ViewVC Help

View File | Revision Log | Show Annotations | View Changeset | Root Listing

root/OpenMD/trunk/src/parallel/ForceMatrixDecomposition.cpp

Comparing:
branches/development/src/parallel/ForceDecomposition.cpp (file contents), Revision 1538 by chuckv, Tue Jan 11 18:58:12 2011 UTC vs.
branches/development/src/parallel/ForceMatrixDecomposition.cpp (file contents), Revision 1756 by gezelter, Mon Jun 18 18:23:20 2012 UTC

#	Line 1 | Line 1
1	<	/**
2	<	* @file ForceDecomposition.cpp
3	<	* @author Charles Vardeman <cvardema.at.nd.edu>
4	<	* @date 08/18/2010
5	<	* @time 11:56am
6	<	* @version 1.0
1	>	/*
2	>	* Copyright (c) 2005 The University of Notre Dame. All Rights Reserved.
3		*
8	–	* @section LICENSE
9	–	* Copyright (c) 2010 The University of Notre Dame. All Rights Reserved.
10	–	*
4		* The University of Notre Dame grants you ("Licensee") a
5		* non-exclusive, royalty free, license to use, modify and
6		* redistribute this software in source and binary code form, provided
#	Line 43 | Line 36
36		* [1]  Meineke, et al., J. Comp. Chem. 26, 252-271 (2005).
37		* [2]  Fennell & Gezelter, J. Chem. Phys. 124, 234104 (2006).
38		* [3]  Sun, Lin & Gezelter, J. Chem. Phys. 128, 24107 (2008).
39	<	* [4]  Vardeman & Gezelter, in progress (2009).
39	>	* [4]  Kuang & Gezelter,  J. Chem. Phys. 133, 164101 (2010).
40	>	* [5]  Vardeman, Stocker & Gezelter, J. Chem. Theory Comput. 7, 834 (2011).
41		*/
42	+	#include "parallel/ForceMatrixDecomposition.hpp"
43	+	#include "math/SquareMatrix3.hpp"
44	+	#include "nonbonded/NonBondedInteraction.hpp"
45	+	#include "brains/SnapshotManager.hpp"
46	+	#include "brains/PairList.hpp"
47
48	+	using namespace std;
49	+	namespace OpenMD {
50
51	+	ForceMatrixDecomposition::ForceMatrixDecomposition(SimInfo* info, InteractionManager* iMan) : ForceDecomposition(info, iMan) {
52
53	<	/*  -*- c++ -*-  */
54	<	#include "config.h"
55	<	#include <stdlib.h>
53	>	// In a parallel computation, row and colum scans must visit all
54	>	// surrounding cells (not just the 14 upper triangular blocks that
55	>	// are used when the processor can see all pairs)
56		#ifdef IS_MPI
57	<	#include <mpi.h>
58	<	#endif
57	>	cellOffsets_.clear();
58	>	cellOffsets_.push_back( Vector3i(-1,-1,-1) );
59	>	cellOffsets_.push_back( Vector3i( 0,-1,-1) );
60	>	cellOffsets_.push_back( Vector3i( 1,-1,-1) );
61	>	cellOffsets_.push_back( Vector3i(-1, 0,-1) );
62	>	cellOffsets_.push_back( Vector3i( 0, 0,-1) );
63	>	cellOffsets_.push_back( Vector3i( 1, 0,-1) );
64	>	cellOffsets_.push_back( Vector3i(-1, 1,-1) );
65	>	cellOffsets_.push_back( Vector3i( 0, 1,-1) );
66	>	cellOffsets_.push_back( Vector3i( 1, 1,-1) );
67	>	cellOffsets_.push_back( Vector3i(-1,-1, 0) );
68	>	cellOffsets_.push_back( Vector3i( 0,-1, 0) );
69	>	cellOffsets_.push_back( Vector3i( 1,-1, 0) );
70	>	cellOffsets_.push_back( Vector3i(-1, 0, 0) );
71	>	cellOffsets_.push_back( Vector3i( 0, 0, 0) );
72	>	cellOffsets_.push_back( Vector3i( 1, 0, 0) );
73	>	cellOffsets_.push_back( Vector3i(-1, 1, 0) );
74	>	cellOffsets_.push_back( Vector3i( 0, 1, 0) );
75	>	cellOffsets_.push_back( Vector3i( 1, 1, 0) );
76	>	cellOffsets_.push_back( Vector3i(-1,-1, 1) );
77	>	cellOffsets_.push_back( Vector3i( 0,-1, 1) );
78	>	cellOffsets_.push_back( Vector3i( 1,-1, 1) );
79	>	cellOffsets_.push_back( Vector3i(-1, 0, 1) );
80	>	cellOffsets_.push_back( Vector3i( 0, 0, 1) );
81	>	cellOffsets_.push_back( Vector3i( 1, 0, 1) );
82	>	cellOffsets_.push_back( Vector3i(-1, 1, 1) );
83	>	cellOffsets_.push_back( Vector3i( 0, 1, 1) );
84	>	cellOffsets_.push_back( Vector3i( 1, 1, 1) );
85	>	#endif
86	>	}
87
58	–	#include <iostream>
59	–	#include <vector>
60	–	#include <algorithm>
61	–	#include <cmath>
62	–	#include "parallel/ForceDecomposition.hpp"
88
89	+	/**
90	+	* distributeInitialData is essentially a copy of the older fortran
91	+	* SimulationSetup
92	+	*/
93	+	void ForceMatrixDecomposition::distributeInitialData() {
94	+	snap_ = sman_->getCurrentSnapshot();
95	+	storageLayout_ = sman_->getStorageLayout();
96	+	ff_ = info_->getForceField();
97	+	nLocal_ = snap_->getNumberOfAtoms();
98	+
99	+	nGroups_ = info_->getNLocalCutoffGroups();
100	+	// gather the information for atomtype IDs (atids):
101	+	idents = info_->getIdentArray();
102	+	AtomLocalToGlobal = info_->getGlobalAtomIndices();
103	+	cgLocalToGlobal = info_->getGlobalGroupIndices();
104	+	vector<int> globalGroupMembership = info_->getGlobalGroupMembership();
105
106	<	using namespace std;
66	<	using namespace OpenMD;
106	>	massFactors = info_->getMassFactors();
107
108	<	//__static
108	>	PairList* excludes = info_->getExcludedInteractions();
109	>	PairList* oneTwo = info_->getOneTwoInteractions();
110	>	PairList* oneThree = info_->getOneThreeInteractions();
111	>	PairList* oneFour = info_->getOneFourInteractions();
112	>
113	>	if (needVelocities_)
114	>	snap_->cgData.setStorageLayout(DataStorage::dslPosition |
115	>	DataStorage::dslVelocity);
116	>	else
117	>	snap_->cgData.setStorageLayout(DataStorage::dslPosition);
118	>
119		#ifdef IS_MPI
120	<	static vector<MPI:Comm> communictors;
121	<	#endif
120	>
121	>	MPI::Intracomm row = rowComm.getComm();
122	>	MPI::Intracomm col = colComm.getComm();
123
124	<	//____ MPITypeTraits
125	<	template<typename T>
126	<	struct MPITypeTraits;
124	>	AtomPlanIntRow = new Plan<int>(row, nLocal_);
125	>	AtomPlanRealRow = new Plan<RealType>(row, nLocal_);
126	>	AtomPlanVectorRow = new Plan<Vector3d>(row, nLocal_);
127	>	AtomPlanMatrixRow = new Plan<Mat3x3d>(row, nLocal_);
128	>	AtomPlanPotRow = new Plan<potVec>(row, nLocal_);
129
130	<	#ifdef IS_MPI
131	<	template<>
132	<	struct MPITypeTraits<RealType> {
133	<	static const MPI::Datatype datatype;
134	<	};
82	<	const MPI_Datatype MPITypeTraits<RealType>::datatype = MY_MPI_REAL;
130	>	AtomPlanIntColumn = new Plan<int>(col, nLocal_);
131	>	AtomPlanRealColumn = new Plan<RealType>(col, nLocal_);
132	>	AtomPlanVectorColumn = new Plan<Vector3d>(col, nLocal_);
133	>	AtomPlanMatrixColumn = new Plan<Mat3x3d>(col, nLocal_);
134	>	AtomPlanPotColumn = new Plan<potVec>(col, nLocal_);
135
136	<	template<>
137	<	struct MPITypeTraits<int> {
138	<	static const MPI::Datatype datatype;
139	<	};
88	<	const MPI::Datatype MPITypeTraits<int>::datatype = MPI_INT;
89	<	#endif
136	>	cgPlanIntRow = new Plan<int>(row, nGroups_);
137	>	cgPlanVectorRow = new Plan<Vector3d>(row, nGroups_);
138	>	cgPlanIntColumn = new Plan<int>(col, nGroups_);
139	>	cgPlanVectorColumn = new Plan<Vector3d>(col, nGroups_);
140
141	<	/**
142	<	* Constructor for ForceDecomposition Parallel Decomposition Method
143	<	* Will try to construct a symmetric grid of processors. Ideally, the
144	<	* number of processors will be a square ex: 4, 9, 16, 25.
