ViewVC Help

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

root/OpenMD/branches/devel_omp/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/devel_omp/src/parallel/ForceMatrixDecomposition.cpp (file contents), Revision 1595 by chuckv, Tue Jul 19 18:50:04 2011 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 45 | Line 38
38		* [3]  Sun, Lin & Gezelter, J. Chem. Phys. 128, 24107 (2008).
39		* [4]  Vardeman & Gezelter, in progress (2009).
40		*/
41	+	#include "parallel/ForceMatrixDecomposition.hpp"
42	+	#include "math/SquareMatrix3.hpp"
43	+	#include "nonbonded/NonBondedInteraction.hpp"
44	+	#include "brains/SnapshotManager.hpp"
45	+	#include "brains/PairList.hpp"
46
47	+	using namespace std;
48	+	namespace OpenMD {
49
50	+	ForceMatrixDecomposition::ForceMatrixDecomposition(SimInfo* info, InteractionManager* iMan) : ForceDecomposition(info, iMan) {
51
52	<	/*  -*- c++ -*-  */
53	<	#include "config.h"
54	<	#include <stdlib.h>
52	>	// In a parallel computation, row and colum scans must visit all
53	>	// surrounding cells (not just the 14 upper triangular blocks that
54	>	// are used when the processor can see all pairs)
55		#ifdef IS_MPI
56	<	#include <mpi.h>
57	<	#endif
56	>	cellOffsets_.push_back( Vector3i(-1, 0, 0) );
57	>	cellOffsets_.push_back( Vector3i(-1,-1, 0) );
58	>	cellOffsets_.push_back( Vector3i( 0,-1, 0) );
59	>	cellOffsets_.push_back( Vector3i( 1,-1, 0) );
60	>	cellOffsets_.push_back( Vector3i( 0, 0,-1) );
61	>	cellOffsets_.push_back( Vector3i(-1, 0, 1) );
62	>	cellOffsets_.push_back( Vector3i(-1,-1,-1) );
63	>	cellOffsets_.push_back( Vector3i( 0,-1,-1) );
64	>	cellOffsets_.push_back( Vector3i( 1,-1,-1) );
65	>	cellOffsets_.push_back( Vector3i( 1, 0,-1) );
66	>	cellOffsets_.push_back( Vector3i( 1, 1,-1) );
67	>	cellOffsets_.push_back( Vector3i( 0, 1,-1) );
68	>	cellOffsets_.push_back( Vector3i(-1, 1,-1) );
69	>	#endif
70	>	}
71
58	–	#include <iostream>
59	–	#include <vector>
60	–	#include <algorithm>
61	–	#include <cmath>
62	–	#include "parallel/ForceDecomposition.hpp"
72
73	+	/**
74	+	* distributeInitialData is essentially a copy of the older fortran
75	+	* SimulationSetup
76	+	*/
77	+	void ForceMatrixDecomposition::distributeInitialData() {
78	+	snap_ = sman_->getCurrentSnapshot();
79	+	storageLayout_ = sman_->getStorageLayout();
80	+	ff_ = info_->getForceField();
81	+	nLocal_ = snap_->getNumberOfAtoms();
82	+
83	+	nGroups_ = info_->getNLocalCutoffGroups();
84	+	// gather the information for atomtype IDs (atids):
85	+	idents = info_->getIdentArray();
86	+	AtomLocalToGlobal = info_->getGlobalAtomIndices();
