41 #ifndef TPETRA_BLOCKCRSMATRIX_DECL_HPP
42 #define TPETRA_BLOCKCRSMATRIX_DECL_HPP
47 #include "Tpetra_CrsGraph.hpp"
48 #include "Tpetra_RowMatrix.hpp"
49 #include "Tpetra_BlockMultiVector_decl.hpp"
52 #include "KokkosSparse_BsrMatrix.hpp"
54 #if KOKKOSKERNELS_VERSION >= 40299
55 #include "Tpetra_Details_MatrixApplyHelper.hpp"
60 template<
class BlockCrsMatrixType>
61 Teuchos::RCP<BlockCrsMatrixType>
62 importAndFillCompleteBlockCrsMatrix (
const Teuchos::RCP<const BlockCrsMatrixType>& sourceMatrix,
63 const Import<
typename BlockCrsMatrixType::local_ordinal_type,
64 typename BlockCrsMatrixType::global_ordinal_type,
65 typename BlockCrsMatrixType::node_type>& importer);
66 template<
class BlockCrsMatrixType>
67 Teuchos::RCP<BlockCrsMatrixType>
68 exportAndFillCompleteBlockCrsMatrix (
const Teuchos::RCP<const BlockCrsMatrixType>& sourceMatrix,
69 const Export<
typename BlockCrsMatrixType::local_ordinal_type,
70 typename BlockCrsMatrixType::global_ordinal_type,
71 typename BlockCrsMatrixType::node_type>& exporter);
153 #if defined(TPETRA_ENABLE_BLOCKCRS_LITTLEBLOCK_LAYOUTLEFT)
160 template<
class Scalar,
171 using STS = Teuchos::ScalarTraits<Scalar>;
212 typedef ::Tpetra::MultiVector<Scalar, LO, GO, node_type>
mv_type;
220 Kokkos::MemoryTraits<Kokkos::Unmanaged> >
222 typedef typename little_block_type::HostMirror little_block_host_type;
228 Kokkos::MemoryTraits<Kokkos::Unmanaged> >
232 typedef typename BMV::little_host_vec_type little_host_vec_type;
236 typedef typename BMV::const_little_host_vec_type const_host_little_vec_type;
239 using local_inds_device_view_type =
240 typename row_matrix_type::local_inds_device_view_type;
241 using local_inds_host_view_type =
242 typename row_matrix_type::local_inds_host_view_type;
243 using nonconst_local_inds_host_view_type =
244 typename row_matrix_type::nonconst_local_inds_host_view_type;
246 using global_inds_device_view_type =
247 typename row_matrix_type::global_inds_device_view_type;
248 using global_inds_host_view_type =
249 typename row_matrix_type::global_inds_host_view_type;
250 using nonconst_global_inds_host_view_type =
251 typename row_matrix_type::nonconst_global_inds_host_view_type;
253 using values_device_view_type =
254 typename row_matrix_type::values_device_view_type;
255 using values_host_view_type =
256 typename row_matrix_type::values_host_view_type;
257 using nonconst_values_host_view_type =
258 typename row_matrix_type::nonconst_values_host_view_type;
262 using local_matrix_device_type =
267 typename local_graph_device_type::size_type>;
269 #if KOKKOSKERNELS_VERSION >= 40299
273 using local_matrix_int_rowptrs_device_type =
282 local_matrix_device_type,
283 local_matrix_int_rowptrs_device_type,
284 typename mv_type::device_view_type>;
293 mutable std::shared_ptr<ApplyHelper> applyHelper;
298 using local_matrix_host_type =
299 typename local_matrix_device_type::HostMirror;
320 const typename local_matrix_device_type::values_type& values,
343 Teuchos::RCP<const map_type>
getDomainMap ()
const override;
346 Teuchos::RCP<const map_type>
getRangeMap ()
const override;
349 Teuchos::RCP<const map_type>
getRowMap ()
const override;
352 Teuchos::RCP<const map_type>
getColMap ()
const override;
374 Teuchos::ETransp mode = Teuchos::NO_TRANS,
375 Scalar alpha = Teuchos::ScalarTraits<Scalar>::one (),
376 Scalar beta = Teuchos::ScalarTraits<Scalar>::zero ())
const override;
420 describe (Teuchos::FancyOStream& out,
421 const Teuchos::EVerbosityLevel verbLevel)