95	<	*
96	<	*/
141	>	nAtomsInRow_ = AtomPlanIntRow->getSize();
142	>	nAtomsInCol_ = AtomPlanIntColumn->getSize();
143	>	nGroupsInRow_ = cgPlanIntRow->getSize();
144	>	nGroupsInCol_ = cgPlanIntColumn->getSize();
145
146	<	ForceDecomposition::ForceDecomposition() {
146	>	// Modify the data storage objects with the correct layouts and sizes:
147	>	atomRowData.resize(nAtomsInRow_);
148	>	atomRowData.setStorageLayout(storageLayout_);
149	>	atomColData.resize(nAtomsInCol_);
150	>	atomColData.setStorageLayout(storageLayout_);
151	>	cgRowData.resize(nGroupsInRow_);
152	>	cgRowData.setStorageLayout(DataStorage::dslPosition);
153	>	cgColData.resize(nGroupsInCol_);
154	>	if (needVelocities_)
155	>	// we only need column velocities if we need them.
156	>	cgColData.setStorageLayout(DataStorage::dslPosition |
157	>	DataStorage::dslVelocity);
158	>	else
159	>	cgColData.setStorageLayout(DataStorage::dslPosition);
160	>
161	>	identsRow.resize(nAtomsInRow_);
162	>	identsCol.resize(nAtomsInCol_);
163	>
164	>	AtomPlanIntRow->gather(idents, identsRow);
165	>	AtomPlanIntColumn->gather(idents, identsCol);
166	>
167	>	// allocate memory for the parallel objects
168	>	atypesRow.resize(nAtomsInRow_);
169	>	atypesCol.resize(nAtomsInCol_);
170
171	<	#ifdef IS_MPI
172	<	int nProcs = MPI::COMM_WORLD.Get_size();
173	<	int worldRank = MPI::COMM_WORLD.Get_rank();
171	>	for (int i = 0; i < nAtomsInRow_; i++)
172	>	atypesRow[i] = ff_->getAtomType(identsRow[i]);
173	>	for (int i = 0; i < nAtomsInCol_; i++)
174	>	atypesCol[i] = ff_->getAtomType(identsCol[i]);
175	>
176	>	pot_row.resize(nAtomsInRow_);
177	>	pot_col.resize(nAtomsInCol_);
178	>
179	>	AtomRowToGlobal.resize(nAtomsInRow_);
180	>	AtomColToGlobal.resize(nAtomsInCol_);
181	>	AtomPlanIntRow->gather(AtomLocalToGlobal, AtomRowToGlobal);
182	>	AtomPlanIntColumn->gather(AtomLocalToGlobal, AtomColToGlobal);
183	>
184	>	cgRowToGlobal.resize(nGroupsInRow_);
185	>	cgColToGlobal.resize(nGroupsInCol_);
186	>	cgPlanIntRow->gather(cgLocalToGlobal, cgRowToGlobal);
187	>	cgPlanIntColumn->gather(cgLocalToGlobal, cgColToGlobal);
188	>
189	>	massFactorsRow.resize(nAtomsInRow_);
190	>	massFactorsCol.resize(nAtomsInCol_);
191	>	AtomPlanRealRow->gather(massFactors, massFactorsRow);
192	>	AtomPlanRealColumn->gather(massFactors, massFactorsCol);
193	>
194	>	groupListRow_.clear();
195	>	groupListRow_.resize(nGroupsInRow_);
196	>	for (int i = 0; i < nGroupsInRow_; i++) {
197	>	int gid = cgRowToGlobal[i];
198	>	for (int j = 0; j < nAtomsInRow_; j++) {
199	>	int aid = AtomRowToGlobal[j];
200	>	if (globalGroupMembership[aid] == gid)
201	>	groupListRow_[i].push_back(j);
202	>	}
203	>	}
204	>
205	>	groupListCol_.clear();
206	>	groupListCol_.resize(nGroupsInCol_);
207	>	for (int i = 0; i < nGroupsInCol_; i++) {
208	>	int gid = cgColToGlobal[i];
209	>	for (int j = 0; j < nAtomsInCol_; j++) {
210	>	int aid = AtomColToGlobal[j];
211	>	if (globalGroupMembership[aid] == gid)
212	>	groupListCol_[i].push_back(j);
213	>	}
214	>	}
215	>
216	>	excludesForAtom.clear();
217	>	excludesForAtom.resize(nAtomsInRow_);
218	>	toposForAtom.clear();
219	>	toposForAtom.resize(nAtomsInRow_);
220	>	topoDist.clear();
221	>	topoDist.resize(nAtomsInRow_);
222	>	for (int i = 0; i < nAtomsInRow_; i++) {
223	>	int iglob = AtomRowToGlobal[i];
224	>
225	>	for (int j = 0; j < nAtomsInCol_; j++) {
226	>	int jglob = AtomColToGlobal[j];
227	>
228	>	if (excludes->hasPair(iglob, jglob))
229	>	excludesForAtom[i].push_back(j);
230	>
231	>	if (oneTwo->hasPair(iglob, jglob)) {
232	>	toposForAtom[i].push_back(j);
233	>	topoDist[i].push_back(1);
234	>	} else {
235	>	if (oneThree->hasPair(iglob, jglob)) {
236	>	toposForAtom[i].push_back(j);
237	>	topoDist[i].push_back(2);
238	>	} else {
239	>	if (oneFour->hasPair(iglob, jglob)) {
240	>	toposForAtom[i].push_back(j);
241	>	topoDist[i].push_back(3);
242	>	}
243	>	}
244	>	}
245	>	}
246	>	}
247	>
248	>	#else
249	>	excludesForAtom.clear();
250	>	excludesForAtom.resize(nLocal_);
251	>	toposForAtom.clear();
252	>	toposForAtom.resize(nLocal_);
253	>	topoDist.clear();
254	>	topoDist.resize(nLocal_);
255	>
256	>	for (int i = 0; i < nLocal_; i++) {
257	>	int iglob = AtomLocalToGlobal[i];
258	>
259	>	for (int j = 0; j < nLocal_; j++) {
260	>	int jglob = AtomLocalToGlobal[j];
261	>
262	>	if (excludes->hasPair(iglob, jglob))
263	>	excludesForAtom[i].push_back(j);
264	>
265	>	if (oneTwo->hasPair(iglob, jglob)) {
266	>	toposForAtom[i].push_back(j);
267	>	topoDist[i].push_back(1);
268	>	} else {
269	>	if (oneThree->hasPair(iglob, jglob)) {
270	>	toposForAtom[i].push_back(j);
271	>	topoDist[i].push_back(2);
272	>	} else {
273	>	if (oneFour->hasPair(iglob, jglob)) {
274	>	toposForAtom[i].push_back(j);
275	>	topoDist[i].push_back(3);
276	>	}
277	>	}
278	>	}
279	>	}
280	>	}
281		#endif
282
283	<	// First time through, construct column stride.
284	<	if (communicators.size() == 0)
285	<	{
286	<	int nColumnsMax = (int) round(sqrt((float) nProcs));
287	<	for (int i = 0; i < nProcs; ++i)
288	<	{
289	<	if (nProcs%i==0) nColumns=i;
283	>	// allocate memory for the parallel objects
284	>	atypesLocal.resize(nLocal_);
285	>
286	>	for (int i = 0; i < nLocal_; i++)
287	>	atypesLocal[i] = ff_->getAtomType(idents[i]);
288	>
289	>	groupList_.clear();
290	>	groupList_.resize(nGroups_);
291	>	for (int i = 0; i < nGroups_; i++) {
292	>	int gid = cgLocalToGlobal[i];
293	>	for (int j = 0; j < nLocal_; j++) {
294	>	int aid = AtomLocalToGlobal[j];
295	>	if (globalGroupMembership[aid] == gid) {
296	>	groupList_[i].push_back(j);
297	>	}
298	>	}
299		}
300
301	<	int nRows = nProcs/nColumns;
302	<	myRank_ = (int) worldRank%nColumns;
301	>
302	>	createGtypeCutoffMap();
303	>
304		}
305	<	else
306	<	{
307	<	myRank_ = myRank/nColumns;
308	<	}
309	<	MPI::Comm newComm = MPI:COMM_WORLD.Split(myRank_,0);
310	<
311	<	isColumn_ = false;
312	<
313	<	}
305	>
306	>	void ForceMatrixDecomposition::createGtypeCutoffMap() {
307	>
308	>	RealType tol = 1e-6;
309	>	largestRcut_ = 0.0;
310	>	RealType rc;
311	>	int atid;
312	>	set<AtomType*> atypes = info_->getSimulatedAtomTypes();
313	>
314	>	map<int, RealType> atypeCutoff;
315	>
316	>	for (set<AtomType*>::iterator at = atypes.begin();
317	>	at != atypes.end(); ++at){
318	>	atid = (*at)->getIdent();
319	>	if (userChoseCutoff_)
320	>	atypeCutoff[atid] = userCutoff_;
321	>	else
322	>	atypeCutoff[atid] = interactionMan_->getSuggestedCutoffRadius(*at);
323	>	}
324	>
325	>	vector<RealType> gTypeCutoffs;
326	>	// first we do a single loop over the cutoff groups to find the
327	>	// largest cutoff for any atypes present in this group.