87	+	cgLocalToGlobal = info_->getGlobalGroupIndices();
88	+	vector<int> globalGroupMembership = info_->getGlobalGroupMembership();
89
90	<	using namespace std;
66	<	using namespace OpenMD;
90	>	massFactors = info_->getMassFactors();
91
92	<	//__static
92	>	PairList* excludes = info_->getExcludedInteractions();
93	>	PairList* oneTwo = info_->getOneTwoInteractions();
94	>	PairList* oneThree = info_->getOneThreeInteractions();
95	>	PairList* oneFour = info_->getOneFourInteractions();
96	>
97		#ifdef IS_MPI
98	<	static vector<MPI:Comm> communictors;
99	<	#endif
98	>
99	>	MPI::Intracomm row = rowComm.getComm();
100	>	MPI::Intracomm col = colComm.getComm();
101
102	<	//____ MPITypeTraits
103	<	template<typename T>
104	<	struct MPITypeTraits;
102	>	AtomPlanIntRow = new Plan<int>(row, nLocal_);
103	>	AtomPlanRealRow = new Plan<RealType>(row, nLocal_);
104	>	AtomPlanVectorRow = new Plan<Vector3d>(row, nLocal_);
105	>	AtomPlanMatrixRow = new Plan<Mat3x3d>(row, nLocal_);
106	>	AtomPlanPotRow = new Plan<potVec>(row, nLocal_);
107
108	<	#ifdef IS_MPI
109	<	template<>
110	<	struct MPITypeTraits<RealType> {
111	<	static const MPI::Datatype datatype;
112	<	};
82	<	const MPI_Datatype MPITypeTraits<RealType>::datatype = MY_MPI_REAL;
108	>	AtomPlanIntColumn = new Plan<int>(col, nLocal_);
109	>	AtomPlanRealColumn = new Plan<RealType>(col, nLocal_);
110	>	AtomPlanVectorColumn = new Plan<Vector3d>(col, nLocal_);
111	>	AtomPlanMatrixColumn = new Plan<Mat3x3d>(col, nLocal_);
112	>	AtomPlanPotColumn = new Plan<potVec>(col, nLocal_);
113
114	<	template<>
115	<	struct MPITypeTraits<int> {
116	<	static const MPI::Datatype datatype;
117	<	};
88	<	const MPI::Datatype MPITypeTraits<int>::datatype = MPI_INT;
89	<	#endif
114	>	cgPlanIntRow = new Plan<int>(row, nGroups_);
115	>	cgPlanVectorRow = new Plan<Vector3d>(row, nGroups_);
116	>	cgPlanIntColumn = new Plan<int>(col, nGroups_);
117	>	cgPlanVectorColumn = new Plan<Vector3d>(col, nGroups_);
118
119	<	/**
120	<	* Constructor for ForceDecomposition Parallel Decomposition Method
121	<	* Will try to construct a symmetric grid of processors. Ideally, the
122	<	* number of processors will be a square ex: 4, 9, 16, 25.
95	<	*
96	<	*/
119	>	nAtomsInRow_ = AtomPlanIntRow->getSize();
120	>	nAtomsInCol_ = AtomPlanIntColumn->getSize();
121	>	nGroupsInRow_ = cgPlanIntRow->getSize();
122	>	nGroupsInCol_ = cgPlanIntColumn->getSize();
123
124	<	ForceDecomposition::ForceDecomposition() {
124	>	// Modify the data storage objects with the correct layouts and sizes:
125	>	atomRowData.resize(nAtomsInRow_);
126	>	atomRowData.setStorageLayout(storageLayout_);
127	>	atomColData.resize(nAtomsInCol_);
