Zoltan 2 Version 0.5
Public Member Functions | Static Public Member Functions
Zoltan2::MatrixInput< User > Class Template Reference

MatrixInput defines the interface for input adapters for matrices. More...

#include <Zoltan2_MatrixInput.hpp>

Inheritance diagram for Zoltan2::MatrixInput< User >:
Inheritance graph
[legend]
Collaboration diagram for Zoltan2::MatrixInput< User >:
Collaboration graph
[legend]

List of all members.

Public Member Functions

enum InputAdapterType inputAdapterType () const
 Returns the type of adapter.
virtual ~MatrixInput ()
 Destructor.
virtual size_t getLocalNumRows () const =0
 Returns the number rows on this process.
virtual size_t getLocalNumColumns () const =0
 Returns the number columns on this process.
virtual size_t getRowListView (const gid_t *&rowIds, const lno_t *&offsets, const gid_t *&colIds) const =0
 Sets pointers to this process' matrix entries. If this optional call is defined in the adapter, it can save a memory copy of application data.
virtual int getRowWeightDimension () const =0
 Returns the dimension (0 or greater) of row weights.
virtual size_t getRowWeights (int weightDim, const scalar_t *&weights, int &stride) const =0
 Provide a pointer to the row weights, if any.
virtual bool getRowWeightIsNumberOfNonZeros (int dim) const =0
 Is the row weight for a dimension the number of row non-zeros?
virtual int getCoordinateDimension () const =0
 Returns the dimension of the geometry, if any.
virtual size_t getRowCoordinates (int coordDim, const scalar_t *&coords, int &stride) const =0
 Provide a pointer to one dimension of row coordinates.
template<typename Adapter >
size_t applyPartitioningSolution (const User &in, User *&out, const PartitioningSolution< Adapter > &solution) const
 Apply the solution to a partitioning problem to an input.
virtual string inputAdapterName () const =0
 Returns a descriptive name that identifies the concrete adapter.
virtual size_t getLocalNumberOfObjects () const =0
 Returns the number of objects in the input.
virtual int getNumberOfWeightsPerObject () const =0
 Returns the number of weights per object. Number of weights per object should be zero or greater. If zero, then it is assumed that all objects are equally weighted.
virtual size_t getObjectWeights (int dim, const scalar_t *&wgt, int &stride) const =0
 Provide pointer to a weight array with stride.

Static Public Member Functions

static string inputAdapterTypeName (InputAdapterType iaType)
 Returns the name of the input adapter.

Detailed Description

template<typename User>
class Zoltan2::MatrixInput< User >

MatrixInput defines the interface for input adapters for matrices.

InputAdapter objects provide access for Zoltan2 to the user's data. Many built-in adapters are already defined for common data structures, such as Tpetra and Epetra objects and C-language pointers to arrays.

Data types:

See IdentifierTraits to understand why the user's global ID type (gid_t) may differ from that used by Zoltan2 (gno_t).

The Kokkos node type can be safely ignored.

The template parameter User is a user-defined data type which, through a traits mechanism, provides the actual data types with which the Zoltan2 library will be compiled. User may be the actual class or structure used by application to represent a vector, or it may be the helper class BasicUserTypes. See InputTraits for more information.

The scalar_t type, representing use data such as matrix values, is used by Zoltan2 for weights, coordinates, part sizes and quality metrics. Some User types (like Tpetra::CrsMatrix) have an inherent scalar type, and some (like Tpetra::CrsGraph) do not. For such objects, the scalar type is set by Zoltan2 to float. If you wish to change it to double, set the second template parameter to double.

Todo:

Create BasicCrsMatrixInput subclass

Do we want to require input adapters to give us the global number of rows, columns etc? We can figure that out.

This is a row-oriented matrix. Do we need a column-oriented matrix? In particular - we assumed coordinates are for rows.

If the user can tell us there are no diagonal entries in a square matrix, it can save time if we have to remove them for the algorithm. Should we have a set method in subclasses for setMatrixHasDiagonalEntries yes, no and maybe?

Definition at line 104 of file Zoltan2_MatrixInput.hpp.


Constructor & Destructor Documentation

template<typename User >
virtual Zoltan2::MatrixInput< User >::~MatrixInput ( ) [inline, virtual]

Destructor.