328	>	#ifdef IS_MPI
329	>	vector<RealType> groupCutoffRow(nGroupsInRow_, 0.0);
330	>	groupRowToGtype.resize(nGroupsInRow_);
331	>	for (int cg1 = 0; cg1 < nGroupsInRow_; cg1++) {
332	>	vector<int> atomListRow = getAtomsInGroupRow(cg1);
333	>	for (vector<int>::iterator ia = atomListRow.begin();
334	>	ia != atomListRow.end(); ++ia) {
335	>	int atom1 = (*ia);
336	>	atid = identsRow[atom1];
337	>	if (atypeCutoff[atid] > groupCutoffRow[cg1]) {
338	>	groupCutoffRow[cg1] = atypeCutoff[atid];
339	>	}
340	>	}
341
342	<	ForceDecomposition::gather(sendbuf, receivebuf){
343	<	communicators(myIndex_).Allgatherv();
344	<	}
342	>	bool gTypeFound = false;
343	>	for (int gt = 0; gt < gTypeCutoffs.size(); gt++) {
344	>	if (abs(groupCutoffRow[cg1] - gTypeCutoffs[gt]) < tol) {
345	>	groupRowToGtype[cg1] = gt;
346	>	gTypeFound = true;
347	>	}
348	>	}
349	>	if (!gTypeFound) {
350	>	gTypeCutoffs.push_back( groupCutoffRow[cg1] );
351	>	groupRowToGtype[cg1] = gTypeCutoffs.size() - 1;
352	>	}
353	>
354	>	}
355	>	vector<RealType> groupCutoffCol(nGroupsInCol_, 0.0);
356	>	groupColToGtype.resize(nGroupsInCol_);
357	>	for (int cg2 = 0; cg2 < nGroupsInCol_; cg2++) {
358	>	vector<int> atomListCol = getAtomsInGroupColumn(cg2);
359	>	for (vector<int>::iterator jb = atomListCol.begin();
360	>	jb != atomListCol.end(); ++jb) {
361	>	int atom2 = (*jb);
362	>	atid = identsCol[atom2];
363	>	if (atypeCutoff[atid] > groupCutoffCol[cg2]) {
364	>	groupCutoffCol[cg2] = atypeCutoff[atid];
365	>	}
366	>	}
367	>	bool gTypeFound = false;
368	>	for (int gt = 0; gt < gTypeCutoffs.size(); gt++) {
369	>	if (abs(groupCutoffCol[cg2] - gTypeCutoffs[gt]) < tol) {
370	>	groupColToGtype[cg2] = gt;
371	>	gTypeFound = true;
372	>	}
373	>	}
374	>	if (!gTypeFound) {
375	>	gTypeCutoffs.push_back( groupCutoffCol[cg2] );
376	>	groupColToGtype[cg2] = gTypeCutoffs.size() - 1;
377	>	}
378	>	}
379	>	#else
380
381	+	vector<RealType> groupCutoff(nGroups_, 0.0);
382	+	groupToGtype.resize(nGroups_);
383	+	for (int cg1 = 0; cg1 < nGroups_; cg1++) {
384	+	groupCutoff[cg1] = 0.0;
385	+	vector<int> atomList = getAtomsInGroupRow(cg1);
386	+	for (vector<int>::iterator ia = atomList.begin();
387	+	ia != atomList.end(); ++ia) {
388	+	int atom1 = (*ia);
389	+	atid = idents[atom1];
390	+	if (atypeCutoff[atid] > groupCutoff[cg1])
391	+	groupCutoff[cg1] = atypeCutoff[atid];
392	+	}
393	+
394	+	bool gTypeFound = false;
395	+	for (int gt = 0; gt < gTypeCutoffs.size(); gt++) {
396	+	if (abs(groupCutoff[cg1] - gTypeCutoffs[gt]) < tol) {
397	+	groupToGtype[cg1] = gt;
398	+	gTypeFound = true;
399	+	}
400	+	}
401	+	if (!gTypeFound) {
402	+	gTypeCutoffs.push_back( groupCutoff[cg1] );
403	+	groupToGtype[cg1] = gTypeCutoffs.size() - 1;
404	+	}
405	+	}
406	+	#endif
407
408	+	// Now we find the maximum group cutoff value present in the simulation
409
410	<	ForceDecomposition::scatter(sbuffer, rbuffer){
411	<	communicators(myIndex_).Reduce_scatter(sbuffer, recevbuf. recvcounts, MPI::DOUBLE, MPI::SUM);
412	<	}
410	>	RealType groupMax = *max_element(gTypeCutoffs.begin(),
411	>	gTypeCutoffs.end());
412	>
413	>	#ifdef IS_MPI
414	>	MPI::COMM_WORLD.Allreduce(&groupMax, &groupMax, 1, MPI::REALTYPE,
415	>	MPI::MAX);
416	>	#endif
417	>
418	>	RealType tradRcut = groupMax;
419	>
420	>	for (int i = 0; i < gTypeCutoffs.size();  i++) {
421	>	for (int j = 0; j < gTypeCutoffs.size();  j++) {
422	>	RealType thisRcut;
423	>	switch(cutoffPolicy_) {
424	>	case TRADITIONAL:
425	>	thisRcut = tradRcut;
426	>	break;
427	>	case MIX:
428	>	thisRcut = 0.5 * (gTypeCutoffs[i] + gTypeCutoffs[j]);
429	>	break;
430	>	case MAX:
431	>	thisRcut = max(gTypeCutoffs[i], gTypeCutoffs[j]);
432	>	break;
433	>	default:
434	>	sprintf(painCave.errMsg,
435	>	"ForceMatrixDecomposition::createGtypeCutoffMap "
436	>	"hit an unknown cutoff policy!\n");
437	>	painCave.severity = OPENMD_ERROR;
438	>	painCave.isFatal = 1;
439	>	simError();
440	>	break;
441	>	}
442	>
443	>	pair<int,int> key = make_pair(i,j);
444	>	gTypeCutoffMap[key].first = thisRcut;
445	>	if (thisRcut > largestRcut_) largestRcut_ = thisRcut;
446	>	gTypeCutoffMap[key].second = thisRcut*thisRcut;
447	>	gTypeCutoffMap[key].third = pow(thisRcut + skinThickness_, 2);
448	>	// sanity check
449	>
450	>	if (userChoseCutoff_) {
451	>	if (abs(gTypeCutoffMap[key].first - userCutoff_) > 0.0001) {
452	>	sprintf(painCave.errMsg,
453	>	"ForceMatrixDecomposition::createGtypeCutoffMap "
454	>	"user-specified rCut (%lf) does not match computed group Cutoff\n", userCutoff_);
455	>	painCave.severity = OPENMD_ERROR;
456	>	painCave.isFatal = 1;
457	>	simError();
458	>	}
459	>	}
460	>	}
461	>	}
462	>	}
463	>
464	>	groupCutoffs ForceMatrixDecomposition::getGroupCutoffs(int cg1, int cg2) {
465	>	int i, j;
466	>	#ifdef IS_MPI
467	>	i = groupRowToGtype[cg1];
468	>	j = groupColToGtype[cg2];
469	>	#else
470	>	i = groupToGtype[cg1];
471	>	j = groupToGtype[cg2];
472	>	#endif
473	>	return gTypeCutoffMap[make_pair(i,j)];
474	>	}
475	>
476	>	int ForceMatrixDecomposition::getTopologicalDistance(int atom1, int atom2) {
477	>	for (int j = 0; j < toposForAtom[atom1].size(); j++) {
478	>	if (toposForAtom[atom1][j] == atom2)
479	>	return topoDist[atom1][j];
480	>	}
481	>	return 0;
482	>	}
483	>
484	>	void ForceMatrixDecomposition::zeroWorkArrays() {
485	>	pairwisePot = 0.0;
486	>	embeddingPot = 0.0;
487	>
488	>	#ifdef IS_MPI
489	>	if (storageLayout_ & DataStorage::dslForce) {
490	>	fill(atomRowData.force.begin(), atomRowData.force.end(), V3Zero);
491	>	fill(atomColData.force.begin(), atomColData.force.end(), V3Zero);