const override;
431 virtual Teuchos::RCP<const ::Tpetra::RowGraph<LO,GO,Node> >
getGraph ()
const override;
440 applyBlock (
const BlockMultiVector<Scalar, LO, GO, Node>& X,
441 BlockMultiVector<Scalar, LO, GO, Node>& Y,
442 Teuchos::ETransp mode = Teuchos::NO_TRANS,
443 const Scalar alpha = Teuchos::ScalarTraits<Scalar>::one (),
444 const Scalar beta = Teuchos::ScalarTraits<Scalar>::zero ());
450 const Import<LO, GO, Node>& importer)
const;
456 const Export<LO, GO, Node>& exporter)
const;
489 const LO numColInds)
const;
521 const LO numColInds)
const;
558 local_inds_host_view_type &indices,
559 values_host_view_type &values)
const override;
565 local_inds_host_view_type &indices,
566 nonconst_values_host_view_type &values)
const;
571 nonconst_local_inds_host_view_type &Indices,
572 nonconst_values_host_view_type &Values,
573 size_t& NumEntries)
const override;
575 getLocalBlockDeviceNonConst (
const LO localRowInd,
const LO localColInd)
const;
577 little_block_host_type
578 getLocalBlockHostNonConst (
const LO localRowInd,
const LO localColInd)
const;
608 const LO numColInds)
const;
617 const ptrdiff_t offsets[],
619 const LO numOffsets)
const;
622 absMaxLocalValuesByOffsets (
const LO localRowInd,
623 const ptrdiff_t offsets[],
625 const LO numOffsets)
const;
634 const ptrdiff_t offsets[],
636 const LO numOffsets)
const;
686 return (*errs_).is_null () ? std::string (
"") : (*errs_)->str ();
722 Kokkos::MemoryUnmanaged>& offsets)
const;
740 Kokkos::MemoryUnmanaged>& diag,
742 Kokkos::MemoryUnmanaged>& offsets)
const;
763 const LO numColInds)
const;
779 virtual bool checkSizes (const ::Tpetra::SrcDistObject& source)
override;
790 const size_t numSameIDs,
811 Kokkos::DualView<
size_t*,
813 size_t& constantNumPackets)
override;
828 Kokkos::DualView<
size_t*,
830 const size_t constantNumPackets,
837 Teuchos::RCP<crs_graph_type> graphRCP_;
877 using graph_row_offset_host_type =
typename crs_graph_type::local_graph_device_type::row_map_type::HostMirror;
878 graph_row_offset_host_type ptrHost_;
885 using graph_column_indices_host_type =
typename crs_graph_type::local_graph_device_type::entries_type::HostMirror;
886 graph_column_indices_host_type indHost_;
893 using impl_scalar_type_dualview = Kokkos::DualView<impl_scalar_type*, device_type>;
918 Teuchos::RCP<Teuchos::RCP<BMV> > X_colMap_;
922 Teuchos::RCP<Teuchos::RCP<BMV> > Y_rowMap_;
931 Teuchos::RCP<Teuchos::RCP<typename crs_graph_type::import_type> > pointImporter_;
947 Teuchos::RCP<bool> localError_;
956 Teuchos::RCP<Teuchos::RCP<std::ostringstream> > errs_;
959 std::ostream& markLocalErrorAndGetStream ();
964 template<
class Device>
966 #if defined(KOKKOS_ENABLE_CUDA)
972 static constexpr
bool value =
973 std::is_same<typename Device::execution_space, Kokkos::Cuda>::value;
975 #elif defined(KOKKOS_ENABLE_HIP)
976 static constexpr
bool value =
977 std::is_same<typename Device::execution_space, Kokkos::HIP>::value;
978 #elif defined(KOKKOS_ENABLE_SYCL)
979 static constexpr
bool value =
980 std::is_same<typename Device::execution_space, Kokkos::Experimental::SYCL>::value;
982 static constexpr
bool value =
false;
987 typename impl_scalar_type_dualview::t_host::const_type
988 getValuesHost()
const;
990 typename impl_scalar_type_dualview::t_dev::const_type
991 getValuesDevice()
const;
1011 typename impl_scalar_type_dualview::t_host
1014 typename impl_scalar_type_dualview::t_dev
1015 getValuesDeviceNonConst()
const;
1018 typename impl_scalar_type_dualview::t_host::const_type