Definition at line 122 of file Zoltan2_MatrixInput.hpp.


Member Function Documentation

template<typename User >
enum InputAdapterType Zoltan2::MatrixInput< User >::inputAdapterType ( ) const [inline, virtual]

Returns the type of adapter.

Implements Zoltan2::InputAdapter< User >.

Definition at line 118 of file Zoltan2_MatrixInput.hpp.

template<typename User >
virtual size_t Zoltan2::MatrixInput< User >::getLocalNumRows ( ) const [pure virtual]

Returns the number rows on this process.

Implemented in Zoltan2::XpetraCrsMatrixInput< User >.

template<typename User >
virtual size_t Zoltan2::MatrixInput< User >::getLocalNumColumns ( ) const [pure virtual]

Returns the number columns on this process.

Implemented in Zoltan2::XpetraCrsMatrixInput< User >.

template<typename User >
virtual size_t Zoltan2::MatrixInput< User >::getRowListView ( const gid_t *&  rowIds,
const lno_t *&  offsets,
const gid_t *&  colIds 
) const [pure virtual]

Sets pointers to this process' matrix entries. If this optional call is defined in the adapter, it can save a memory copy of application data.

Parameters:
rowIdswill on return a pointer to row global Ids
offsetsis an array of size numRows + 1. The column Ids for rowId[i] begin at colIds[offsets[i]]. The last element of offsets is the size of the colIds array.
colIdson return will point to the global column Ids for the non-zeros for each row.
Returns:
The number of ids in the rowIds list.

Implemented in Zoltan2::XpetraCrsMatrixInput< User >.

template<typename User >
virtual int Zoltan2::MatrixInput< User >::getRowWeightDimension ( ) const [pure virtual]

Returns the dimension (0 or greater) of row weights.

Row weights may be used when partitioning matrix rows. If no weights are provided, and rowWeightIsNumberOfNonZeros() is false, then it is assumed rows are equally weighted.

Implemented in Zoltan2::XpetraCrsMatrixInput< User >.

template<typename User >
virtual size_t Zoltan2::MatrixInput< User >::getRowWeights ( int  weightDim,
const scalar_t *&  weights,
int &  stride 
) const [pure virtual]

Provide a pointer to the row weights, if any.

Parameters:
weightDimranges from zero to one less than getRowWeightDimension().
weightsis the list of weights of the given dimension for the rows returned in getRowListView(). If weights for this dimension are to be uniform for all rows in the global problem, the weights should be a NULL pointer.
strideThe k'th weight is located at weights[stride*k]
Returns:
The number of weights listed, which should be at least the local number of rows times the stride for non-uniform weights, zero otherwise.

Zoltan2 does not copy your data. The data pointed to by weights must remain valid for the lifetime of this InputAdapter.

Implemented in Zoltan2::XpetraCrsMatrixInput< User >.

template<typename User >
virtual bool Zoltan2::MatrixInput< User >::getRowWeightIsNumberOfNonZeros ( int  dim) const [pure virtual]

Is the row weight for a dimension the number of row non-zeros?

Parameters:
dima value between zero and getRowWeightDimension() that specifies the weight dimension that will be equal to the number of row non-zeros.

If you wish for Zoltan2 to automatically assign for one of the weight dimensions for each row a weight that is equal to the number of non-zeros in the row, then return true here. Otherwise return false.

Implemented in Zoltan2::XpetraCrsMatrixInput< User >.

template<typename User >
virtual int Zoltan2::MatrixInput< User >::getCoordinateDimension ( ) const [pure virtual]

Returns the dimension of the geometry, if any.

Some algorithms can use geometric row or column coordinate information if it is present. Given the problem parameters supplied by the user, it may make sense to use row coordinates or it may make sense to use column coordinates.

Implemented in Zoltan2::XpetraCrsMatrixInput< User >.

template<typename User >
virtual size_t Zoltan2::MatrixInput< User >::getRowCoordinates ( int  coordDim,
const scalar_t *&  coords,
int &  stride 
) const [pure virtual]

Provide a pointer to one dimension of row coordinates.