492	>	}
493	>
494	>	if (storageLayout_ & DataStorage::dslTorque) {
495	>	fill(atomRowData.torque.begin(), atomRowData.torque.end(), V3Zero);
496	>	fill(atomColData.torque.begin(), atomColData.torque.end(), V3Zero);
497	>	}
498	>
499	>	fill(pot_row.begin(), pot_row.end(),
500	>	Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
501	>
502	>	fill(pot_col.begin(), pot_col.end(),
503	>	Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
504	>
505	>	if (storageLayout_ & DataStorage::dslParticlePot) {
506	>	fill(atomRowData.particlePot.begin(), atomRowData.particlePot.end(),
507	>	0.0);
508	>	fill(atomColData.particlePot.begin(), atomColData.particlePot.end(),
509	>	0.0);
510	>	}
511	>
512	>	if (storageLayout_ & DataStorage::dslDensity) {
513	>	fill(atomRowData.density.begin(), atomRowData.density.end(), 0.0);
514	>	fill(atomColData.density.begin(), atomColData.density.end(), 0.0);
515	>	}
516	>
517	>	if (storageLayout_ & DataStorage::dslFunctional) {
518	>	fill(atomRowData.functional.begin(), atomRowData.functional.end(),
519	>	0.0);
520	>	fill(atomColData.functional.begin(), atomColData.functional.end(),
521	>	0.0);
522	>	}
523	>
524	>	if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
525	>	fill(atomRowData.functionalDerivative.begin(),
526	>	atomRowData.functionalDerivative.end(), 0.0);
527	>	fill(atomColData.functionalDerivative.begin(),
528	>	atomColData.functionalDerivative.end(), 0.0);
529	>	}
530	>
531	>	if (storageLayout_ & DataStorage::dslSkippedCharge) {
532	>	fill(atomRowData.skippedCharge.begin(),
533	>	atomRowData.skippedCharge.end(), 0.0);
534	>	fill(atomColData.skippedCharge.begin(),
535	>	atomColData.skippedCharge.end(), 0.0);
536	>	}
537	>
538	>	if (storageLayout_ & DataStorage::dslFlucQForce) {
539	>	fill(atomRowData.flucQFrc.begin(),
540	>	atomRowData.flucQFrc.end(), 0.0);
541	>	fill(atomColData.flucQFrc.begin(),
542	>	atomColData.flucQFrc.end(), 0.0);
543	>	}
544	>
545	>	if (storageLayout_ & DataStorage::dslElectricField) {
546	>	fill(atomRowData.electricField.begin(),
547	>	atomRowData.electricField.end(), V3Zero);
548	>	fill(atomColData.electricField.begin(),
549	>	atomColData.electricField.end(), V3Zero);
550	>	}
551	>
552	>	if (storageLayout_ & DataStorage::dslFlucQForce) {
553	>	fill(atomRowData.flucQFrc.begin(), atomRowData.flucQFrc.end(),
554	>	0.0);
555	>	fill(atomColData.flucQFrc.begin(), atomColData.flucQFrc.end(),
556	>	0.0);
557	>	}
558	>
559	>	#endif
560	>	// even in parallel, we need to zero out the local arrays:
561	>
562	>	if (storageLayout_ & DataStorage::dslParticlePot) {
563	>	fill(snap_->atomData.particlePot.begin(),
564	>	snap_->atomData.particlePot.end(), 0.0);
565	>	}
566	>
567	>	if (storageLayout_ & DataStorage::dslDensity) {
568	>	fill(snap_->atomData.density.begin(),
569	>	snap_->atomData.density.end(), 0.0);
570	>	}
571	>
572	>	if (storageLayout_ & DataStorage::dslFunctional) {
573	>	fill(snap_->atomData.functional.begin(),
574	>	snap_->atomData.functional.end(), 0.0);
575	>	}
576	>
577	>	if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
578	>	fill(snap_->atomData.functionalDerivative.begin(),
579	>	snap_->atomData.functionalDerivative.end(), 0.0);
580	>	}
581	>
582	>	if (storageLayout_ & DataStorage::dslSkippedCharge) {
583	>	fill(snap_->atomData.skippedCharge.begin(),
584	>	snap_->atomData.skippedCharge.end(), 0.0);
585	>	}
586	>
587	>	if (storageLayout_ & DataStorage::dslElectricField) {
588	>	fill(snap_->atomData.electricField.begin(),
589	>	snap_->atomData.electricField.end(), V3Zero);
590	>	}
591	>	}
592	>
593	>
594	>	void ForceMatrixDecomposition::distributeData()  {
595	>	snap_ = sman_->getCurrentSnapshot();
596	>	storageLayout_ = sman_->getStorageLayout();
597	>	#ifdef IS_MPI
598	>
599	>	// gather up the atomic positions
600	>	AtomPlanVectorRow->gather(snap_->atomData.position,
601	>	atomRowData.position);
602	>	AtomPlanVectorColumn->gather(snap_->atomData.position,
603	>	atomColData.position);
604	>
605	>	// gather up the cutoff group positions
606	>
607	>	cgPlanVectorRow->gather(snap_->cgData.position,
608	>	cgRowData.position);
609	>
610	>	cgPlanVectorColumn->gather(snap_->cgData.position,
611	>	cgColData.position);
612	>
613	>
614	>
615	>	if (needVelocities_) {
616	>	// gather up the atomic velocities
617	>	AtomPlanVectorColumn->gather(snap_->atomData.velocity,
618	>	atomColData.velocity);
619	>
620	>	cgPlanVectorColumn->gather(snap_->cgData.velocity,
621	>	cgColData.velocity);
622	>	}
623	>
624	>
625	>	// if needed, gather the atomic rotation matrices
626	>	if (storageLayout_ & DataStorage::dslAmat) {
627	>	AtomPlanMatrixRow->gather(snap_->atomData.aMat,
628	>	atomRowData.aMat);
629	>	AtomPlanMatrixColumn->gather(snap_->atomData.aMat,
630	>	atomColData.aMat);
631	>	}
632	>
633	>	// if needed, gather the atomic eletrostatic frames
634	>	if (storageLayout_ & DataStorage::dslElectroFrame) {
635	>	AtomPlanMatrixRow->gather(snap_->atomData.electroFrame,
636	>	atomRowData.electroFrame);
637	>	AtomPlanMatrixColumn->gather(snap_->atomData.electroFrame,
638	>	atomColData.electroFrame);
639	>	}
640	>
641	>	// if needed, gather the atomic fluctuating charge values
642	>	if (storageLayout_ & DataStorage::dslFlucQPosition) {
643	>	AtomPlanRealRow->gather(snap_->atomData.flucQPos,
644	>	atomRowData.flucQPos);
645	>	AtomPlanRealColumn->gather(snap_->atomData.flucQPos,
646	>	atomColData.flucQPos);
647	>	}
648	>
649	>	#endif
650	>	}
651	>
652	>	/* collects information obtained during the pre-pair loop onto local
653	>	* data structures.