128	>	atomColData.setStorageLayout(storageLayout_);
129	>	cgRowData.resize(nGroupsInRow_);
130	>	cgRowData.setStorageLayout(DataStorage::dslPosition);
131	>	cgColData.resize(nGroupsInCol_);
132	>	cgColData.setStorageLayout(DataStorage::dslPosition);
133	>
134	>	identsRow.resize(nAtomsInRow_);
135	>	identsCol.resize(nAtomsInCol_);
136	>
137	>	AtomPlanIntRow->gather(idents, identsRow);
138	>	AtomPlanIntColumn->gather(idents, identsCol);
139	>
140	>	// allocate memory for the parallel objects
141	>	atypesRow.resize(nAtomsInRow_);
142	>	atypesCol.resize(nAtomsInCol_);
143
144	<	#ifdef IS_MPI
145	<	int nProcs = MPI::COMM_WORLD.Get_size();
146	<	int worldRank = MPI::COMM_WORLD.Get_rank();
144	>	for (int i = 0; i < nAtomsInRow_; i++)
145	>	atypesRow[i] = ff_->getAtomType(identsRow[i]);
146	>	for (int i = 0; i < nAtomsInCol_; i++)
147	>	atypesCol[i] = ff_->getAtomType(identsCol[i]);
148	>
149	>	pot_row.resize(nAtomsInRow_);
150	>	pot_col.resize(nAtomsInCol_);
151	>
152	>	AtomRowToGlobal.resize(nAtomsInRow_);
153	>	AtomColToGlobal.resize(nAtomsInCol_);
154	>	AtomPlanIntRow->gather(AtomLocalToGlobal, AtomRowToGlobal);
155	>	AtomPlanIntColumn->gather(AtomLocalToGlobal, AtomColToGlobal);
156	>
157	>	cerr << "Atoms in Local:\n";
158	>	for (int i = 0; i < AtomLocalToGlobal.size(); i++) {
159	>	cerr << "i =\t" << i << "\t localAt =\t" << AtomLocalToGlobal[i] << "\n";
160	>	}
161	>	cerr << "Atoms in Row:\n";
162	>	for (int i = 0; i < AtomRowToGlobal.size(); i++) {
163	>	cerr << "i =\t" << i << "\t rowAt =\t" << AtomRowToGlobal[i] << "\n";
164	>	}
165	>	cerr << "Atoms in Col:\n";
166	>	for (int i = 0; i < AtomColToGlobal.size(); i++) {
167	>	cerr << "i =\t" << i << "\t colAt =\t" << AtomColToGlobal[i] << "\n";
168	>	}
169	>
170	>	cgRowToGlobal.resize(nGroupsInRow_);
171	>	cgColToGlobal.resize(nGroupsInCol_);
172	>	cgPlanIntRow->gather(cgLocalToGlobal, cgRowToGlobal);
173	>	cgPlanIntColumn->gather(cgLocalToGlobal, cgColToGlobal);
174	>
175	>	cerr << "Gruops in Local:\n";
176	>	for (int i = 0; i < cgLocalToGlobal.size(); i++) {
177	>	cerr << "i =\t" << i << "\t localCG =\t" << cgLocalToGlobal[i] << "\n";
178	>	}
179	>	cerr << "Groups in Row:\n";
180	>	for (int i = 0; i < cgRowToGlobal.size(); i++) {
181	>	cerr << "i =\t" << i << "\t rowCG =\t" << cgRowToGlobal[i] << "\n";
182	>	}
183	>	cerr << "Groups in Col:\n";
184	>	for (int i = 0; i < cgColToGlobal.size(); i++) {
185	>	cerr << "i =\t" << i << "\t colCG =\t" << cgColToGlobal[i] << "\n";
186	>	}
187	>
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		#endif
249
250	<	// First time through, construct column stride.
