/usr/include/trilinos/MueLu_CoupledAggregationCommHelper_decl.hpp is in libtrilinos-muelu-dev 12.4.2-2.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 | // @HEADER
//
// ***********************************************************************
//
// MueLu: A package for multigrid based preconditioning
// Copyright 2012 Sandia Corporation
//
// Under the terms of Contract DE-AC04-94AL85000 with Sandia Corporation,
// the U.S. Government retains certain rights in this software.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// 1. Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
//
// 2. Redistributions in binary form must reproduce the above copyright
// notice, this list of conditions and the following disclaimer in the
// documentation and/or other materials provided with the distribution.
//
// 3. Neither the name of the Corporation nor the names of the
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY SANDIA CORPORATION "AS IS" AND ANY
// EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
// PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL SANDIA CORPORATION OR THE
// CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
// LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
// NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
//
// Questions? Contact
// Jonathan Hu (jhu@sandia.gov)
// Andrey Prokopenko (aprokop@sandia.gov)
// Ray Tuminaro (rstumin@sandia.gov)
//
// ***********************************************************************
//
// @HEADER
#ifndef MUELU_COUPLEDAGGREGATIONCOMMHELPER_DECL_HPP
#define MUELU_COUPLEDAGGREGATIONCOMMHELPER_DECL_HPP
#include <Xpetra_Import_fwd.hpp>
#include <Xpetra_ImportFactory_fwd.hpp>
#include <Xpetra_Vector_fwd.hpp>
#include "MueLu_ConfigDefs.hpp"
#include "MueLu_BaseClass.hpp"
#include "MueLu_CoupledAggregationCommHelper_fwd.hpp"
#include "MueLu_Aggregates.hpp"
namespace MueLu {
/*!
@class CoupledAggregationCommHelper
@brief Helper class for providing arbitrated communication across processors
For more details, see the comments for the ArbitrateAndCommunicate methods.
*/
template <class LocalOrdinal = int, class GlobalOrdinal = LocalOrdinal, class Node = KokkosClassic::DefaultNode::DefaultNodeType>
class CoupledAggregationCommHelper : public BaseClass {
typedef double Scalar; // Scalar type only used for weight: always a double.
#undef MUELU_COUPLEDAGGREGATIONCOMMHELPER_SHORT
#include "MueLu_UseShortNames.hpp"
public:
//! @name Constructors/Destructors.
//@{
//! Constructor.
CoupledAggregationCommHelper(const RCP<const Map> & uniqueMap, const RCP<const Map> & nonUniqueMap);
//! Destructor.
~CoupledAggregationCommHelper() { }
//@}
/*!
@brief This method assigns unknowns to aggregates.
Tie-breaking is possible is using random weights.
@param[in] weights vector of weights that help determine ownership.
@param[in,out] aggregates aggregate data structure
@param[in] perturb flag indicating whether weights should be randomly perturbed for tie-breaking purposes.
*/
void ArbitrateAndCommunicate(Vector &weights, Aggregates &aggregates, const bool perturb) const {
ArbitrateAndCommunicate(weights, *aggregates.GetProcWinner(), &*aggregates.GetVertex2AggId(), perturb);
}
/*!
@brief This class uses a weighted rendezvous algorithm to do a global reduction on a vector that may be based on a non unique map.
A non-unique map is one that has at least one global ID that occurs on two or more processes. For each repeated ID \f$i\f$, the
algorithm finds the maximum value \f$v[i]\f$ in the weight vector \f$v\f$. This value is communicated to all processors that
have \f$i\f$ in their local map. More details are below.
For each GlobalId \f$K\f$ associated with weight.getMap():
-# Find the maximum absolute value of \f$weight[K]\f$ across all
processors and assign this to all local elements of weight[] (across
processors) that are associated with \f$K\f$.
-# Set procWinner[] to the MyPid() that had the largest element.
procWinner[] is still set if only one processor owns a GlobalId.
The ONLY CASE when procWinner[i] is NOT set corresponds to when
all local weights associated with a GlobalId are zero. This allows
one to effectively skip the maximum/winner calculation for a subset
of GlobalId's. This might occur when a processor has already
claimed ownership for a GlobalId and so all local copies have
the same value. We want to skip the maximum calculation with
tiebreaking to avoid another processor claiming ownership.
-# Optionally, set companion[] (across all relevant processors) to the
local companion value associated with the procWinner[] processor.
@param weight[in,out]
- On input, vector of NONNEGATIVE weights.
- On output, \f$ \mbox{weight}[k] \Leftarrow \max(\mbox{weight}[k_{p1}],\dots,\mbox{weight}[k_{pn}]) \f$
where \f$ \mbox{weight}[k_{pj}] \f$ is processor \f$pj\f$'s value for GID \f$k\f$.