654	>	*/
655	>	void ForceMatrixDecomposition::collectIntermediateData() {
656	>	snap_ = sman_->getCurrentSnapshot();
657	>	storageLayout_ = sman_->getStorageLayout();
658	>	#ifdef IS_MPI
659	>
660	>	if (storageLayout_ & DataStorage::dslDensity) {
661	>
662	>	AtomPlanRealRow->scatter(atomRowData.density,
663	>	snap_->atomData.density);
664	>
665	>	int n = snap_->atomData.density.size();
666	>	vector<RealType> rho_tmp(n, 0.0);
667	>	AtomPlanRealColumn->scatter(atomColData.density, rho_tmp);
668	>	for (int i = 0; i < n; i++)
669	>	snap_->atomData.density[i] += rho_tmp[i];
670	>	}
671	>
672	>	if (storageLayout_ & DataStorage::dslElectricField) {
673	>
674	>	AtomPlanVectorRow->scatter(atomRowData.electricField,
675	>	snap_->atomData.electricField);
676	>
677	>	int n = snap_->atomData.electricField.size();
678	>	vector<Vector3d> field_tmp(n, V3Zero);
679	>	AtomPlanVectorColumn->scatter(atomColData.electricField, field_tmp);
680	>	for (int i = 0; i < n; i++)
681	>	snap_->atomData.electricField[i] += field_tmp[i];
682	>	}
683	>	#endif
684	>	}
685	>
686	>	/*
687	>	* redistributes information obtained during the pre-pair loop out to
688	>	* row and column-indexed data structures
689	>	*/
690	>	void ForceMatrixDecomposition::distributeIntermediateData() {
691	>	snap_ = sman_->getCurrentSnapshot();
692	>	storageLayout_ = sman_->getStorageLayout();
693	>	#ifdef IS_MPI
694	>	if (storageLayout_ & DataStorage::dslFunctional) {
695	>	AtomPlanRealRow->gather(snap_->atomData.functional,
696	>	atomRowData.functional);
697	>	AtomPlanRealColumn->gather(snap_->atomData.functional,
698	>	atomColData.functional);
699	>	}
700	>
701	>	if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
702	>	AtomPlanRealRow->gather(snap_->atomData.functionalDerivative,
703	>	atomRowData.functionalDerivative);
704	>	AtomPlanRealColumn->gather(snap_->atomData.functionalDerivative,
705	>	atomColData.functionalDerivative);
706	>	}
707	>	#endif
708	>	}
709	>
710	>
711	>	void ForceMatrixDecomposition::collectData() {
712	>	snap_ = sman_->getCurrentSnapshot();
713	>	storageLayout_ = sman_->getStorageLayout();
714	>	#ifdef IS_MPI
715	>	int n = snap_->atomData.force.size();
716	>	vector<Vector3d> frc_tmp(n, V3Zero);
717	>
718	>	AtomPlanVectorRow->scatter(atomRowData.force, frc_tmp);
719	>	for (int i = 0; i < n; i++) {
720	>	snap_->atomData.force[i] += frc_tmp[i];
721	>	frc_tmp[i] = 0.0;
722	>	}
723	>
724	>	AtomPlanVectorColumn->scatter(atomColData.force, frc_tmp);
725	>	for (int i = 0; i < n; i++) {
726	>	snap_->atomData.force[i] += frc_tmp[i];
727	>	}
728	>
729	>	if (storageLayout_ & DataStorage::dslTorque) {
730	>
731	>	int nt = snap_->atomData.torque.size();
732	>	vector<Vector3d> trq_tmp(nt, V3Zero);
733	>
734	>	AtomPlanVectorRow->scatter(atomRowData.torque, trq_tmp);
735	>	for (int i = 0; i < nt; i++) {
736	>	snap_->atomData.torque[i] += trq_tmp[i];
737	>	trq_tmp[i] = 0.0;
738	>	}
739	>
740	>	AtomPlanVectorColumn->scatter(atomColData.torque, trq_tmp);
741	>	for (int i = 0; i < nt; i++)
742	>	snap_->atomData.torque[i] += trq_tmp[i];
743	>	}
744	>
745	>	if (storageLayout_ & DataStorage::dslSkippedCharge) {
746
747	+	int ns = snap_->atomData.skippedCharge.size();
748	+	vector<RealType> skch_tmp(ns, 0.0);
749
750	+	AtomPlanRealRow->scatter(atomRowData.skippedCharge, skch_tmp);
751	+	for (int i = 0; i < ns; i++) {
752	+	snap_->atomData.skippedCharge[i] += skch_tmp[i];
753	+	skch_tmp[i] = 0.0;
754	+	}
755	+
756	+	AtomPlanRealColumn->scatter(atomColData.skippedCharge, skch_tmp);
757	+	for (int i = 0; i < ns; i++)
758	+	snap_->atomData.skippedCharge[i] += skch_tmp[i];
759	+
760	+	}
761	+
762	+	if (storageLayout_ & DataStorage::dslFlucQForce) {
763	+
764	+	int nq = snap_->atomData.flucQFrc.size();
765	+	vector<RealType> fqfrc_tmp(nq, 0.0);
766	+
767	+	AtomPlanRealRow->scatter(atomRowData.flucQFrc, fqfrc_tmp);
768	+	for (int i = 0; i < nq; i++) {
769	+	snap_->atomData.flucQFrc[i] += fqfrc_tmp[i];
770	+	fqfrc_tmp[i] = 0.0;
771	+	}
772	+
773	+	AtomPlanRealColumn->scatter(atomColData.flucQFrc, fqfrc_tmp);
774	+	for (int i = 0; i < nq; i++)
775	+	snap_->atomData.flucQFrc[i] += fqfrc_tmp[i];
776	+
777	+	}
778	+
779	+	nLocal_ = snap_->getNumberOfAtoms();
780	+
781	+	vector<potVec> pot_temp(nLocal_,
782	+	Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
783	+
784	+	// scatter/gather pot_row into the members of my column
785	+
786	+	AtomPlanPotRow->scatter(pot_row, pot_temp);
787	+
788	+	for (int ii = 0;  ii < pot_temp.size(); ii++ )
789	+	pairwisePot += pot_temp[ii];
790	+
791	+	if (storageLayout_ & DataStorage::dslParticlePot) {
792	+	// This is the pairwise contribution to the particle pot.  The
793	+	// embedding contribution is added in each of the low level
794	+	// non-bonded routines.  In single processor, this is done in
795	+	// unpackInteractionData, not in collectData.
796	+	for (int ii = 0; ii < N_INTERACTION_FAMILIES; ii++) {
797	+	for (int i = 0; i < nLocal_; i++) {
798	+	// factor of two is because the total potential terms are divided
799	+	// by 2 in parallel due to row/ column scatter
800	+	snap_->atomData.particlePot[i] += 2.0 * pot_temp[i](ii);
801	+	}
802	+	}
803	+	}
804	+
805	+	fill(pot_temp.begin(), pot_temp.end(),
806	+	Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
807	+
808	+	AtomPlanPotColumn->scatter(pot_col, pot_temp);
809	+
810	+	for (int ii = 0;  ii < pot_temp.size(); ii++ )
811	+	pairwisePot += pot_temp[ii];
812	+
813	+	if (storageLayout_ & DataStorage::dslParticlePot) {
814	+	// This is the pairwise contribution to the particle pot.  The
815	+	// embedding contribution is added in each of the low level
816	+	// non-bonded routines.  In single processor, this is done in
817	+	// unpackInteractionData, not in collectData.
818	+	for (int ii = 0; ii < N_INTERACTION_FAMILIES; ii++) {
819	+	for (int i = 0; i < nLocal_; i++) {
820	+	// factor of two is because the total potential terms are divided
821	+	// by 2 in parallel due to row/ column scatter
822	+	snap_->atomData.particlePot[i] += 2.0 * pot_temp[i](ii);
823	+	}
824	+	}
825	+	}
826	+
827	+	if (storageLayout_ & DataStorage::dslParticlePot) {
828	+	int npp = snap_->atomData.particlePot.size();
829	+	vector<RealType> ppot_temp(npp, 0.0);
830	+
831	+	// This is the direct or embedding contribution to the particle
832	+	// pot.
833	+
834	+	AtomPlanRealRow->scatter(atomRowData.particlePot, ppot_temp);
835	+	for (int i = 0; i < npp; i++) {
836	+	snap_->atomData.particlePot[i] += ppot_temp[i];
837	+	}
838	+
839	+	fill(ppot_temp.begin(), ppot_temp.end(), 0.0);
840	+
841	+	AtomPlanRealColumn->scatter(atomColData.particlePot, ppot_temp);
842	+	for (int i = 0; i < npp; i++) {
843	+	snap_->atomData.particlePot[i] += ppot_temp[i];
844	+	}
845	+	}
846	+
847	+	for (int ii = 0; ii < N_INTERACTION_FAMILIES; ii++) {
848	+	RealType ploc1 = pairwisePot[ii];
849	+	RealType ploc2 = 0.0;
850	+	MPI::COMM_WORLD.Allreduce(&ploc1, &ploc2, 1, MPI::REALTYPE, MPI::SUM);
851	+	pairwisePot[ii] = ploc2;
852	+	}
853	+
854	+	// Here be dragons.
855	+	MPI::Intracomm col = colComm.getComm();
856	+
857	+	col.Allreduce(MPI::IN_PLACE,
858	+	&snap_->frameData.conductiveHeatFlux[0], 3,
859	+	MPI::REALTYPE, MPI::SUM);
860	+
861	+
862	+	#endif
863	+
864	+	}
865	+
866	+	/**
867	+	* Collects information obtained during the post-pair (and embedding
868	+	* functional) loops onto local data structures.
869	+	*/
870	+	void ForceMatrixDecomposition::collectSelfData() {
871	+	snap_ = sman_->getCurrentSnapshot();
872	+	storageLayout_ = sman_->getStorageLayout();
873	+
874	+	#ifdef IS_MPI
875	+	for (int ii = 0; ii < N_INTERACTION_FAMILIES; ii++) {
876	+	RealType ploc1 = embeddingPot[ii];
877	+	RealType ploc2 = 0.0;
878	+	MPI::COMM_WORLD.Allreduce(&ploc1, &ploc2, 1, MPI::REALTYPE, MPI::SUM);
879	+	embeddingPot[ii] = ploc2;
880	+	}
881	+	#endif
882	+
883	+	}
884	+
885	+
886	+
887	+	int ForceMatrixDecomposition::getNAtomsInRow() {
888	+	#ifdef IS_MPI
889	+	return nAtomsInRow_;
890	+	#else
891	+	return nLocal_;
892	+	#endif
893	+	}
894	+
895	+	/**
896	+	* returns the list of atoms belonging to this group.