Parameters:
coordDim[input] is a value from 0 to one less than getCoordinateDimension() specifying which dimension is being provided in the coords list.
coords[output] points to a list of coordinate values for the dimension. The order of coords should correspond to the order of rowIds in getRowListView().
stride[output] describes the layout of the coordinate values in the coords list. If stride is one, then the ith coordinate value is coords[i], but if stride is two, then the ith coordinate value is coords[2*i].
Returns:
The length of the coords list. This may be more than getLocalNumberOfVertices() because the stride may be more than one. It may also be zero if column coordinates but not row coordinates are supplied.

Zoltan2 does not copy your data. The data pointed to by coords must remain valid for the lifetime of this InputAdapter.

Implemented in Zoltan2::XpetraCrsMatrixInput< User >.

template<typename User >
template<typename Adapter >
size_t Zoltan2::MatrixInput< User >::applyPartitioningSolution ( const User &  in,
User *&  out,
const PartitioningSolution< Adapter > &  solution 
) const [inline]

Apply the solution to a partitioning problem to an input.

This is not a required part of the MatrixInput interface.

Parameters:
inAn input object with a structure and assignment of of global Ids to processes that matches that of the input data that instantiated this InputAdapter.
outOn return this should point to a newly created object with the specified partitioning.
solutionThe Solution object created by a Problem should be supplied as the third argument. It must have been templated on user data that has the same global ID distribution as this user data.
Returns:
Returns the number of Ids in the new partitioning.

Reimplemented in Zoltan2::XpetraCrsMatrixInput< User >.

Definition at line 239 of file Zoltan2_MatrixInput.hpp.

template<typename User >
virtual string Zoltan2::InputAdapter< User >::inputAdapterName ( ) const [pure virtual, inherited]
template<typename User >
virtual size_t Zoltan2::InputAdapter< User >::getLocalNumberOfObjects ( ) const [pure virtual, inherited]

Returns the number of objects in the input.

Objects may be coordinates, graph vertices, matrix rows, etc. They are the objects to be partitioned, ordered, or colored.

Implemented in Zoltan2::BasicCoordinateInput< User >, Zoltan2::BasicIdentifierInput< User >, Zoltan2::BasicVectorInput< User >, Zoltan2::XpetraCrsGraphInput< User >, Zoltan2::XpetraCrsMatrixInput< User >, Zoltan2::XpetraMultiVectorInput< User >, and Zoltan2::XpetraVectorInput< User >.

template<typename User >
virtual int Zoltan2::InputAdapter< User >::getNumberOfWeightsPerObject ( ) const [pure virtual, inherited]

Returns the number of weights per object. Number of weights per object should be zero or greater. If zero, then it is assumed that all objects are equally weighted.

Implemented in Zoltan2::BasicCoordinateInput< User >, Zoltan2::BasicIdentifierInput< User >, Zoltan2::BasicVectorInput< User >, Zoltan2::XpetraCrsGraphInput< User >, Zoltan2::XpetraCrsMatrixInput< User >, Zoltan2::XpetraMultiVectorInput< User >, and Zoltan2::XpetraVectorInput< User >.

template<typename User >
virtual size_t Zoltan2::InputAdapter< User >::getObjectWeights ( int  dim,
const scalar_t *&  wgt,
int &  stride 
) const [pure virtual, inherited]

Provide pointer to a weight array with stride.

Parameters:
dimthe weight dimension, zero or greater
wgton return a pointer to the weights for this dimension
strideon return, the value such that the nth weight should be found at wgt[n*stride] .
Returns:
the length of the wgt array, which should be at least equal to getLocalNumberOfObjects() * stride .

Implemented in Zoltan2::BasicCoordinateInput< User >, Zoltan2::BasicIdentifierInput< User >, Zoltan2::BasicVectorInput< User >, Zoltan2::XpetraCrsGraphInput< User >, Zoltan2::XpetraCrsMatrixInput< User >, Zoltan2::XpetraMultiVectorInput< User >, and Zoltan2::XpetraVectorInput< User >.

template<typename User >
string Zoltan2::InputAdapter< User >::inputAdapterTypeName ( InputAdapterType  iaType) [static, inherited]

Returns the name of the input adapter.

Definition at line 134 of file Zoltan2_InputAdapter.hpp.


The documentation for this class was generated from the following file: