMatrixMultiplyOp (const Teuchos::RCP< const CrsMatrix< MatScalar, LocalOrdinal, GlobalOrdinal, Node > > &A) |
Non-member function to create a CrsMatrixMultiplyOp. More... | |
template<class OpScalar , class MatScalar , class LocalOrdinal , class GlobalOrdinal , class Node > | |
Teuchos::RCP< CrsMatrixSolveOp< OpScalar, MatScalar, LocalOrdinal, GlobalOrdinal, Node > > | createCrsMatrixSolveOp (const Teuchos::RCP< const CrsMatrix< MatScalar, LocalOrdinal, GlobalOrdinal, Node > > &A) |
Nonmember function that wraps a CrsMatrix in a CrsMatrixSolveOp. More... | |
template<class DistObjectType > | |
void | removeEmptyProcessesInPlace (Teuchos::RCP< DistObjectType > &input, const Teuchos::RCP< const Map< typename DistObjectType::local_ordinal_type, typename DistObjectType::global_ordinal_type, typename DistObjectType::node_type > > &newMap) |
Remove processes which contain no elements in this object's Map. More... | |
template<class DistObjectType > | |
void | removeEmptyProcessesInPlace (Teuchos::RCP< DistObjectType > &input) |
Remove processes which contain no elements in this object's Map. More... | |
Array< std::string > | distributorSendTypes () |
Valid values for Distributor's "Send type" parameter. More... | |
template<class LocalOrdinal , class GlobalOrdinal , class Node > | |
Teuchos::RCP< const Export< LocalOrdinal, GlobalOrdinal, Node > > | createExport (const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > &src, const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > &tgt) |
Non-member constructor for Export objects. More... | |
template<class LocalOrdinal , class GlobalOrdinal , class Node > | |
Teuchos::RCP< const Import< LocalOrdinal, GlobalOrdinal, Node > > | createImport (const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > &src, const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > &tgt) |
Nonmember constructor for Import. More... | |
template<class LocalOrdinal , class GlobalOrdinal , class Node > | |
Teuchos::RCP< const Import< LocalOrdinal, GlobalOrdinal, Node > > | createImport (const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > &src, const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > &tgt, const Teuchos::RCP< Teuchos::ParameterList > &plist) |
Nonmember constructor for Import that takes a ParameterList. More... | |
template<class LocalOrdinal , class GlobalOrdinal > | |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal > > | createLocalMap (size_t numElements, const Teuchos::RCP< const Teuchos::Comm< int > > &comm) |
Nonmember constructor for a locally replicated Map with the default Kokkos Node. More... | |
template<class LocalOrdinal , class GlobalOrdinal , class Node > | |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > | createLocalMapWithNode (size_t numElements, const Teuchos::RCP< const Teuchos::Comm< int > > &comm, const Teuchos::RCP< Node > &node=defaultArgNode< Node >()) |
Nonmember constructor for a locally replicated Map with a specified Kokkos Node. More... | |
template<class LocalOrdinal , class GlobalOrdinal > | |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal > > | createUniformContigMap (global_size_t numElements, const Teuchos::RCP< const Teuchos::Comm< int > > &comm) |
Non-member constructor for a uniformly distributed, contiguous Map with the default Kokkos Node. More... | |
template<class LocalOrdinal , class GlobalOrdinal , class Node > | |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > | createUniformContigMapWithNode (global_size_t numElements, const Teuchos::RCP< const Teuchos::Comm< int > > &comm, const Teuchos::RCP< Node > &node=defaultArgNode< Node >()) |
Non-member constructor for a uniformly distributed, contiguous Map with a user-specified Kokkos Node. More... | |
template<class LocalOrdinal , class GlobalOrdinal > | |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal > > | createContigMap (global_size_t numElements, size_t localNumElements, const Teuchos::RCP< const Teuchos::Comm< int > > &comm) |
Non-member constructor for a (potentially) non-uniformly distributed, contiguous Map with the default Kokkos Node. More... | |
template<class LocalOrdinal , class GlobalOrdinal , class Node > | |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > | createContigMapWithNode (global_size_t numElements, size_t localNumElements, const Teuchos::RCP< const Teuchos::Comm< int > > &comm, const Teuchos::RCP< Node > &node=defaultArgNode< Node >()) |
Non-member constructor for a (potentially) non-uniformly distributed, contiguous Map with a user-specified Kokkos Node. More... | |
template<class LocalOrdinal , class GlobalOrdinal > | |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal > > | createNonContigMap (const ArrayView< const GlobalOrdinal > &elementList, const RCP< const Teuchos::Comm< int > > &comm) |
Non-member constructor for a non-contiguous Map with the default Kokkos Node. More... | |
template<class LocalOrdinal , class GlobalOrdinal , class Node > | |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > | createNonContigMapWithNode (const ArrayView< const GlobalOrdinal > &elementList, const RCP< const Teuchos::Comm< int > > &comm, const RCP< Node > &node) |
Non-member constructor for a non-contiguous Map with a user-specified Kokkos Node. More... | |
template<class LocalOrdinal , class GlobalOrdinal , class Node > | |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > | createWeightedContigMapWithNode (int thisNodeWeight, global_size_t numElements, const Teuchos::RCP< const Teuchos::Comm< int > > &comm, const Teuchos::RCP< Node > &node) |
Non-member constructor for a contiguous Map with user-defined weights and a user-specified Kokkos Node. More... | |
template<class LocalOrdinal , class GlobalOrdinal , class Node > | |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > | createOneToOne (const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > &M) |
Creates a one-to-one version of the given Map where each GID is owned by only one process. More... | |
template<class LocalOrdinal , class GlobalOrdinal , class Node > | |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > | createOneToOne (const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > &M, const Tpetra::Details::TieBreak< LocalOrdinal, GlobalOrdinal > &tie_break) |
Creates a one-to-one version of the given Map where each GID is owned by only one process. The rule to break ties is specifed by the tie break object. More... | |
template<class DS , class DL , class DG , class DN , const bool dstClassic, class SS , class SL , class SG , class SN , const bool srcClassic> | |
void | deep_copy (MultiVector< DS, DL, DG, DN, dstClassic > &dst, const MultiVector< SS, SL, SG, SN, srcClassic > &src) |
Copy the contents of the MultiVector src into dst . More... | |
template<class ST , class LO , class GO , class NT , const bool classic = NT::classic> | |
MultiVector< ST, LO, GO, NT, classic > | createCopy (const MultiVector< ST, LO, GO, NT, classic > &src) |
Return a deep copy of the given MultiVector. More... | |
template<class Scalar , class LocalOrdinal , class GlobalOrdinal , class Node , const bool classic = Node::classic> | |
Teuchos::RCP< MultiVector< Scalar, LocalOrdinal, GlobalOrdinal, Node, classic > > | createMultiVector (const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > &map, const size_t numVectors) |
Nonmember MultiVector "constructor": Create a MultiVector from a given Map. More... | |
template<class ST , class LO , class GO , class NT , const bool classic> | |
void | deep_copy (MultiVector< ST, LO, GO, NT, classic > &dst, const MultiVector< ST, LO, GO, NT, classic > &src) |
Specialization of deep_copy for MultiVector objects with the same template parameters. More... | |
template<typename MapType , typename KeyArgType , typename ValueArgType > | |
MapType::iterator | efficientAddOrUpdate (MapType &m, const KeyArgType &k, const ValueArgType &v) |
Efficiently insert or replace an entry in an std::map. More... | |
template<class IT1 , class IT2 > | |
void | sort2 (const IT1 &first1, const IT1 &last1, const IT2 &first2) |
Sort the first array, and apply the resulting permutation to the second array. More... | |
template<class IT1 , class IT2 , class IT3 > | |
void | sort3 (const IT1 &first1, const IT1 &last1, const IT2 &first2, const IT3 &first3) |
Sort the first array, and apply the same permutation to the second and third arrays. More... | |
template<class IT1 , class IT2 > | |
void | merge2 (IT1 &indResultOut, IT2 &valResultOut, IT1 indBeg, IT1 indEnd, IT2 valBeg, IT2 valEnd) |
Merge values in place, additively, with the same index. More... | |
template<class IT1 , class IT2 , class BinaryFunction > | |
void | merge2 (IT1 &indResultOut, IT2 &valResultOut, IT1 indBeg, IT1 indEnd, IT2 valBeg, IT2 valEnd, BinaryFunction f) |
Merge values in place with the same index, using any associative binary function. More... | |
template<class KeyInputIterType , class ValueInputIterType , class KeyOutputIterType , class ValueOutputIterType , class BinaryFunction > | |
void | keyValueMerge (KeyInputIterType keyBeg1, KeyInputIterType keyEnd1, ValueInputIterType valBeg1, ValueInputIterType valEnd1, KeyInputIterType keyBeg2, KeyInputIterType keyEnd2, ValueInputIterType valBeg2, ValueInputIterType valEnd2, KeyOutputIterType keyOut, ValueOutputIterType valOut, BinaryFunction f) |
Merge two sorted (by keys) sequences of unique (key,value) pairs by combining pairs with equal keys. More... | |
template<class Scalar , class LocalOrdinal , class GlobalOrdinal , class Node , const bool classic = Node::classic> | |
Vector< Scalar, LocalOrdinal, GlobalOrdinal, Node, classic > | createCopy (const Vector< Scalar, LocalOrdinal, GlobalOrdinal, Node, classic > &src) |
Return a deep copy of the given Vector. More... | |
template<class Scalar , class LocalOrdinal , class GlobalOrdinal , class Node , const bool classic = Node::classic> | |
Teuchos::RCP< Vector< Scalar, LocalOrdinal, GlobalOrdinal, Node, classic > > | createVector (const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > &map) |
Nonmember Vector "constructor": Create a Vector from a given Map. More... | |
Namespace Tpetra contains the class and methods constituting the Tpetra library.