251	<	if (communicators.size() == 0)
252	<	{
253	<	int nColumnsMax = (int) round(sqrt((float) nProcs));
254	<	for (int i = 0; i < nProcs; ++i)
255	<	{
256	<	if (nProcs%i==0) nColumns=i;
250	>	// allocate memory for the parallel objects
251	>	atypesLocal.resize(nLocal_);
252	>
253	>	for (int i = 0; i < nLocal_; i++)
254	>	atypesLocal[i] = ff_->getAtomType(idents[i]);
255	>
256	>	groupList_.clear();
257	>	groupList_.resize(nGroups_);
258	>	for (int i = 0; i < nGroups_; i++) {
259	>	int gid = cgLocalToGlobal[i];
260	>	for (int j = 0; j < nLocal_; j++) {
261	>	int aid = AtomLocalToGlobal[j];
262	>	if (globalGroupMembership[aid] == gid) {
263	>	groupList_[i].push_back(j);
264	>	}
265	>	}
266		}
267
268	<	int nRows = nProcs/nColumns;
269	<	myRank_ = (int) worldRank%nColumns;
268	>	excludesForAtom.clear();
269	>	excludesForAtom.resize(nLocal_);
270	>	toposForAtom.clear();
271	>	toposForAtom.resize(nLocal_);
272	>	topoDist.clear();
273	>	topoDist.resize(nLocal_);
274	>
275	>	for (int i = 0; i < nLocal_; i++) {
276	>	int iglob = AtomLocalToGlobal[i];
277	>
278	>	for (int j = 0; j < nLocal_; j++) {
279	>	int jglob = AtomLocalToGlobal[j];
280	>
281	>	if (excludes->hasPair(iglob, jglob))
282	>	excludesForAtom[i].push_back(j);
283	>
284	>	if (oneTwo->hasPair(iglob, jglob)) {
285	>	toposForAtom[i].push_back(j);
286	>	topoDist[i].push_back(1);
287	>	} else {
288	>	if (oneThree->hasPair(iglob, jglob)) {
289	>	toposForAtom[i].push_back(j);
290	>	topoDist[i].push_back(2);
291	>	} else {
292	>	if (oneFour->hasPair(iglob, jglob)) {
293	>	toposForAtom[i].push_back(j);
294	>	topoDist[i].push_back(3);
295	>	}
296	>	}
297	>	}
298	>	}
299	>	}
300	>
301	>	createGtypeCutoffMap();
302	>
303		}
304	<	else
305	<	{
306	<	myRank_ = myRank/nColumns;
307	<	}
308	<	MPI::Comm newComm = MPI:COMM_WORLD.Split(myRank_,0);
309	<
310	<	isColumn_ = false;
311	<
312	<	}
304	>
305	>	void ForceMatrixDecomposition::createGtypeCutoffMap() {
306	>
307	>	RealType tol = 1e-6;
308	>	largestRcut_ = 0.0;
309	>	RealType rc;
310	>	int atid;
311	>	set<AtomType*> atypes = info_->getSimulatedAtomTypes();
312	>
313	>	map<int, RealType> atypeCutoff;
314	>
315	>	for (set<AtomType*>::iterator at = atypes.begin();
316	>	at != atypes.end(); ++at){
317	>	atid = (*at)->getIdent();
318	>	if (userChoseCutoff_)
319	>	atypeCutoff[atid] = userCutoff_;
320	>	else
321	>	atypeCutoff[atid] = interactionMan_->getSuggestedCutoffRadius(*at);
322	>	}
323	>
324	>	vector<RealType> gTypeCutoffs;
325	>	// first we do a single loop over the cutoff groups to find the
326	>	// largest cutoff for any atypes present in this group.
327	>	#ifdef IS_MPI
328	>	vector<RealType> groupCutoffRow(nGroupsInRow_, 0.0);
329	>	groupRowToGtype.resize(nGroupsInRow_);
330	>	for (int cg1 = 0; cg1 < nGroupsInRow_; cg1++) {
331	>	vector<int> atomListRow = getAtomsInGroupRow(cg1);
332	>	for (vector<int>::iterator ia = atomListRow.begin();
333	>	ia != atomListRow.end(); ++ia) {
334	>	int atom1 = (*ia);
335	>	atid = identsRow[atom1];
336	>	if (atypeCutoff[atid] > groupCutoffRow[cg1]) {
337	>	groupCutoffRow[cg1] = atypeCutoff[atid];
338	>	}
339	>	}
340
341	<	ForceDecomposition::gather(sendbuf, receivebuf){
342	<	communicators(myIndex_).Allgatherv();
343	<	}
341	>	bool gTypeFound = false;
342	>	for (int gt = 0; gt < gTypeCutoffs.size(); gt++) {