1019 getValuesHost (
const LO& lclRow)
const;
1022 typename impl_scalar_type_dualview::t_dev::const_type
1023 getValuesDevice (
const LO& lclRow)
const;
1026 typename impl_scalar_type_dualview::t_host
1030 typename impl_scalar_type_dualview::t_dev
1031 getValuesDeviceNonConst (
const LO& lclRow);
1048 const Teuchos::ETransp mode,
1118 findRelOffsetOfColumnIndex (
const LO localRowIndex,
1119 const LO colIndexToFind,
1120 const LO hint = 0)
const;
1124 LO offsetPerBlock ()
const;
1127 getConstLocalBlockFromInput (
const impl_scalar_type* val,
const size_t pointOffset)
const;
1130 getNonConstLocalBlockFromInput (
impl_scalar_type* val,
const size_t pointOffset)
const;
1132 little_block_host_type
1133 getNonConstLocalBlockFromInputHost (
impl_scalar_type* val,
const size_t pointOffset)
const;
1140 virtual Teuchos::RCP<const Teuchos::Comm<int> >
getComm()
const override;
1171 virtual bool hasColMap ()
const override;
1228 nonconst_global_inds_host_view_type &Indices,
1229 nonconst_values_host_view_type &Values,
1230 size_t& NumEntries)
const override;
1257 global_inds_host_view_type & indices,
1258 values_host_view_type & values)
const override;
1282 virtual void leftScale (const ::Tpetra::Vector<Scalar, LO, GO, Node>& x)
override;
1289 virtual void rightScale (const ::Tpetra::Vector<Scalar, LO, GO, Node>& x)
override;
1299 virtual typename ::Tpetra::RowMatrix<Scalar, LO, GO, Node>::mag_type
1304 template<
class BlockCrsMatrixType>
1305 friend Teuchos::RCP<BlockCrsMatrixType>
1306 Tpetra::importAndFillCompleteBlockCrsMatrix (
const Teuchos::RCP<const BlockCrsMatrixType>& sourceMatrix,
1307 const Import<
typename BlockCrsMatrixType::local_ordinal_type,
1308 typename BlockCrsMatrixType::global_ordinal_type,
1309 typename BlockCrsMatrixType::node_type>& importer);
1311 template<
class BlockCrsMatrixType>
1312 friend Teuchos::RCP<BlockCrsMatrixType>
1313 Tpetra::exportAndFillCompleteBlockCrsMatrix (
const Teuchos::RCP<const BlockCrsMatrixType>& sourceMatrix,
1314 const Export<
typename BlockCrsMatrixType::local_ordinal_type,
1315 typename BlockCrsMatrixType::global_ordinal_type,
1316 typename BlockCrsMatrixType::node_type>& exporter);
1320 template<
class BlockCrsMatrixType>
1321 Teuchos::RCP<BlockCrsMatrixType>
1322 importAndFillCompleteBlockCrsMatrix (
const Teuchos::RCP<const BlockCrsMatrixType>& sourceMatrix,
1323 const Import<
typename BlockCrsMatrixType::local_ordinal_type,
1324 typename BlockCrsMatrixType::global_ordinal_type,
1325 typename BlockCrsMatrixType::node_type>& importer)
1327 Teuchos::RCP<BlockCrsMatrixType> destMatrix;
1328 sourceMatrix->importAndFillComplete (destMatrix, importer);
1333 template<
class BlockCrsMatrixType>
1334 Teuchos::RCP<BlockCrsMatrixType>
1335 exportAndFillCompleteBlockCrsMatrix (
const Teuchos::RCP<const BlockCrsMatrixType>& sourceMatrix,
1336 const Export<
typename BlockCrsMatrixType::local_ordinal_type,
1337 typename BlockCrsMatrixType::global_ordinal_type,
1338 typename BlockCrsMatrixType::node_type>& exporter)
1340 Teuchos::RCP<BlockCrsMatrixType> destMatrix;
1341 sourceMatrix->exportAndFillComplete (destMatrix, exporter);
1347 #endif // TPETRA_BLOCKCRSMATRIX_DECL_HPP
virtual void getLocalRowCopy(LO LocalRow, nonconst_local_inds_host_view_type &Indices, nonconst_values_host_view_type &Values, size_t &NumEntries) const override
Not implemented.
size_t getNumEntriesInLocalRow(const LO localRowInd) const override
Return the number of entries in the given row on the calling process.