Namespace for Tpetra classes and methods.
Tpetra namespace
typedef Teuchos_Ordinal Tpetra::Array_size_type |
Size type for Teuchos Array objects.
Definition at line 54 of file Tpetra_ConfigDefs.hpp.
typedef size_t Tpetra::global_size_t |
Global size_t object.
This type is intended to support scenarios where the global memory allocation is larger than that of a single node.
Currently, it is typedefed to size_t
.
Definition at line 111 of file Tpetra_ConfigDefs.hpp.
enum Tpetra::CombineMode |
Rule for combining data in an Import or Export.
Import or Export (data redistribution) operations might need to combine data received from other processes with existing data on the calling process. This enum tells Tpetra how to do that for a specific Import or Export operation. Each Tpetra object may interpret the CombineMode in a different way, so you should check the Tpetra object's documentation for details.
Here is the list of supported combine modes:
ADD and REPLACE are intended for modifying values that already exist. Tpetra objects will generally work correctly if those values don't already exist. (For example, ADD will behave like INSERT if the entry does not yet exist on the calling process.) However, performance may suffer.
The ZERO combine mode is a special case that bypasses communication. It may seem odd to include a "combine mode" that doesn't actually combine. However, this is useful for computations like domain decomposition with overlap. A ZERO combine mode with overlap is different than an ADD combine mode without overlap. (See Ifpack2::AdditiveSchwarz, which inspired inclusion of this combine mode.) Furthermore, Import and Export also encapsulate a local permutation; if you want only to execute the local permutation without communication, you may use the ZERO combine mode.
Definition at line 94 of file Tpetra_CombineMode.hpp.
enum Tpetra::LocalGlobal |
Enum for local versus global allocation of Map entries.
LocallyReplicated
means that the Map's entries are locally replicated across all processes.
GloballyDistributed
means that the Map's entries are globally distributed across all processes.
Definition at line 120 of file Tpetra_ConfigDefs.hpp.
enum Tpetra::LookupStatus |
Return status of Map remote index lookup (getRemoteIndexList()).
Enumerator | |
---|---|
AllIDsPresent |
All queried indices were present in the Map |
IDNotPresent |
At least one of the specified indices was not present in the Map |
Definition at line 126 of file Tpetra_ConfigDefs.hpp.
enum Tpetra::ProfileType |
Allocation profile for matrix/graph entries
Enumerator | |
---|---|
StaticProfile |
Single, static allocation (strict and more efficient) |
DynamicProfile |
Multiple, dynamic allocations (flexibile, but less efficient) |
Definition at line 132 of file Tpetra_ConfigDefs.hpp.
Optimize storage option
Enumerator | |
---|---|
DoOptimizeStorage |
Indicates that storage should be optimized |
DoNotOptimizeStorage |
Indicates that storage should not be optimized |
Definition at line 138 of file Tpetra_ConfigDefs.hpp.
Sweep direction for Gauss-Seidel or Successive Over-Relaxation (SOR).
Definition at line 300 of file Tpetra_ConfigDefs.hpp.
void Tpetra::setCombineModeParameter | ( | Teuchos::ParameterList & | plist, |
const std::string & | paramName | ||
) |
Set CombineMode parameter in a Teuchos::ParameterList.
If you are constructing a Teuchos::ParameterList with a CombineMode parameter, set the parameter by using this function. This will use a special feature of Teuchos – custom parameter list validation – so that users can specify CombineMode values by string, rather than enum value. The strings are the same as the enum names: "ADD", "INSERT", "REPLACE", "ABSMAX", and "ZERO". They are not case sensitive.
Using this function to set a CombineMode parameter will ensure that the XML serialization of the resulting Teuchos::ParameterList will refer to the CombineMode enum values using human-readable string names, rather than raw integers.
plist | [out] Teuchos::ParameterList to which you want to add the Tpetra::CombineMode parameter. |
paramName | [in] String name to use for the parameter. For example, you might wish to call the parameter "Combine Mode", "Tpetra::CombineMode", or "combine mode". The parameter's name is case sensitive, even though the string values are not. |
Definition at line 48 of file Tpetra_CombineMode.cpp.
bool Tpetra::isInitialized | ( | ) |
Whether Tpetra is in an initialized state.
Initialize Tpetra by calling one of the versions of initialize(). After initialize() returns, Tpetra is initialized. Once finalize() returns, Tpetra is no longer initialized.
Definition at line 93 of file Tpetra_Core.cpp.
Teuchos::RCP< const Teuchos::Comm< int > > Tpetra::getDefaultComm | ( | ) |
Get Tpetra's default communicator.
Definition at line 97 of file Tpetra_Core.cpp.
void Tpetra::initialize | ( | int * | argc, |
char *** | argv | ||
) |
Initialize Tpetra.