343	>	if (abs(groupCutoffRow[cg1] - gTypeCutoffs[gt]) < tol) {
344	>	groupRowToGtype[cg1] = gt;
345	>	gTypeFound = true;
346	>	}
347	>	}
348	>	if (!gTypeFound) {
349	>	gTypeCutoffs.push_back( groupCutoffRow[cg1] );
350	>	groupRowToGtype[cg1] = gTypeCutoffs.size() - 1;
351	>	}
352	>
353	>	}
354	>	vector<RealType> groupCutoffCol(nGroupsInCol_, 0.0);
355	>	groupColToGtype.resize(nGroupsInCol_);
356	>	for (int cg2 = 0; cg2 < nGroupsInCol_; cg2++) {
357	>	vector<int> atomListCol = getAtomsInGroupColumn(cg2);
358	>	for (vector<int>::iterator jb = atomListCol.begin();
359	>	jb != atomListCol.end(); ++jb) {
360	>	int atom2 = (*jb);
361	>	atid = identsCol[atom2];
362	>	if (atypeCutoff[atid] > groupCutoffCol[cg2]) {
363	>	groupCutoffCol[cg2] = atypeCutoff[atid];
364	>	}
365	>	}
366	>	bool gTypeFound = false;
367	>	for (int gt = 0; gt < gTypeCutoffs.size(); gt++) {
368	>	if (abs(groupCutoffCol[cg2] - gTypeCutoffs[gt]) < tol) {
369	>	groupColToGtype[cg2] = gt;
370	>	gTypeFound = true;
371	>	}
372	>	}
373	>	if (!gTypeFound) {
374	>	gTypeCutoffs.push_back( groupCutoffCol[cg2] );
375	>	groupColToGtype[cg2] = gTypeCutoffs.size() - 1;
376	>	}
377	>	}
378	>	#else
379
380	+	vector<RealType> groupCutoff(nGroups_, 0.0);
381	+	groupToGtype.resize(nGroups_);
382	+	for (int cg1 = 0; cg1 < nGroups_; cg1++) {
383	+	groupCutoff[cg1] = 0.0;
384	+	vector<int> atomList = getAtomsInGroupRow(cg1);
385	+	for (vector<int>::iterator ia = atomList.begin();
386	+	ia != atomList.end(); ++ia) {
387	+	int atom1 = (*ia);
388	+	atid = idents[atom1];
389	+	if (atypeCutoff[atid] > groupCutoff[cg1])
390	+	groupCutoff[cg1] = atypeCutoff[atid];
391	+	}
392	+
393	+	bool gTypeFound = false;
394	+	for (int gt = 0; gt < gTypeCutoffs.size(); gt++) {
395	+	if (abs(groupCutoff[cg1] - gTypeCutoffs[gt]) < tol) {
396	+	groupToGtype[cg1] = gt;
397	+	gTypeFound = true;
398	+	}
399	+	}
400	+	if (!gTypeFound) {
401	+	gTypeCutoffs.push_back( groupCutoff[cg1] );
402	+	groupToGtype[cg1] = gTypeCutoffs.size() - 1;
403	+	}
404	+	}
405	+	#endif
406
407	+	// Now we find the maximum group cutoff value present in the simulation
408
409	<	ForceDecomposition::scatter(sbuffer, rbuffer){
410	<	communicators(myIndex_).Reduce_scatter(sbuffer, recevbuf. recvcounts, MPI::DOUBLE, MPI::SUM);
409	>	RealType groupMax = *max_element(gTypeCutoffs.begin(),
410	>	gTypeCutoffs.end());
411	>
412	>	#ifdef IS_MPI
413	>	MPI::COMM_WORLD.Allreduce(&groupMax, &groupMax, 1, MPI::REALTYPE,
414	>	MPI::MAX);
415	>	#endif
416	>
417	>	RealType tradRcut = groupMax;
418	>
419	>	for (int i = 0; i < gTypeCutoffs.size();  i++) {
420	>	for (int j = 0; j < gTypeCutoffs.size();  j++) {
421	>	RealType thisRcut;
422	>	switch(cutoffPolicy_) {
423	>	case TRADITIONAL:
424	>	thisRcut = tradRcut;
425	>	break;
426	>	case MIX:
427	>	thisRcut = 0.5 * (gTypeCutoffs[i] + gTypeCutoffs[j]);
428	>	break;
429	>	case MAX:
430	>	thisRcut = max(gTypeCutoffs[i], gTypeCutoffs[j]);
431	>	break;
432	>	default:
433	>	sprintf(painCave.errMsg,
434	>	"ForceMatrixDecomposition::createGtypeCutoffMap "
435	>	"hit an unknown cutoff policy!\n");
436	>	painCave.severity = OPENMD_ERROR;
437	>	painCave.isFatal = 1;
438	>	simError();
439	>	break;
440	>	}
441	>
442	>	pair<int,int> key = make_pair(i,j);
443	>	gTypeCutoffMap[key].first = thisRcut;
444	>	if (thisRcut > largestRcut_) largestRcut_ = thisRcut;
