Tpetra parallel linear algebra  Version of the Day
TpetraExt_MatrixMatrix_ExtraKernels_decl.hpp
1 // @HEADER
2 // ***********************************************************************
3 //
4 // Tpetra: Templated Linear Algebra Services Package
5 // Copyright (2008) Sandia Corporation
6 //
7 // Under the terms of Contract DE-AC04-94AL85000 with Sandia Corporation,
8 // the U.S. Government retains certain rights in this software.
9 //
10 // Redistribution and use in source and binary forms, with or without
11 // modification, are permitted provided that the following conditions are
12 // met:
13 //
14 // 1. Redistributions of source code must retain the above copyright
15 // notice, this list of conditions and the following disclaimer.
16 //
17 // 2. Redistributions in binary form must reproduce the above copyright
18 // notice, this list of conditions and the following disclaimer in the
19 // documentation and/or other materials provided with the distribution.
20 //
21 // 3. Neither the name of the Corporation nor the names of the
22 // contributors may be used to endorse or promote products derived from
23 // this software without specific prior written permission.
24 //
25 // THIS SOFTWARE IS PROVIDED BY SANDIA CORPORATION "AS IS" AND ANY
26 // EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 // IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
28 // PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL SANDIA CORPORATION OR THE
29 // CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
30 // EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
31 // PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
32 // PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33 // LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34 // NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35 // SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36 //
37 // Questions? Contact Michael A. Heroux (maherou@sandia.gov)
38 //
39 // ************************************************************************
40 // @HEADER
41 
42 #ifndef TPETRA_MATRIXMATRIX_EXTRAKERNELS_DECL_HPP
43 #define TPETRA_MATRIXMATRIX_EXTRAKERNELS_DECL_HPP
45 
46 
47 namespace Tpetra {
48 
49 namespace MatrixMatrix {
50 
51  // This guy allows us to easily get an Unmanaged Kokkos View from a ManagedOne
52  template <typename View>
53  using UnmanagedView = Kokkos::View< typename View::data_type
54  , typename View::array_layout
55  , typename View::device_type
56  , typename Kokkos::MemoryTraits< Kokkos::Unmanaged>
57  >;
58 
59  namespace ExtraKernels {
60 
61  template<class CrsMatrixType>
62  size_t C_estimate_nnz_per_row(CrsMatrixType & A, CrsMatrixType &B);
63 
64  template<class InRowptrArrayType, class InColindArrayType, class InValsArrayType,
65  class OutRowptrType, class OutColindType, class OutValsType>
66  void copy_out_from_thread_memory(const InRowptrArrayType & Inrowptr, const InColindArrayType &Incolind, const InValsArrayType & Invals,
67  size_t m, double thread_chunk,
68  OutRowptrType & Outrowptr, OutColindType &Outcolind, OutValsType & Outvals);
69 
70  /***************************** OpenMP Only Kernels *****************************/
71 #ifdef HAVE_TPETRA_INST_OPENMP
72  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class LocalOrdinalViewType>
73  static inline void mult_A_B_newmatrix_LowThreadGustavsonKernel(CrsMatrixStruct<Scalar, LocalOrdinal, GlobalOrdinal, Kokkos::Compat::KokkosOpenMPWrapperNode>& Aview,
74  CrsMatrixStruct<Scalar, LocalOrdinal, GlobalOrdinal, Kokkos::Compat::KokkosOpenMPWrapperNode>& Bview,
75  const LocalOrdinalViewType & Acol2Brow,
76  const LocalOrdinalViewType & Acol2Irow,
77  const LocalOrdinalViewType & Bcol2Ccol,
78  const LocalOrdinalViewType & Icol2Ccol,
79  CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Kokkos::Compat::KokkosOpenMPWrapperNode>& C,
80  Teuchos::RCP<const Import<LocalOrdinal,GlobalOrdinal,Kokkos::Compat::KokkosOpenMPWrapperNode> > Cimport,