This initializes the following if they have not already been initialized:
If Trilinos was built with MPI enabled, this function sets the default communicator to MPI_COMM_WORLD (wrapped in a Teuchos wrapper). Otherwise, it sets the default communicator to a Teuchos::SerialComm instance.
argc | [in/out] Same as first argument of MPI_Init() |
argv | [in/out] Same as second argument of MPI_Init() |
The argc
and argv
arguments are both passed by pointer, in order to match MPI_Init's interface. MPI_Init() reserves the right to modify command-line arguments, e.g., by reading and removing those that pertain to MPI. Thus, in main(), one would write
Definition at line 112 of file Tpetra_Core.cpp.
void Tpetra::initialize | ( | int * | argc, |
char *** | argv, | ||
const Teuchos::RCP< const Teuchos::Comm< int > > & | comm | ||
) |
Initialize Tpetra.
This initializes the following if they have not already been initialized:
argc | [in/out] Same as first argument of MPI_Init() |
argv | [in/out] Same as second argument of MPI_Init() |
comm | [in] Tpetra's default communicator, wrapped in a Teuchos wrapper. This may be either a Teuchos::MpiComm or a Teuchos::SerialComm instance. |
The argc
and argv
arguments are both passed by pointer, in order to match MPI_Init's interface. MPI_Init() reserves the right to modify command-line arguments, e.g., by reading and removing those that pertain to MPI. Thus, in main(), one would write
Definition at line 169 of file Tpetra_Core.cpp.
Teuchos::RCP< CrsGraph< LocalOrdinal, GlobalOrdinal, Node, classic > > createCrsGraph | ( | const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > & | map, |
size_t | maxNumEntriesPerRow = 0 , |
||
const Teuchos::RCP< Teuchos::ParameterList > & | params = Teuchos::null |
||
) |
Nonmember function to create an empty CrsGraph given a row Map and the max number of entries allowed locally per row.
Definition at line 2610 of file Tpetra_CrsGraph_decl.hpp.
Teuchos::RCP< CrsMatrix< Scalar, LocalOrdinal, GlobalOrdinal, Node, classic > > createCrsMatrix | ( | const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > & | map, |
size_t | maxNumEntriesPerRow = 0 , |
||
const Teuchos::RCP< Teuchos::ParameterList > & | params = Teuchos::null |
||
) |
Non-member function to create an empty CrsMatrix given a row map and a non-zero profile.
Definition at line 3559 of file Tpetra_CrsMatrix_decl.hpp.
Teuchos::RCP< CrsMatrixType > importAndFillCompleteCrsMatrix | ( | const Teuchos::RCP< const CrsMatrixType > & | sourceMatrix, |
const Import< typename CrsMatrixType::local_ordinal_type, typename CrsMatrixType::global_ordinal_type, typename CrsMatrixType::node_type > & | importer, | ||
const Teuchos::RCP< const Map< typename CrsMatrixType::local_ordinal_type, typename CrsMatrixType::global_ordinal_type, typename CrsMatrixType::node_type > > & | domainMap = Teuchos::null , |
||
const Teuchos::RCP< const Map< typename CrsMatrixType::local_ordinal_type, typename CrsMatrixType::global_ordinal_type, typename CrsMatrixType::node_type > > & | rangeMap = Teuchos::null , |
||
const Teuchos::RCP< Teuchos::ParameterList > & | params = Teuchos::null |
||
) |
Nonmember CrsMatrix constructor that fuses Import and fillComplete().
CrsMatrixType | A specialization of CrsMatrix. |
A common use case is to create an empty destination CrsMatrix, redistribute from a source CrsMatrix (by an Import or Export operation), then call fillComplete() on the destination CrsMatrix. This constructor fuses these three cases, for an Import redistribution.
Fusing redistribution and fillComplete() exposes potential optimizations. For example, it may make constructing the column Map faster, and it may avoid intermediate unoptimized storage in the destination CrsMatrix. These optimizations may improve performance for specialized kernels like sparse matrix-matrix multiply, as well as for redistributing data after doing load balancing.
The resulting matrix is fill complete (in the sense of isFillComplete()) and has optimized storage (in the sense of isStorageOptimized()). By default, its domain Map is the domain Map of the source matrix, and its range Map is the range Map of the source matrix.
sourceMatrix | [in] The source matrix from which to import. The source of an Import must have a nonoverlapping distribution. |
importer | [in] The Import instance containing a precomputed redistribution plan. The source Map of the Import must be the same as the rowMap of sourceMatrix unless the "Reverse Mode" option on the params list, in which case the targetMap of Import must match the rowMap of the sourceMatrix |
domainMap | [in] Domain Map of the returned matrix. If null, we use the default, which is the domain Map of the source matrix. |
rangeMap | [in] Range Map of the returned matrix. If null, we use the default, which is the range Map of the source matrix. |
params | [in/out] Optional list of parameters. If not null, any missing parameters will be filled in with their default values. |
Definition at line 3621 of file Tpetra_CrsMatrix_decl.hpp.