Communication plan for data redistribution from a uniquely-owned to a (possibly) multiply-owned distr...
virtual size_t getLocalNumEntries() const override
The local number of stored (structurally nonzero) entries.
virtual global_size_t getGlobalNumCols() const override
The global number of columns of this matrix.
::Tpetra::MultiVector< Scalar, LO, GO, node_type > mv_type
The implementation of MultiVector that this class uses.
Kokkos::View< const impl_scalar_type *, device_type > const_little_vec_type
"Const block view" of all degrees of freedom at a mesh point, for a single column of the MultiVector...
typename BMV::impl_scalar_type impl_scalar_type
The implementation type of entries in the matrix.
void getLocalDiagCopy(const Kokkos::View< impl_scalar_type ***, device_type, Kokkos::MemoryUnmanaged > &diag, const Kokkos::View< const size_t *, device_type, Kokkos::MemoryUnmanaged > &offsets) const
Variant of getLocalDiagCopy() that uses precomputed offsets and puts diagonal blocks in a 3-D Kokkos:...
void describe(Teuchos::FancyOStream &out, const Teuchos::EVerbosityLevel verbLevel) const override
Print a description of this object to the given output stream.
Kokkos::View< impl_scalar_type **, Impl::BlockCrsMatrixLittleBlockArrayLayout, device_type, Kokkos::MemoryTraits< Kokkos::Unmanaged > > little_block_type
The type used to access nonconst matrix blocks.
virtual typename::Tpetra::RowMatrix< Scalar, LO, GO, Node >::mag_type getFrobeniusNorm() const override
The Frobenius norm of the matrix.
LO replaceLocalValues(const LO localRowInd, const LO colInds[], const Scalar vals[], const LO numColInds) const
Replace values at the given (mesh, i.e., block) column indices, in the given (mesh, i.e., block) row.
virtual void leftScale(const ::Tpetra::Vector< Scalar, LO, GO, Node > &x) override
Scale the RowMatrix on the left with the given Vector x.
virtual LO getBlockSize() const override
The number of degrees of freedom per mesh point.
LO replaceLocalValuesByOffsets(const LO localRowInd, const ptrdiff_t offsets[], const Scalar vals[], const LO numOffsets) const
Like replaceLocalValues, but avoids computing row offsets.
std::string description() const override
One-line description of this object.
Sparse matrix whose entries are small dense square blocks, all of the same dimensions.
typename DistObject< Scalar, LO, GO, Node >::buffer_device_type buffer_device_type
Kokkos::Device specialization for communication buffers.
void exportAndFillComplete(Teuchos::RCP< BlockCrsMatrix< Scalar, LO, GO, Node > > &destMatrix, const Export< LO, GO, Node > &exporter) const
Import from this to the given destination matrix, and make the result fill complete.
virtual bool hasColMap() const override
Whether this matrix has a well-defined column Map.
virtual void copyAndPermute(const SrcDistObject &source, const size_t numSameIDs, const Kokkos::DualView< const local_ordinal_type *, buffer_device_type > &permuteToLIDs, const Kokkos::DualView< const local_ordinal_type *, buffer_device_type > &permuteFromLIDs, const CombineMode CM)
Perform copies and permutations that are local to the calling (MPI) process.
virtual bool isLocallyIndexed() const override
Whether matrix indices are locally indexed.
bool hasTransposeApply() const override
Whether it is valid to apply the transpose or conjugate transpose of this matrix. ...
LO local_ordinal_type
The type of local indices.
void getLocalRowViewNonConst(LO LocalRow, local_inds_host_view_type &indices, nonconst_values_host_view_type &values) const
::Tpetra::CrsGraph< LO, GO, node_type > crs_graph_type
The implementation of CrsGraph that this class uses.
Kokkos::View< impl_scalar_type *, device_type > little_vec_type
"Block view" of all degrees of freedom at a mesh point, for a single column of the MultiVector...
void setAllToScalar(const Scalar &alpha)
Set all matrix entries equal to alpha.
Node node_type
The Node type.