81  const std::string& label,
82  const Teuchos::RCP<Teuchos::ParameterList>& params);
83 
84  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class LocalOrdinalViewType>
85  static inline void mult_A_B_reuse_LowThreadGustavsonKernel(CrsMatrixStruct<Scalar, LocalOrdinal, GlobalOrdinal, Kokkos::Compat::KokkosOpenMPWrapperNode>& Aview,
86  CrsMatrixStruct<Scalar, LocalOrdinal, GlobalOrdinal, Kokkos::Compat::KokkosOpenMPWrapperNode>& Bview,
87  const LocalOrdinalViewType & Acol2Brow,
88  const LocalOrdinalViewType & Acol2Irow,
89  const LocalOrdinalViewType & Bcol2Ccol,
90  const LocalOrdinalViewType & Icol2Ccol,
91  CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Kokkos::Compat::KokkosOpenMPWrapperNode>& C,
92  Teuchos::RCP<const Import<LocalOrdinal,GlobalOrdinal,Kokkos::Compat::KokkosOpenMPWrapperNode> > Cimport,
93  const std::string& label,
94  const Teuchos::RCP<Teuchos::ParameterList>& params);
95 
96  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class LocalOrdinalViewType>
97  static inline void jacobi_A_B_newmatrix_LowThreadGustavsonKernel(Scalar omega,
98  const Vector<Scalar,LocalOrdinal,GlobalOrdinal, Kokkos::Compat::KokkosOpenMPWrapperNode> & Dinv,
99  CrsMatrixStruct<Scalar, LocalOrdinal, GlobalOrdinal, Kokkos::Compat::KokkosOpenMPWrapperNode>& Aview,
100  CrsMatrixStruct<Scalar, LocalOrdinal, GlobalOrdinal, Kokkos::Compat::KokkosOpenMPWrapperNode>& Bview,
101  const LocalOrdinalViewType & Acol2Brow,
102  const LocalOrdinalViewType & Acol2Irow,
103  const LocalOrdinalViewType & Bcol2Ccol,
104  const LocalOrdinalViewType & Icol2Ccol,
105  CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Kokkos::Compat::KokkosOpenMPWrapperNode>& C,
106  Teuchos::RCP<const Import<LocalOrdinal,GlobalOrdinal,Kokkos::Compat::KokkosOpenMPWrapperNode> > Cimport,
107  const std::string& label,
108  const Teuchos::RCP<Teuchos::ParameterList>& params);
109 
110  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class LocalOrdinalViewType>
111  static inline void jacobi_A_B_reuse_LowThreadGustavsonKernel(Scalar omega,
112  const Vector<Scalar,LocalOrdinal,GlobalOrdinal, Kokkos::Compat::KokkosOpenMPWrapperNode> & Dinv,
113  CrsMatrixStruct<Scalar, LocalOrdinal, GlobalOrdinal, Kokkos::Compat::KokkosOpenMPWrapperNode>& Aview,
114  CrsMatrixStruct<Scalar, LocalOrdinal, GlobalOrdinal, Kokkos::Compat::KokkosOpenMPWrapperNode>& Bview,
115  const LocalOrdinalViewType & Acol2Brow,
116  const LocalOrdinalViewType & Acol2Irow,
117  const LocalOrdinalViewType & Bcol2Ccol,
118  const LocalOrdinalViewType & Icol2Ccol,
119  CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Kokkos::Compat::KokkosOpenMPWrapperNode>& C,
120  Teuchos::RCP<const Import<LocalOrdinal,GlobalOrdinal,Kokkos::Compat::KokkosOpenMPWrapperNode> > Cimport,
121  const std::string& label,
122  const Teuchos::RCP<Teuchos::ParameterList>& params);
123 #endif
124 
125  /***************************** Generic Kernels *****************************/
126  template<class Scalar, class LocalOrdinal, class GlobalOrdinal, class Node, class LocalOrdinalViewType>
127  static inline void jacobi_A_B_newmatrix_MultiplyScaleAddKernel(Scalar omega,
128  const Vector<Scalar,LocalOrdinal,GlobalOrdinal, Node> & Dinv,
129  CrsMatrixStruct<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Aview,
130  CrsMatrixStruct<Scalar, LocalOrdinal, GlobalOrdinal, Node>& Bview,
131  const LocalOrdinalViewType & Acol2Brow,
132  const LocalOrdinalViewType & Acol2Irow,
133  const LocalOrdinalViewType & Bcol2Ccol,
134  const LocalOrdinalViewType & Icol2Ccol,
135  CrsMatrix<Scalar, LocalOrdinal, GlobalOrdinal, Node>& C,
136  Teuchos::RCP<const Import<LocalOrdinal,GlobalOrdinal,Node> > Cimport,
137  const std::string& label,
138  const Teuchos::RCP<Teuchos::ParameterList>& params);
139 
140  }// ExtraKernels
141 }//MatrixMatrix
142 }//Tpetra
143 
144 
145 
146 #endif
TpetraExt_MatrixMatrix_decl.hpp
Tpetra
Namespace Tpetra contains the class and methods constituting the Tpetra library.