Teuchos::RCP< CrsMatrixType > exportAndFillCompleteCrsMatrix | ( | const Teuchos::RCP< const CrsMatrixType > & | sourceMatrix, |
const Export< typename CrsMatrixType::local_ordinal_type, typename CrsMatrixType::global_ordinal_type, typename CrsMatrixType::node_type > & | exporter, | ||
const Teuchos::RCP< const Map< typename CrsMatrixType::local_ordinal_type, typename CrsMatrixType::global_ordinal_type, typename CrsMatrixType::node_type > > & | domainMap = Teuchos::null , |
||
const Teuchos::RCP< const Map< typename CrsMatrixType::local_ordinal_type, typename CrsMatrixType::global_ordinal_type, typename CrsMatrixType::node_type > > & | rangeMap = Teuchos::null , |
||
const Teuchos::RCP< Teuchos::ParameterList > & | params = Teuchos::null |
||
) |
Nonmember CrsMatrix constructor that fuses Export and fillComplete().
CrsMatrixType | A specialization of CrsMatrix. |
For justification, see the documentation of importAndFillCompleteCrsMatrix() (which is the Import analog of this function).
The resulting matrix is fill complete (in the sense of isFillComplete()) and has optimized storage (in the sense of isStorageOptimized()). By default, its domain Map is the domain Map of the source matrix, and its range Map is the range Map of the source matrix.
sourceMatrix | [in] The source matrix from which to export. Its row Map may be overlapping, since the source of an Export may be overlapping. |
exporter | [in] The Export instance containing a precomputed redistribution plan. The source Map of the Export must be the same as the row Map of sourceMatrix. |
domainMap | [in] Domain Map of the returned matrix. If null, we use the default, which is the domain Map of the source matrix. |
rangeMap | [in] Range Map of the returned matrix. If null, we use the default, which is the range Map of the source matrix. |
params | [in/out] Optional list of parameters. If not null, any missing parameters will be filled in with their default values. |
Definition at line 3673 of file Tpetra_CrsMatrix_decl.hpp.
Teuchos::RCP< CrsMatrixMultiplyOp< OpScalar, MatScalar, LocalOrdinal, GlobalOrdinal, Node > > createCrsMatrixMultiplyOp | ( | const Teuchos::RCP< const CrsMatrix< MatScalar, LocalOrdinal, GlobalOrdinal, Node > > & | A | ) |
Non-member function to create a CrsMatrixMultiplyOp.
The function has the same template parameters of CrsMatrixMultiplyOp.
A | [in] The CrsMatrix instance to wrap in an CrsMatrixMultiplyOp. |
Definition at line 1187 of file Tpetra_CrsMatrixMultiplyOp.hpp.
Teuchos::RCP< CrsMatrixSolveOp< OpScalar, MatScalar, LocalOrdinal, GlobalOrdinal, Node > > createCrsMatrixSolveOp | ( | const Teuchos::RCP< const CrsMatrix< MatScalar, LocalOrdinal, GlobalOrdinal, Node > > & | A | ) |
Nonmember function that wraps a CrsMatrix in a CrsMatrixSolveOp.
The function has the same template parameters of CrsMatrixSolveOp.
A | [in] The CrsMatrix instance to wrap in an CrsMatrixSolveOp. |
Definition at line 379 of file Tpetra_CrsMatrixSolveOp.hpp.
void Tpetra::removeEmptyProcessesInPlace | ( | Teuchos::RCP< DistObjectType > & | input, |
const Teuchos::RCP< const Map< typename DistObjectType::local_ordinal_type, typename DistObjectType::global_ordinal_type, typename DistObjectType::node_type > > & | newMap | ||
) |
Remove processes which contain no elements in this object's Map.
DistObjectType | A specialization of DistObject. |
Vocabulary:
input->getMap() on input to this method is the "original Map."
The communicator returned by input->getComm() on input to this method is the "original communicator."
All processes in the original communicator which contain zero elements in the original Map are "excluded processes."
All other processes in the original communicator are "included
processes."
Preconditions:
input
is distributed over the original Map.newMap
must be the same as the result of calling removeEmptyProcesses() on the original Map.newMap
must be Teuchos::null
. (This is what getMap()->removeEmptyProcesses()
returns anyway on excluded processes.)This method has collective semantics over the original communicator. On included processes, reassign this object's Map (that would be returned by getMap()) to the input
newMap
, and do any work that needs to be done to restore correct semantics. The input DistObject input
will be nonnull on return. On excluded processes, free any data in input
that need freeing, do any other work that needs to be done to restore correct semantics, and set input
to null before returning.
The two-argument version of this function is useful if you have already precomputed the new Map that excludes processes with zero elements. For example, you might want to apply this Map to several different MultiVector instances. The one-argument version of this function is useful if you want the DistObject to compute the new Map itself, because you only plan to use it for that one DistObject instance.
Here is a sample use case. Suppose that
input
is some subclass of DistObject, like MultiVector, CrsGraph, or CrsMatrix. Suppose also that map_type
is the corresponding specialization of Map.
input
. Calling any methods (other than the destructor) on the input on excluded processes has undefined behavior in that case, and may result in deadlock.Definition at line 950 of file Tpetra_DistObject_def.hpp.
void Tpetra::removeEmptyProcessesInPlace | ( | Teuchos::RCP< DistObjectType > & | input | ) |
Remove processes which contain no elements in this object's Map.