445	>	gTypeCutoffMap[key].second = thisRcut*thisRcut;
446	>	gTypeCutoffMap[key].third = pow(thisRcut + skinThickness_, 2);
447	>	// sanity check
448	>
449	>	if (userChoseCutoff_) {
450	>	if (abs(gTypeCutoffMap[key].first - userCutoff_) > 0.0001) {
451	>	sprintf(painCave.errMsg,
452	>	"ForceMatrixDecomposition::createGtypeCutoffMap "
453	>	"user-specified rCut (%lf) does not match computed group Cutoff\n", userCutoff_);
454	>	painCave.severity = OPENMD_ERROR;
455	>	painCave.isFatal = 1;
456	>	simError();
457	>	}
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	>	#endif
539	>	// even in parallel, we need to zero out the local arrays:
540	>
541	>	if (storageLayout_ & DataStorage::dslParticlePot) {
542	>	fill(snap_->atomData.particlePot.begin(),
543	>	snap_->atomData.particlePot.end(), 0.0);
544	>	}
545	>
546	>	if (storageLayout_ & DataStorage::dslDensity) {
547	>	fill(snap_->atomData.density.begin(),
548	>	snap_->atomData.density.end(), 0.0);
549	>	}
550	>	if (storageLayout_ & DataStorage::dslFunctional) {
551	>	fill(snap_->atomData.functional.begin(),
552	>	snap_->atomData.functional.end(), 0.0);
553	>	}
554	>	if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
555	>	fill(snap_->atomData.functionalDerivative.begin(),
556	>	snap_->atomData.functionalDerivative.end(), 0.0);
557	>	}
558	>	if (storageLayout_ & DataStorage::dslSkippedCharge) {
559	>	fill(snap_->atomData.skippedCharge.begin(),
560	>	snap_->atomData.skippedCharge.end(), 0.0);
561	>	}
562	>
563	>	}
564	>
565	>
566	>	void ForceMatrixDecomposition::distributeData()  {
567	>	snap_ = sman_->getCurrentSnapshot();
568	>	storageLayout_ = sman_->getStorageLayout();
569	>	#ifdef IS_MPI
570	>
571	>	// gather up the atomic positions
572	>	AtomPlanVectorRow->gather(snap_->atomData.position,
573	>	atomRowData.position);
574	>	AtomPlanVectorColumn->gather(snap_->atomData.position,
575	>	atomColData.position);
576	>
577	>	// gather up the cutoff group positions
578	>
579	>	cerr  << "before gather\n";
580	>	for (int i = 0; i < snap_->cgData.position.size(); i++) {
581	>	cerr << "cgpos = " << snap_->cgData.position[i] << "\n";
582	>	}
583	>
584	>	cgPlanVectorRow->gather(snap_->cgData.position,
585	>	cgRowData.position);
586	>
587	>	cerr  << "after gather\n";
588	>	for (int i = 0; i < cgRowData.position.size(); i++) {
589	>	cerr << "cgRpos = " << cgRowData.position[i] << "\n";
590	>	}
591	>
592	>	cgPlanVectorColumn->gather(snap_->cgData.position,
593	>	cgColData.position);
594	>	for (int i = 0; i < cgColData.position.size(); i++) {
595	>	cerr << "cgCpos = " << cgColData.position[i] << "\n";
596	>	}
597	>
598	>
599	>	// if needed, gather the atomic rotation matrices
600	>	if (storageLayout_ & DataStorage::dslAmat) {
601	>	AtomPlanMatrixRow->gather(snap_->atomData.aMat,
602	>	atomRowData.aMat);
603	>	AtomPlanMatrixColumn->gather(snap_->atomData.aMat,
604	>	atomColData.aMat);
605	>	}
606	>
607	>	// if needed, gather the atomic eletrostatic frames
608	>	if (storageLayout_ & DataStorage::dslElectroFrame) {
609	>	AtomPlanMatrixRow->gather(snap_->atomData.electroFrame,
610	>	atomRowData.electroFrame);
611	>	AtomPlanMatrixColumn->gather(snap_->atomData.electroFrame,
612	>	atomColData.electroFrame);
613	>	}
614	>
615	>	#endif
616	>	}
617	>
618	>	/* collects information obtained during the pre-pair loop onto local
619	>	* data structures.