897	+	*/
898	+	vector<int> ForceMatrixDecomposition::getAtomsInGroupRow(int cg1){
899	+	#ifdef IS_MPI
900	+	return groupListRow_[cg1];
901	+	#else
902	+	return groupList_[cg1];
903	+	#endif
904	+	}
905	+
906	+	vector<int> ForceMatrixDecomposition::getAtomsInGroupColumn(int cg2){
907	+	#ifdef IS_MPI
908	+	return groupListCol_[cg2];
909	+	#else
910	+	return groupList_[cg2];
911	+	#endif
912	+	}
913	+
914	+	Vector3d ForceMatrixDecomposition::getIntergroupVector(int cg1, int cg2){
915	+	Vector3d d;
916	+
917	+	#ifdef IS_MPI
918	+	d = cgColData.position[cg2] - cgRowData.position[cg1];
919	+	#else
920	+	d = snap_->cgData.position[cg2] - snap_->cgData.position[cg1];
921	+	#endif
922	+
923	+	snap_->wrapVector(d);
924	+	return d;
925	+	}
926	+
927	+	Vector3d ForceMatrixDecomposition::getGroupVelocityColumn(int cg2){
928	+	#ifdef IS_MPI
929	+	return cgColData.velocity[cg2];
930	+	#else
931	+	return snap_->cgData.velocity[cg2];
932	+	#endif
933	+	}
934	+
935	+	Vector3d ForceMatrixDecomposition::getAtomVelocityColumn(int atom2){
936	+	#ifdef IS_MPI
937	+	return atomColData.velocity[atom2];
938	+	#else
939	+	return snap_->atomData.velocity[atom2];
940	+	#endif
941	+	}
942	+
943	+
944	+	Vector3d ForceMatrixDecomposition::getAtomToGroupVectorRow(int atom1, int cg1){
945	+
946	+	Vector3d d;
947	+
948	+	#ifdef IS_MPI
949	+	d = cgRowData.position[cg1] - atomRowData.position[atom1];
950	+	#else
951	+	d = snap_->cgData.position[cg1] - snap_->atomData.position[atom1];
952	+	#endif
953	+
954	+	snap_->wrapVector(d);
955	+	return d;
956	+	}
957	+
958	+	Vector3d ForceMatrixDecomposition::getAtomToGroupVectorColumn(int atom2, int cg2){
959	+	Vector3d d;
960	+
961	+	#ifdef IS_MPI
962	+	d = cgColData.position[cg2] - atomColData.position[atom2];
963	+	#else
964	+	d = snap_->cgData.position[cg2] - snap_->atomData.position[atom2];
965	+	#endif
966	+
967	+	snap_->wrapVector(d);
968	+	return d;
969	+	}
970	+
971	+	RealType ForceMatrixDecomposition::getMassFactorRow(int atom1) {
972	+	#ifdef IS_MPI
973	+	return massFactorsRow[atom1];
974	+	#else
975	+	return massFactors[atom1];
976	+	#endif
977	+	}
978	+
979	+	RealType ForceMatrixDecomposition::getMassFactorColumn(int atom2) {
980	+	#ifdef IS_MPI
981	+	return massFactorsCol[atom2];
982	+	#else
983	+	return massFactors[atom2];
984	+	#endif
985	+
986	+	}
987	+
988	+	Vector3d ForceMatrixDecomposition::getInteratomicVector(int atom1, int atom2){
989	+	Vector3d d;
990	+
991	+	#ifdef IS_MPI
992	+	d = atomColData.position[atom2] - atomRowData.position[atom1];
993	+	#else
994	+	d = snap_->atomData.position[atom2] - snap_->atomData.position[atom1];
995	+	#endif
996	+
997	+	snap_->wrapVector(d);
998	+	return d;
999	+	}
1000	+
1001	+	vector<int> ForceMatrixDecomposition::getExcludesForAtom(int atom1) {
1002	+	return excludesForAtom[atom1];
1003	+	}
1004	+
1005	+	/**
1006	+	* We need to exclude some overcounted interactions that result from
1007	+	* the parallel decomposition.
1008	+	*/
1009	+	bool ForceMatrixDecomposition::skipAtomPair(int atom1, int atom2, int cg1, int cg2) {
1010	+	int unique_id_1, unique_id_2, group1, group2;
1011	+
1012	+	#ifdef IS_MPI
1013	+	// in MPI, we have to look up the unique IDs for each atom
1014	+	unique_id_1 = AtomRowToGlobal[atom1];
1015	+	unique_id_2 = AtomColToGlobal[atom2];
1016	+	group1 = cgRowToGlobal[cg1];
1017	+	group2 = cgColToGlobal[cg2];
1018	+	#else
1019	+	unique_id_1 = AtomLocalToGlobal[atom1];
1020	+	unique_id_2 = AtomLocalToGlobal[atom2];
1021	+	group1 = cgLocalToGlobal[cg1];
1022	+	group2 = cgLocalToGlobal[cg2];
1023	+	#endif
1024	+
1025	+	if (unique_id_1 == unique_id_2) return true;
1026	+
1027	+	#ifdef IS_MPI
1028	+	// this prevents us from doing the pair on multiple processors
1029	+	if (unique_id_1 < unique_id_2) {
1030	+	if ((unique_id_1 + unique_id_2) % 2 == 0) return true;
1031	+	} else {
1032	+	if ((unique_id_1 + unique_id_2) % 2 == 1) return true;
1033	+	}
1034	+	#endif
1035	+
1036	+	#ifndef IS_MPI
1037	+	if (group1 == group2) {
1038	+	if (unique_id_1 < unique_id_2) return true;
1039	+	}
1040	+	#endif
1041	+
1042	+	return false;
1043	+	}
1044	+
1045	+	/**
1046	+	* We need to handle the interactions for atoms who are involved in
1047	+	* the same rigid body as well as some short range interactions
1048	+	* (bonds, bends, torsions) differently from other interactions.
1049	+	* We'll still visit the pairwise routines, but with a flag that
1050	+	* tells those routines to exclude the pair from direct long range
1051	+	* interactions.  Some indirect interactions (notably reaction
1052	+	* field) must still be handled for these pairs.
1053	+	*/
1054	+	bool ForceMatrixDecomposition::excludeAtomPair(int atom1, int atom2) {
1055	+
1056	+	// excludesForAtom was constructed to use row/column indices in the MPI
1057	+	// version, and to use local IDs in the non-MPI version:
1058	+
1059	+	for (vector<int>::iterator i = excludesForAtom[atom1].begin();
1060	+	i != excludesForAtom[atom1].end(); ++i) {
1061	+	if ( (*i) == atom2 ) return true;
1062	+	}
1063	+
1064	+	return false;
1065	+	}
1066	+
1067	+
1068	+	void ForceMatrixDecomposition::addForceToAtomRow(int atom1, Vector3d fg){
1069	+	#ifdef IS_MPI
1070	+	atomRowData.force[atom1] += fg;
1071	+	#else
1072	+	snap_->atomData.force[atom1] += fg;
1073	+	#endif
1074	+	}
1075	+
1076	+	void ForceMatrixDecomposition::addForceToAtomColumn(int atom2, Vector3d fg){
1077	+	#ifdef IS_MPI
1078	+	atomColData.force[atom2] += fg;
1079	+	#else
1080	+	snap_->atomData.force[atom2] += fg;
1081	+	#endif
1082	+	}
1083	+
1084	+	// filling interaction blocks with pointers
1085	+	void ForceMatrixDecomposition::fillInteractionData(InteractionData &idat,
1086	+	int atom1, int atom2) {
1087	+
1088	+	idat.excluded = excludeAtomPair(atom1, atom2);
1089	+
1090	+	#ifdef IS_MPI
1091	+	idat.atypes = make_pair( atypesRow[atom1], atypesCol[atom2]);
1092	+	//idat.atypes = make_pair( ff_->getAtomType(identsRow[atom1]),
1093	+	//                         ff_->getAtomType(identsCol[atom2]) );
1094	+
1095	+	if (storageLayout_ & DataStorage::dslAmat) {
1096	+	idat.A1 = &(atomRowData.aMat[atom1]);
1097	+	idat.A2 = &(atomColData.aMat[atom2]);
1098	+	}
1099	+
1100	+	if (storageLayout_ & DataStorage::dslElectroFrame) {
1101	+	idat.eFrame1 = &(atomRowData.electroFrame[atom1]);
1102	+	idat.eFrame2 = &(atomColData.electroFrame[atom2]);
1103	+	}
1104	+
1105	+	if (storageLayout_ & DataStorage::dslTorque) {
1106	+	idat.t1 = &(atomRowData.torque[atom1]);
1107	+	idat.t2 = &(atomColData.torque[atom2]);
1108	+	}
1109	+
1110	+	if (storageLayout_ & DataStorage::dslDensity) {
1111	+	idat.rho1 = &(atomRowData.density[atom1]);
1112	+	idat.rho2 = &(atomColData.density[atom2]);
1113	+	}
1114	+
1115	+	if (storageLayout_ & DataStorage::dslFunctional) {