Kokkos::StaticCrsGraph< local_ordinal_type, Kokkos::LayoutLeft, device_type, void, size_t > local_graph_device_type
The type of the part of the sparse graph on each MPI process.
bool localError() const
Whether this object had an error on the calling process.
Declaration of the Tpetra::CrsMatrix class.
Scalar scalar_type
The type of entries in the matrix (that is, of each entry in each block).
size_t getLocalNumRows() const override
get the local number of block rows
virtual void unpackAndCombine(const Kokkos::DualView< const local_ordinal_type *, buffer_device_type > &importLIDs, Kokkos::DualView< packet_type *, buffer_device_type > imports, Kokkos::DualView< size_t *, buffer_device_type > numPacketsPerLID, const size_t constantNumPackets, const CombineMode combineMode)
Perform any unpacking and combining after communication.
virtual Teuchos::RCP< const Teuchos::Comm< int > > getComm() const override
The communicator over which this matrix is distributed.
Kokkos::LayoutRight BlockCrsMatrixLittleBlockArrayLayout
give an option to use layoutleft
virtual void rightScale(const ::Tpetra::Vector< Scalar, LO, GO, Node > &x) override
Scale the RowMatrix on the right with the given Vector x.
::Tpetra::Map< LO, GO, node_type > map_type
The implementation of Map that this class uses.
void getLocalRowView(LO LocalRow, local_inds_host_view_type &indices, values_host_view_type &values) const override
Get a view of the (mesh, i.e., block) row, using local (mesh, i.e., block) indices.
virtual ~BlockCrsMatrix()
Destructor (declared virtual for memory safety).
MultiVector for multiple degrees of freedom per mesh point.
virtual void copyAndPermute(const SrcDistObject &sourceObj, const size_t numSameIDs, const Kokkos::DualView< const local_ordinal_type *, buffer_device_type > &permuteToLIDs, const Kokkos::DualView< const local_ordinal_type *, buffer_device_type > &permuteFromLIDs, const CombineMode CM) override
Teuchos::RCP< const map_type > getColMap() const override
get the (mesh) map for the columns of this block matrix.
GO global_ordinal_type
The type of global indices.
virtual size_t getGlobalMaxNumRowEntries() const override
The maximum number of entries in any row over all processes in the matrix's communicator.
size_t global_size_t
Global size_t object.
virtual void getGlobalRowView(GO GlobalRow, global_inds_host_view_type &indices, values_host_view_type &values) const override
Get a constant, nonpersisting, globally indexed view of the given row of the matrix.
BlockCrsMatrix()
Default constructor: Makes an empty block matrix.
LO absMaxLocalValues(const LO localRowInd, const LO colInds[], const Scalar vals[], const LO numColInds) const
Variant of getLocalDiagCopy() that uses precomputed offsets and puts diagonal blocks in a 3-D Kokkos:...
Teuchos::RCP< const map_type > getDomainMap() const override
Get the (point) domain Map of this matrix.
device_type::memory_space memory_space
The Kokkos memory space that this class uses.
Teuchos::RCP< const map_type > getRowMap() const override
get the (mesh) map for the rows of this block matrix.
std::string errorMessages() const
The current stream of error messages.
LO sumIntoLocalValuesByOffsets(const LO localRowInd, const ptrdiff_t offsets[], const Scalar vals[], const LO numOffsets) const
Like sumIntoLocalValues, but avoids computing row offsets.
global_size_t getGlobalNumRows() const override
get the global number of block rows
BMV::const_little_vec_type const_little_vec_type
The type used to access const vector blocks.
Communication plan for data redistribution from a (possibly) multiply-owned to a uniquely-owned distr...
virtual Teuchos::RCP< const ::Tpetra::RowGraph< LO, GO, Node > > getGraph() const override
Get the (mesh) graph.
virtual GO getIndexBase() const override
The index base for global indices in this matrix.
virtual void packAndPrepare(const SrcDistObject &source, const Kokkos::DualView< const local_ordinal_type *, buffer_device_type > &exportLIDs, Kokkos::DualView< packet_type *, buffer_device_type > &exports, Kokkos::DualView< size_t *, buffer_device_type > numPacketsPerLID, size_t &constantNumPackets)
Pack data and metadata for communication (sends).