DistObjectType | A specialization of DistObject. |
This method behaves just like the two-argument version of removeEmptyProcessesInPlace(), except that it first calls removeEmptyProcesses() on the input DistObject's Map to compute the new Map.
The two-argument version of this function is useful if you have already precomputed the new Map that excludes processes with zero elements. For example, you might want to apply this Map to several different MultiVector instances. The one-argument version of this function is useful if you want the DistObject to compute the new Map itself, because you only plan to use it for that one DistObject instance.
Here is a sample use case. Suppose that input
is some subclass of DistObject, like MultiVector, CrsGraph, or CrsMatrix. Suppose also that map_type
is the corresponding specialization of Map.
Definition at line 963 of file Tpetra_DistObject_def.hpp.
Array< std::string > Tpetra::distributorSendTypes | ( | ) |
Valid values for Distributor's "Send type" parameter.
This is mainly useful as an implementation detail of Distributor. You may use it if you would like a programmatic way to get all possible values of the "Send type" parameter of Distributor.
Definition at line 90 of file Tpetra_Distributor.cpp.
Teuchos::RCP< const Export< LocalOrdinal, GlobalOrdinal, Node > > createExport | ( | const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > & | src, |
const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > & | tgt | ||
) |
Non-member constructor for Export objects.
Creates a Export object from the given source and target maps.
src != null
tgt != null
src == tgt
, returns null
. (Debug mode: throws std::runtime_error if one of src
or tgt
is null
.) Definition at line 329 of file Tpetra_Export_decl.hpp.
Teuchos::RCP< const Import< LocalOrdinal, GlobalOrdinal, Node > > createImport | ( | const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > & | src, |
const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > & | tgt | ||
) |
Nonmember constructor for Import.
Create a Import object from the given source and target Maps.
src != null
tgt != null
src == tgt
, returns null
. (Debug mode: throws std::runtime_error if one of src
or tgt
is null
.) Definition at line 511 of file Tpetra_Import_decl.hpp.
Teuchos::RCP< const Import< LocalOrdinal, GlobalOrdinal, Node > > createImport | ( | const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > & | src, |
const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > & | tgt, | ||
const Teuchos::RCP< Teuchos::ParameterList > & | plist | ||
) |
Nonmember constructor for Import that takes a ParameterList.
Create a Import object from the given source and target Maps, using the given list of parameters.
src != null
tgt != null
src == tgt
, returns null
. (Debug mode: throws std::runtime_error if one of src
or tgt
is null
.) Definition at line 541 of file Tpetra_Import_decl.hpp.
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal > > createLocalMap | ( | size_t | numElements, |
const Teuchos::RCP< const Teuchos::Comm< int > > & | comm | ||
) |
Nonmember constructor for a locally replicated Map with the default Kokkos Node.
This method returns a Map instantiated on the default Kokkos Node type. The Map is configured to use zero-based indexing.
numElements | [in] Number of elements on each process. Each process gets the same set of elements, namely 0, 1, ..., numElements - 1 . |
comm | [in] The Map's communicator. |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > createLocalMapWithNode | ( | size_t | numElements, |
const Teuchos::RCP< const Teuchos::Comm< int > > & | comm, | ||
const Teuchos::RCP< Node > & | node = defaultArgNode< Node >() |
||
) |
Nonmember constructor for a locally replicated Map with a specified Kokkos Node.
This method returns a Map instantiated on the given Kokkos Node instance. The Map is configured to use zero-based indexing.
numElements | [in] Number of elements on each process. Each process gets the same set of elements, namely 0, 1, ..., numElements - 1 . |
comm | [in] The Map's communicator. |
node | [in] The Kokkos Node instance. If not provided, we will construct an instance of the correct type for you. |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal > > createUniformContigMap | ( | global_size_t | numElements, |
const Teuchos::RCP< const Teuchos::Comm< int > > & | comm | ||
) |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > createUniformContigMapWithNode | ( | global_size_t | numElements, |
const Teuchos::RCP< const Teuchos::Comm< int > > & | comm, | ||
const Teuchos::RCP< Node > & | node = defaultArgNode< Node >() |
||
) |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal > > createContigMap | ( | global_size_t | numElements, |
size_t | localNumElements, | ||
const Teuchos::RCP< const Teuchos::Comm< int > > & | comm | ||
) |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > createContigMapWithNode | ( | global_size_t | numElements, |
size_t | localNumElements, | ||
const Teuchos::RCP< const Teuchos::Comm< int > > & | comm, | ||
const Teuchos::RCP< Node > & | node = defaultArgNode< Node >() |
||
) |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal > > createNonContigMap | ( | const ArrayView< const GlobalOrdinal > & | elementList, |
const RCP< const Teuchos::Comm< int > > & | comm | ||
) |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > createNonContigMapWithNode | ( | const ArrayView< const GlobalOrdinal > & | elementList, |
const RCP< const Teuchos::Comm< int > > & | comm, | ||
const RCP< Node > & | node | ||
) |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > createWeightedContigMapWithNode | ( | int | thisNodeWeight, |
global_size_t | numElements, | ||
const Teuchos::RCP< const Teuchos::Comm< int > > & | comm, | ||
const Teuchos::RCP< Node > & | node | ||
) |
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > createOneToOne | ( | const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > & | M | ) |
Creates a one-to-one version of the given Map where each GID is owned by only one process.