1116	+	idat.frho1 = &(atomRowData.functional[atom1]);
1117	+	idat.frho2 = &(atomColData.functional[atom2]);
1118	+	}
1119	+
1120	+	if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
1121	+	idat.dfrho1 = &(atomRowData.functionalDerivative[atom1]);
1122	+	idat.dfrho2 = &(atomColData.functionalDerivative[atom2]);
1123	+	}
1124	+
1125	+	if (storageLayout_ & DataStorage::dslParticlePot) {
1126	+	idat.particlePot1 = &(atomRowData.particlePot[atom1]);
1127	+	idat.particlePot2 = &(atomColData.particlePot[atom2]);
1128	+	}
1129	+
1130	+	if (storageLayout_ & DataStorage::dslSkippedCharge) {
1131	+	idat.skippedCharge1 = &(atomRowData.skippedCharge[atom1]);
1132	+	idat.skippedCharge2 = &(atomColData.skippedCharge[atom2]);
1133	+	}
1134	+
1135	+	if (storageLayout_ & DataStorage::dslFlucQPosition) {
1136	+	idat.flucQ1 = &(atomRowData.flucQPos[atom1]);
1137	+	idat.flucQ2 = &(atomColData.flucQPos[atom2]);
1138	+	}
1139	+
1140	+	#else
1141	+
1142	+	idat.atypes = make_pair( atypesLocal[atom1], atypesLocal[atom2]);
1143	+
1144	+	if (storageLayout_ & DataStorage::dslAmat) {
1145	+	idat.A1 = &(snap_->atomData.aMat[atom1]);
1146	+	idat.A2 = &(snap_->atomData.aMat[atom2]);
1147	+	}
1148	+
1149	+	if (storageLayout_ & DataStorage::dslElectroFrame) {
1150	+	idat.eFrame1 = &(snap_->atomData.electroFrame[atom1]);
1151	+	idat.eFrame2 = &(snap_->atomData.electroFrame[atom2]);
1152	+	}
1153	+
1154	+	if (storageLayout_ & DataStorage::dslTorque) {
1155	+	idat.t1 = &(snap_->atomData.torque[atom1]);
1156	+	idat.t2 = &(snap_->atomData.torque[atom2]);
1157	+	}
1158	+
1159	+	if (storageLayout_ & DataStorage::dslDensity) {
1160	+	idat.rho1 = &(snap_->atomData.density[atom1]);
1161	+	idat.rho2 = &(snap_->atomData.density[atom2]);
1162	+	}
1163	+
1164	+	if (storageLayout_ & DataStorage::dslFunctional) {
1165	+	idat.frho1 = &(snap_->atomData.functional[atom1]);
1166	+	idat.frho2 = &(snap_->atomData.functional[atom2]);
1167	+	}
1168	+
1169	+	if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
1170	+	idat.dfrho1 = &(snap_->atomData.functionalDerivative[atom1]);
1171	+	idat.dfrho2 = &(snap_->atomData.functionalDerivative[atom2]);
1172	+	}
1173	+
1174	+	if (storageLayout_ & DataStorage::dslParticlePot) {
1175	+	idat.particlePot1 = &(snap_->atomData.particlePot[atom1]);
1176	+	idat.particlePot2 = &(snap_->atomData.particlePot[atom2]);
1177	+	}
1178	+
1179	+	if (storageLayout_ & DataStorage::dslSkippedCharge) {
1180	+	idat.skippedCharge1 = &(snap_->atomData.skippedCharge[atom1]);
1181	+	idat.skippedCharge2 = &(snap_->atomData.skippedCharge[atom2]);
1182	+	}
1183	+
1184	+	if (storageLayout_ & DataStorage::dslFlucQPosition) {
1185	+	idat.flucQ1 = &(snap_->atomData.flucQPos[atom1]);
1186	+	idat.flucQ2 = &(snap_->atomData.flucQPos[atom2]);
1187	+	}
1188	+
1189	+	#endif
1190	+	}
1191	+
1192	+
1193	+	void ForceMatrixDecomposition::unpackInteractionData(InteractionData &idat, int atom1, int atom2) {
1194	+	#ifdef IS_MPI
1195	+	pot_row[atom1] += RealType(0.5) *  *(idat.pot);
1196	+	pot_col[atom2] += RealType(0.5) *  *(idat.pot);
1197	+
1198	+	atomRowData.force[atom1] += *(idat.f1);
1199	+	atomColData.force[atom2] -= *(idat.f1);
1200	+
1201	+	if (storageLayout_ & DataStorage::dslFlucQForce) {
1202	+	atomRowData.flucQFrc[atom1] -= *(idat.dVdFQ1);
1203	+	atomColData.flucQFrc[atom2] -= *(idat.dVdFQ2);
1204	+	}
1205	+
1206	+	if (storageLayout_ & DataStorage::dslElectricField) {
1207	+	atomRowData.electricField[atom1] += *(idat.eField1);
1208	+	atomColData.electricField[atom2] += *(idat.eField2);
1209	+	}
1210	+
1211	+	#else
1212	+	pairwisePot += *(idat.pot);
1213	+
1214	+	snap_->atomData.force[atom1] += *(idat.f1);
1215	+	snap_->atomData.force[atom2] -= *(idat.f1);
1216	+
1217	+	if (idat.doParticlePot) {
1218	+	// This is the pairwise contribution to the particle pot.  The
1219	+	// embedding contribution is added in each of the low level
1220	+	// non-bonded routines.  In parallel, this calculation is done
1221	+	// in collectData, not in unpackInteractionData.
1222	+	snap_->atomData.particlePot[atom1] += *(idat.vpair) * *(idat.sw);
1223	+	snap_->atomData.particlePot[atom2] += *(idat.vpair) * *(idat.sw);
1224	+	}
1225	+
1226	+	if (storageLayout_ & DataStorage::dslFlucQForce) {
1227	+	snap_->atomData.flucQFrc[atom1] -= *(idat.dVdFQ1);
1228	+	snap_->atomData.flucQFrc[atom2] -= *(idat.dVdFQ2);
1229	+	}
1230	+
1231	+	if (storageLayout_ & DataStorage::dslElectricField) {
1232	+	snap_->atomData.electricField[atom1] += *(idat.eField1);
1233	+	snap_->atomData.electricField[atom2] += *(idat.eField2);
1234	+	}
1235	+
1236	+	#endif
1237	+
1238	+	}
1239	+
1240	+	/*
1241	+	* buildNeighborList
1242	+	*
1243	+	* first element of pair is row-indexed CutoffGroup
1244	+	* second element of pair is column-indexed CutoffGroup
1245	+	*/
1246	+	vector<pair<int, int> > ForceMatrixDecomposition::buildNeighborList() {
1247	+
1248	+	vector<pair<int, int> > neighborList;
1249	+	groupCutoffs cuts;
1250	+	bool doAllPairs = false;
1251	+
1252	+	#ifdef IS_MPI
1253	+	cellListRow_.clear();
1254	+	cellListCol_.clear();
1255	+	#else
1256	+	cellList_.clear();
1257	+	#endif
1258	+
1259	+	RealType rList_ = (largestRcut_ + skinThickness_);
1260	+	RealType rl2 = rList_ * rList_;
1261	+	Snapshot* snap_ = sman_->getCurrentSnapshot();
1262	+	Mat3x3d Hmat = snap_->getHmat();
1263	+	Vector3d Hx = Hmat.getColumn(0);
1264	+	Vector3d Hy = Hmat.getColumn(1);
1265	+	Vector3d Hz = Hmat.getColumn(2);
1266	+
1267	+	nCells_.x() = (int) ( Hx.length() )/ rList_;
1268	+	nCells_.y() = (int) ( Hy.length() )/ rList_;
1269	+	nCells_.z() = (int) ( Hz.length() )/ rList_;
1270	+
1271	+	// handle small boxes where the cell offsets can end up repeating cells
1272	+
1273	+	if (nCells_.x() < 3) doAllPairs = true;
1274	+	if (nCells_.y() < 3) doAllPairs = true;
1275	+	if (nCells_.z() < 3) doAllPairs = true;
1276	+
1277	+	Mat3x3d invHmat = snap_->getInvHmat();
1278	+	Vector3d rs, scaled, dr;
1279	+	Vector3i whichCell;
1280	+	int cellIndex;
1281	+	int nCtot = nCells_.x() * nCells_.y() * nCells_.z();
1282	+
1283	+	#ifdef IS_MPI
1284	+	cellListRow_.resize(nCtot);
1285	+	cellListCol_.resize(nCtot);
1286	+	#else
1287	+	cellList_.resize(nCtot);
1288	+	#endif
1289	+
1290	+	if (!doAllPairs) {
1291	+	#ifdef IS_MPI
1292	+
1293	+	for (int i = 0; i < nGroupsInRow_; i++) {
1294	+	rs = cgRowData.position[i];
1295	+
1296	+	// scaled positions relative to the box vectors
1297	+	scaled = invHmat * rs;
1298	+
1299	+	// wrap the vector back into the unit box by subtracting integer box
1300	+	// numbers
1301	+	for (int j = 0; j < 3; j++) {
1302	+	scaled[j] -= roundMe(scaled[j]);
1303	+	scaled[j] += 0.5;
1304	+	}
1305	+
1306	+	// find xyz-indices of cell that cutoffGroup is in.