Kokkos::View< const impl_scalar_type **, Impl::BlockCrsMatrixLittleBlockArrayLayout, device_type, Kokkos::MemoryTraits< Kokkos::Unmanaged > > const_little_block_type
The type used to access const matrix blocks.
void applyBlock(const BlockMultiVector< Scalar, LO, GO, Node > &X, BlockMultiVector< Scalar, LO, GO, Node > &Y, Teuchos::ETransp mode=Teuchos::NO_TRANS, const Scalar alpha=Teuchos::ScalarTraits< Scalar >::one(), const Scalar beta=Teuchos::ScalarTraits< Scalar >::zero())
Version of apply() that takes BlockMultiVector input and output.
CombineMode
Rule for combining data in an Import or Export.
virtual bool isGloballyIndexed() const override
Whether matrix indices are globally indexed.
LO getLocalRowOffsets(const LO localRowInd, ptrdiff_t offsets[], const LO colInds[], const LO numColInds) const
Get relative offsets corresponding to the given rows, given by local row index.
size_t getLocalMaxNumRowEntries() const override
Maximum number of entries in any row of the matrix, on this process.
virtual global_size_t getGlobalNumEntries() const override
The global number of stored (structurally nonzero) entries.
void apply(const mv_type &X, mv_type &Y, Teuchos::ETransp mode=Teuchos::NO_TRANS, Scalar alpha=Teuchos::ScalarTraits< Scalar >::one(), Scalar beta=Teuchos::ScalarTraits< Scalar >::zero()) const override
For this matrix A, compute Y := beta * Y + alpha * Op(A) * X.
Abstract base class for objects that can be the source of an Import or Export operation.
device_type::execution_space execution_space
The Kokkos execution space that this class uses.
Teuchos::RCP< const map_type > getRangeMap() const override
Get the (point) range Map of this matrix.
void getLocalDiagOffsets(const Kokkos::View< size_t *, device_type, Kokkos::MemoryUnmanaged > &offsets) const
Get offsets of the diagonal entries in the matrix.
char packet_type
Implementation detail; tells.
local_matrix_device_type getLocalMatrixDevice() const
A read-only, row-oriented interface to a sparse matrix.
A distributed dense vector.
virtual bool supportsRowViews() const override
Whether this object implements getLocalRowView() and getGlobalRowView().
virtual void packAndPrepare(const SrcDistObject &sourceObj, const Kokkos::DualView< const local_ordinal_type *, buffer_device_type > &exportLIDs, Kokkos::DualView< packet_type *, buffer_device_type > &exports, Kokkos::DualView< size_t *, buffer_device_type > numPacketsPerLID, size_t &constantNumPackets) override
BMV::little_vec_type little_vec_type
The type used to access nonconst vector blocks.
virtual void unpackAndCombine(const Kokkos::DualView< const local_ordinal_type *, buffer_device_type > &importLIDs, Kokkos::DualView< packet_type *, buffer_device_type > imports, Kokkos::DualView< size_t *, buffer_device_type > numPacketsPerLID, const size_t constantNumPackets, const CombineMode combineMode) override
virtual void getGlobalRowCopy(GO GlobalRow, nonconst_global_inds_host_view_type &Indices, nonconst_values_host_view_type &Values, size_t &NumEntries) const override
Get a copy of the given global row's entries.
Node::device_type device_type
The Kokkos::Device specialization that this class uses.
virtual bool isFillComplete() const override
Whether fillComplete() has been called.
impl_scalar_type_dualview::t_host getValuesHostNonConst() const
Get the host or device View of the matrix's values (val_).
Base class for distributed Tpetra objects that support data redistribution.
typename mv_type::impl_scalar_type impl_scalar_type
The implementation type of entries in the object.
LO sumIntoLocalValues(const LO localRowInd, const LO colInds[], const Scalar vals[], const LO numColInds) const
Sum into values at the given (mesh, i.e., block) column indices, in the given (mesh, i.e., block) row.
virtual size_t getLocalNumCols() const override
The number of columns needed to apply the forward operator on this node.
void importAndFillComplete(Teuchos::RCP< BlockCrsMatrix< Scalar, LO, GO, Node > > &destMatrix, const Import< LO, GO, Node > &importer) const
Import from this to the given destination matrix, and make the result fill complete.
virtual size_t getNumEntriesInGlobalRow(GO globalRow) const override
The current number of entries on the calling process in the specified global row. ...