The user must guarantee there are no duplicate GID on the same processor. Unexepected behavior may result.
Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > createOneToOne | ( | const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > & | M, |
const Tpetra::Details::TieBreak< LocalOrdinal, GlobalOrdinal > & | tie_break | ||
) |
Creates a one-to-one version of the given Map where each GID is owned by only one process. The rule to break ties is specifed by the tie break object.
The user must guarantee there are no duplicate GID on the same processor. Unexepected behavior may result.
void Tpetra::deep_copy | ( | MultiVector< DS, DL, DG, DN, dstClassic > & | dst, |
const MultiVector< SS, SL, SG, SN, srcClassic > & | src | ||
) |
Copy the contents of the MultiVector src
into dst
.
src
must be compatible with the Map of dst
. Copy the contents of the MultiVector src
into the MultiVector dst
. ("Copy the contents" means the same thing as "deep
copy.") The two MultiVectors need not necessarily have the same template parameters, but the assignment of their entries must make sense. Furthermore, their Maps must be compatible, that is, the MultiVectors' local dimensions must be the same on all processes.
This method must always be called as a collective operation on all processes over which the multivector is distributed. This is because the method reserves the right to check for compatibility of the two Maps, at least in debug mode, and throw if they are not compatible.
Definition at line 2245 of file Tpetra_MultiVector_decl.hpp.
MultiVector< ST, LO, GO, NT, classic > Tpetra::createCopy | ( | const MultiVector< ST, LO, GO, NT, classic > & | src | ) |
Return a deep copy of the given MultiVector.
Definition at line 4234 of file Tpetra_MultiVector_def.hpp.
Teuchos::RCP< MultiVector< Scalar, LocalOrdinal, GlobalOrdinal, Node, classic > > createMultiVector | ( | const Teuchos::RCP< const Map< LocalOrdinal, GlobalOrdinal, Node > > & | map, |
const size_t | numVectors | ||
) |
Nonmember MultiVector "constructor": Create a MultiVector from a given Map.
map | [in] Map describing the distribution of rows of the resulting MultiVector. |
numVectors | [in] Number of columns of the resulting MultiVector. |
void Tpetra::deep_copy | ( | MultiVector< ST, LO, GO, NT, classic > & | dst, |
const MultiVector< ST, LO, GO, NT, classic > & | src | ||
) |
Specialization of deep_copy for MultiVector objects with the same template parameters.
Definition at line 2232 of file Tpetra_MultiVector_decl.hpp.
MapType::iterator Tpetra::efficientAddOrUpdate | ( | MapType & | m, |
const KeyArgType & | k, | ||
const ValueArgType & | v | ||
) |
Efficiently insert or replace an entry in an std::map.
MapType | Specialization of std::map |
KeyArgType | Type of keys of the std::map |
ValueArgType | Type of values of the std::map |
This function is taken from Scott Meyers' "Effective STL", Item
Definition at line 249 of file Tpetra_Util.hpp.
void Tpetra::sort2 | ( | const IT1 & | first1, |
const IT1 & | last1, | ||
const IT2 & | first2 | ||
) |
Sort the first array, and apply the resulting permutation to the second array.
Sort the values in the first array (represented by the exclusive iterator range first1,last1) in ascending order. Apply the permutation resulting from the sort to the second array (represented by a starting iterator first2).
first1 | A random access iterator pointing to the beginning of the first array. |
last1 | A random access iterator pointing to the end (exclusive) of the first array. |
first2 | A random access iterator pointing to the beginning of the second array. The second array must have no fewer elements than the first array. If the first array has N elements, then the permutation will only be applied to the first N elements of the second array. |
Definition at line 545 of file Tpetra_Util.hpp.
void Tpetra::sort3 | ( | const IT1 & | first1, |
const IT1 & | last1, | ||
const IT2 & | first2, | ||
const IT3 & | first3 | ||
) |
Sort the first array, and apply the same permutation to the second and third arrays.
Sort the values in the first array (represented by the exclusive iterator range first1,last1) in ascending order. Apply the permutation resulting from the sort to the second array (represented by a starting iterator first2) and third array (represented by a starting iterator first3).
first1 | A random access iterator pointing to the beginning of the first array. |
last1 | A random access iterator pointing to the end (exclusive) of the first array. |
first2 | A random access iterator pointing to the beginning of the second array. |
first3 | A random access iterator pointing to the beginning of the third array. |
Definition at line 579 of file Tpetra_Util.hpp.
void Tpetra::merge2 | ( | IT1 & | indResultOut, |
IT2 & | valResultOut, | ||
IT1 | indBeg, | ||
IT1 | indEnd, | ||
IT2 | valBeg, | ||
IT2 | valEnd | ||
) |
Merge values in place, additively, with the same index.