1307	+	whichCell.x() = nCells_.x() * scaled.x();
1308	+	whichCell.y() = nCells_.y() * scaled.y();
1309	+	whichCell.z() = nCells_.z() * scaled.z();
1310	+
1311	+	// find single index of this cell:
1312	+	cellIndex = Vlinear(whichCell, nCells_);
1313	+
1314	+	// add this cutoff group to the list of groups in this cell;
1315	+	cellListRow_[cellIndex].push_back(i);
1316	+	}
1317	+	for (int i = 0; i < nGroupsInCol_; i++) {
1318	+	rs = cgColData.position[i];
1319	+
1320	+	// scaled positions relative to the box vectors
1321	+	scaled = invHmat * rs;
1322	+
1323	+	// wrap the vector back into the unit box by subtracting integer box
1324	+	// numbers
1325	+	for (int j = 0; j < 3; j++) {
1326	+	scaled[j] -= roundMe(scaled[j]);
1327	+	scaled[j] += 0.5;
1328	+	}
1329	+
1330	+	// find xyz-indices of cell that cutoffGroup is in.
1331	+	whichCell.x() = nCells_.x() * scaled.x();
1332	+	whichCell.y() = nCells_.y() * scaled.y();
1333	+	whichCell.z() = nCells_.z() * scaled.z();
1334	+
1335	+	// find single index of this cell:
1336	+	cellIndex = Vlinear(whichCell, nCells_);
1337	+
1338	+	// add this cutoff group to the list of groups in this cell;
1339	+	cellListCol_[cellIndex].push_back(i);
1340	+	}
1341	+
1342	+	#else
1343	+	for (int i = 0; i < nGroups_; i++) {
1344	+	rs = snap_->cgData.position[i];
1345	+
1346	+	// scaled positions relative to the box vectors
1347	+	scaled = invHmat * rs;
1348	+
1349	+	// wrap the vector back into the unit box by subtracting integer box
1350	+	// numbers
1351	+	for (int j = 0; j < 3; j++) {
1352	+	scaled[j] -= roundMe(scaled[j]);
1353	+	scaled[j] += 0.5;
1354	+	}
1355	+
1356	+	// find xyz-indices of cell that cutoffGroup is in.
1357	+	whichCell.x() = nCells_.x() * scaled.x();
1358	+	whichCell.y() = nCells_.y() * scaled.y();
1359	+	whichCell.z() = nCells_.z() * scaled.z();
1360	+
1361	+	// find single index of this cell:
1362	+	cellIndex = Vlinear(whichCell, nCells_);
1363	+
1364	+	// add this cutoff group to the list of groups in this cell;
1365	+	cellList_[cellIndex].push_back(i);
1366	+	}
1367	+
1368	+	#endif
1369	+
1370	+	for (int m1z = 0; m1z < nCells_.z(); m1z++) {
1371	+	for (int m1y = 0; m1y < nCells_.y(); m1y++) {
1372	+	for (int m1x = 0; m1x < nCells_.x(); m1x++) {
1373	+	Vector3i m1v(m1x, m1y, m1z);
1374	+	int m1 = Vlinear(m1v, nCells_);
1375	+
1376	+	for (vector<Vector3i>::iterator os = cellOffsets_.begin();
1377	+	os != cellOffsets_.end(); ++os) {
1378	+
1379	+	Vector3i m2v = m1v + (*os);
1380	+
1381	+
1382	+	if (m2v.x() >= nCells_.x()) {
1383	+	m2v.x() = 0;
1384	+	} else if (m2v.x() < 0) {
1385	+	m2v.x() = nCells_.x() - 1;
1386	+	}
1387	+
1388	+	if (m2v.y() >= nCells_.y()) {
1389	+	m2v.y() = 0;
1390	+	} else if (m2v.y() < 0) {
1391	+	m2v.y() = nCells_.y() - 1;
1392	+	}
1393	+
1394	+	if (m2v.z() >= nCells_.z()) {
1395	+	m2v.z() = 0;
1396	+	} else if (m2v.z() < 0) {
1397	+	m2v.z() = nCells_.z() - 1;
1398	+	}
1399	+
1400	+	int m2 = Vlinear (m2v, nCells_);
1401	+
1402	+	#ifdef IS_MPI
1403	+	for (vector<int>::iterator j1 = cellListRow_[m1].begin();
1404	+	j1 != cellListRow_[m1].end(); ++j1) {
1405	+	for (vector<int>::iterator j2 = cellListCol_[m2].begin();
1406	+	j2 != cellListCol_[m2].end(); ++j2) {
1407	+
1408	+	// In parallel, we need to visit *all* pairs of row
1409	+	// & column indicies and will divide labor in the
1410	+	// force evaluation later.
1411	+	dr = cgColData.position[(*j2)] - cgRowData.position[(*j1)];
1412	+	snap_->wrapVector(dr);
1413	+	cuts = getGroupCutoffs( (*j1), (*j2) );
1414	+	if (dr.lengthSquare() < cuts.third) {
1415	+	neighborList.push_back(make_pair((*j1), (*j2)));
1416	+	}
1417	+	}
1418	+	}
1419	+	#else
1420	+	for (vector<int>::iterator j1 = cellList_[m1].begin();
1421	+	j1 != cellList_[m1].end(); ++j1) {
1422	+	for (vector<int>::iterator j2 = cellList_[m2].begin();
1423	+	j2 != cellList_[m2].end(); ++j2) {
1424	+
1425	+	// Always do this if we're in different cells or if
1426	+	// we're in the same cell and the global index of
1427	+	// the j2 cutoff group is greater than or equal to
1428	+	// the j1 cutoff group.  Note that Rappaport's code
1429	+	// has a "less than" conditional here, but that
1430	+	// deals with atom-by-atom computation.  OpenMD
1431	+	// allows atoms within a single cutoff group to
1432	+	// interact with each other.
1433	+
1434	+
1435	+
1436	+	if (m2 != m1 || (*j2) >= (*j1) ) {
1437	+
1438	+	dr = snap_->cgData.position[(*j2)] - snap_->cgData.position[(*j1)];
1439	+	snap_->wrapVector(dr);
1440	+	cuts = getGroupCutoffs( (*j1), (*j2) );
1441	+	if (dr.lengthSquare() < cuts.third) {
1442	+	neighborList.push_back(make_pair((*j1), (*j2)));
1443	+	}
1444	+	}
1445	+	}
1446	+	}
1447	+	#endif
1448	+	}
1449	+	}
1450	+	}
1451	+	}
1452	+	} else {
1453	+	// branch to do all cutoff group pairs
1454	+	#ifdef IS_MPI
1455	+	for (int j1 = 0; j1 < nGroupsInRow_; j1++) {
1456	+	for (int j2 = 0; j2 < nGroupsInCol_; j2++) {
1457	+	dr = cgColData.position[j2] - cgRowData.position[j1];
1458	+	snap_->wrapVector(dr);
1459	+	cuts = getGroupCutoffs( j1, j2 );
1460	+	if (dr.lengthSquare() < cuts.third) {
1461	+	neighborList.push_back(make_pair(j1, j2));
1462	+	}
1463	+	}
1464	+	}
1465	+	#else
1466	+	// include all groups here.
1467	+	for (int j1 = 0; j1 < nGroups_; j1++) {
1468	+	// include self group interactions j2 == j1
1469	+	for (int j2 = j1; j2 < nGroups_; j2++) {
1470	+	dr = snap_->cgData.position[j2] - snap_->cgData.position[j1];
1471	+	snap_->wrapVector(dr);
1472	+	cuts = getGroupCutoffs( j1, j2 );
1473	+	if (dr.lengthSquare() < cuts.third) {
1474	+	neighborList.push_back(make_pair(j1, j2));
1475	+	}
1476	+	}
1477	+	}
1478	+	#endif
1479	+	}
1480	+
1481	+	// save the local cutoff group positions for the check that is
1482	+	// done on each loop:
1483	+	saved_CG_positions_.clear();
1484	+	for (int i = 0; i < nGroups_; i++)
1485	+	saved_CG_positions_.push_back(snap_->cgData.position[i]);
1486	+
1487	+	return neighborList;
1488	+	}
1489	+	} //end namespace OpenMD

Diff Legend

–	Removed lines
+	Added lines
<	Changed lines
>	Changed lines