@param procWinner[in,out]
- On input, allocated but contents ignored.
- On output, \f$\mbox{procWinner}[k] \Leftarrow pj\f$ such that
\f$\mbox{weight}[k_{pj}] = \max(\mbox{weight}[k_{p1}],...,\mbox{weight}[k_{pn}])\f$, where
\f$ \mbox{weight}[k_{pj}] \f$ is processor \f$pj\f$'s value for GID \f$k\f$.
NOTE: If all input \f$\mbox{weight}[k_{pi}]\f$'s are zero, then \f$\mbox{procWinner}[k]\f$ is left untouched.
@param companion[in,out]
- On input, either NULL or allocated but contents ignored. If NULL, step 3 above is skipped.
- On output, if not null, \f$\mbox{companion}[k] \Leftarrow \mbox{companion}[k_j]\f$ where
\f$\mbox{companion}[k_j]\f$ lives on processor \f$\mbox{procWinner}[k]\f$.
and corresponds to the same GlobalId as \f$k\f$.
NOTE: If for a particular GlobalId, no processor
has a value of procWinner that matches
its MyPid, the corresponding companion
is not altered.
@param perturb[in] Optional arguments that is either true or
false (default: true). weight is perturbed
and the perturbed values are used in step 1)
above. Returned values reflect the perturbed
data. This option avoids having lots of
tiebreaks where the large MyPid() always wins.
*/
/*
Output:
@param weight \f$ weight[k] \Leftarrow \max(weight[k_1],\dots,weight[k_n]) \f$
where \f$ weight[k_j] \f$ live on different processors
but have the same GlobalId as weight[k] on this processor.
@param procWinner procWinner[k] <-- MyPid associated with the
kj yielding the max in
Max(weight[k1],...,weight[kn]) .
See weight Output comments.
NOTE: If all input weight[kj]'s are zero,
then procWinner[k] is left untouched.
@param companion If not null,
companion[k] <-- companion[kj] where
companion[kj] lives on processor procWinner[k].
and corresponds to the same GlobalId as k.
NOTE: If for a particlar GlobalId, no processor
has a value of procWinner that matches
its MyPid, the corresponding companion
is not altered.
*/
void ArbitrateAndCommunicate(Vector &weight, LOVector &procWinner, LOVector *companion, const bool perturb) const; //ArbitrateAndCommunicate(Vector&, LOVector &, LOVector *, const bool) const
/*! @brief Redistribute data in source to dest where both source and dest might have multiple copies of the same global id across many processors.
The source may not have the same value for all of these multiple copies, but on
termination dest will have a unique value for each global id. When multiple
copies exist in source, 'what' determines how they are combined to make a
unique value in dest (see CombineMode).
Input:
@param[in] source Vector where multiple copies of some GlobalIds
might exist and might have different values.
@param[in,out] dest On input, allocated but contents ignored.
On output, contains redistributed data from source where
'what' determines how multiple copies of source
values associated with the same GlobalId are
combined into a unique value on all processors.
@param[in] what Determines how multiple copies of the same
GlobalId are combined (see CombineMode).
*/
void NonUnique2NonUnique(const Vector &source, Vector &dest, const Xpetra::CombineMode what) const;
private:
RCP<const Import> import_;
mutable RCP<const Import> winnerImport_; //FIXME get rid of "mutable"
mutable RCP<Import> pushWinners_; //FIXME get rid of mutable
RCP<Vector> tempVec_;
mutable RCP<Vector> perturbWt_;
mutable RCP<Vector> postComm_;
mutable RCP<Vector> candidateWinners_;
mutable ArrayRCP<GO> myWinners_;
mutable int numMyWinners_;
mutable RCP<Map> winnerMap_;
mutable int numCalls_;
int myPID_;
// uniqueMap A subset of weight.getMap() where each GlobalId
// has only one unique copy on one processor.
// Normally, weight.getMap() would have both locals
// and ghost elements while uniqueMap would just
// have the locals. It should be possible to
// remove this or make it an optional argument
// and use some existing Epetra/Tpetra capability to
// make a uniqueMap.
//
// import_ This corresponds precisely to
// Import import_(
// weight.getMap(), uniqueMap);
// This could also be eliminated and created
// here, but for efficiency user's must pass in.
//
};
}
//JG:
// - procWinner is an array of proc ID -> LocalOrdinal
// - companion == aggregates.GetVertex2AggId() == local aggregate ID -> LocalOrdinal
#define MUELU_COUPLEDAGGREGATIONCOMMHELPER_SHORT
#endif // MUELU_COUPLEDAGGREGATIONCOMMHELPER_DECL_HPP
|