IT1 | Iterator type for the range of indices |
IT2 | Iterator type for the range of values |
indBeg, indEnd defines a half-exclusive (does not include the end) range of indices, and valBeg, valEnd its corresponding range of values. The range of values must have the same number of entries as the range of indices. In every nondecreasing subsequence of indices, this method will merge values that have the same index, by adding the values together. When done, it assigns the new end (exclusive) of the index range to indResultOut, and the new end (exclusive) of the value range to valResultOut. (It is legal for the index range not to be sorted, but then only nondecreasing subsequences will get merged.)
For example, if the indices on input are {0, 1, 1, 3, -1, -1, -1, 0}, and their corresponding values on input are {42.0, -4.0, -3.0, 1.5, 1.0, 2.0, 3.0}, then on exit from this function, the indices are {0, 1, 3, -1, 0}, and the values are {42.0, -7.0, 1.5, 6.0, 100.0}.
On entry to the function, indResultOut may alias indEnd, and valResultOut may alias valEnd. For example, the following code is legal:
However, the following code is not legal, because the return value of std::vector::end()
cannot be modified:
Definition at line 646 of file Tpetra_Util.hpp.
void Tpetra::merge2 | ( | IT1 & | indResultOut, |
IT2 & | valResultOut, | ||
IT1 | indBeg, | ||
IT1 | indEnd, | ||
IT2 | valBeg, | ||
IT2 | valEnd, | ||
BinaryFunction | f | ||
) |
Merge values in place with the same index, using any associative binary function.
IT1 | Iterator type for the range of indices |
IT2 | Iterator type for the range of values |
BinaryFunction | The type of a function that takes two values and returns another value. |
indBeg, indEnd defines a half-exclusive (does not include the end) range of indices, and valBeg, valEnd its corresponding range of values. The range of values must have the same number of entries as the range of indices. In every nondecreasing subsequence of indices, this method will merge values that have the same index, by using the given binary function. When done, it assigns the new end (exclusive) of the index range to indResultOut, and the new end (exclusive) of the value range to valResultOut. (It is legal for the index range not to be sorted, but then only nondecreasing subsequences will get merged.)
For example, if the indices on input are {0, 1, 1, 3, -1, -1, -1, 0}, their corresponding values on input are {42.0, -4.0, -3.0, 1.5, 1.0, 2.0, 3.0}, and the binary function is an instance of std::plus<double>
, then on exit from this function, the indices are {0, 1, 3, -1, 0}, and the values are {42.0, -7.0, 1.5, 6.0, 100.0}.
On entry to the function, indResultOut may alias indEnd, and valResultOut may alias valEnd. For example, the following code is legal:
However, the following code is not legal, because the return value of std::vector::end()
cannot be modified:
Definition at line 735 of file Tpetra_Util.hpp.
void Tpetra::keyValueMerge | ( | KeyInputIterType | keyBeg1, |
KeyInputIterType | keyEnd1, | ||
ValueInputIterType | valBeg1, | ||
ValueInputIterType | valEnd1, | ||
KeyInputIterType | keyBeg2, | ||
KeyInputIterType | keyEnd2, | ||
ValueInputIterType | valBeg2, | ||
ValueInputIterType | valEnd2, | ||
KeyOutputIterType | keyOut, | ||
ValueOutputIterType | valOut, | ||
BinaryFunction | f | ||
) |
Merge two sorted (by keys) sequences of unique (key,value) pairs by combining pairs with equal keys.
keyBeg1 | [in] Start of first sequence of keys. |
keyEnd1 | [in] End (exclusive) of first sequence of keys. |
valBeg1 | [in] Start of first sequence of values. |
valEnd1 | [in] End (exclusive) of first sequence of values. |
keyBeg2 | [in] Start of second sequence of keys. |
keyEnd2 | [in] End (exclusive) of second sequence of keys. |
valBeg2 | [in] Start of second sequence of values. |
valEnd2 | [in] End (exclusive) of second sequence of values. |
keyOut | [in/out] Output sequence of keys. |
valOut | [in/out] Output sequence of values. |
f | [in] Binary associative function to use to combine values whose keys are equal. For example, for simple replacement, use a function like std::project1st (in the SGI extensions to the STL) with both template parameters equal to the value type. For addition, use std::plus with template parameter equal to the value type. |
Definition at line 803 of file Tpetra_Util.hpp.
Vector< Scalar, LocalOrdinal, GlobalOrdinal, Node, classic > Tpetra::createCopy | ( | const Vector< Scalar, LocalOrdinal, GlobalOrdinal, Node, classic > & | src | ) |
Return a deep copy of the given Vector.
Definition at line 301 of file Tpetra_Vector_def.hpp.