620	>	*/
621	>	void ForceMatrixDecomposition::collectIntermediateData() {
622	>	snap_ = sman_->getCurrentSnapshot();
623	>	storageLayout_ = sman_->getStorageLayout();
624	>	#ifdef IS_MPI
625	>
626	>	if (storageLayout_ & DataStorage::dslDensity) {
627	>
628	>	AtomPlanRealRow->scatter(atomRowData.density,
629	>	snap_->atomData.density);
630	>
631	>	int n = snap_->atomData.density.size();
632	>	vector<RealType> rho_tmp(n, 0.0);
633	>	AtomPlanRealColumn->scatter(atomColData.density, rho_tmp);
634	>	for (int i = 0; i < n; i++)
635	>	snap_->atomData.density[i] += rho_tmp[i];
636	>	}
637	>	#endif
638	>	}
639	>
640	>	/*
641	>	* redistributes information obtained during the pre-pair loop out to
642	>	* row and column-indexed data structures
643	>	*/
644	>	void ForceMatrixDecomposition::distributeIntermediateData() {
645	>	snap_ = sman_->getCurrentSnapshot();
646	>	storageLayout_ = sman_->getStorageLayout();
647	>	#ifdef IS_MPI
648	>	if (storageLayout_ & DataStorage::dslFunctional) {
649	>	AtomPlanRealRow->gather(snap_->atomData.functional,
650	>	atomRowData.functional);
651	>	AtomPlanRealColumn->gather(snap_->atomData.functional,
652	>	atomColData.functional);
653	>	}
654	>
655	>	if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
656	>	AtomPlanRealRow->gather(snap_->atomData.functionalDerivative,
657	>	atomRowData.functionalDerivative);
658	>	AtomPlanRealColumn->gather(snap_->atomData.functionalDerivative,
659	>	atomColData.functionalDerivative);
660	>	}
661	>	#endif
662	>	}
663	>
664	>
665	>	void ForceMatrixDecomposition::collectData() {
666	>	snap_ = sman_->getCurrentSnapshot();
667	>	storageLayout_ = sman_->getStorageLayout();
668	>	#ifdef IS_MPI
669	>	int n = snap_->atomData.force.size();
670	>	vector<Vector3d> frc_tmp(n, V3Zero);
671	>
672	>	AtomPlanVectorRow->scatter(atomRowData.force, frc_tmp);
673	>	for (int i = 0; i < n; i++) {
674	>	snap_->atomData.force[i] += frc_tmp[i];
675	>	frc_tmp[i] = 0.0;
676	>	}
677	>
678	>	AtomPlanVectorColumn->scatter(atomColData.force, frc_tmp);
679	>	for (int i = 0; i < n; i++) {
680	>	snap_->atomData.force[i] += frc_tmp[i];
681	>	}
682	>
683	>	if (storageLayout_ & DataStorage::dslTorque) {
684	>
685	>	int nt = snap_->atomData.torque.size();
686	>	vector<Vector3d> trq_tmp(nt, V3Zero);
687	>
688	>	AtomPlanVectorRow->scatter(atomRowData.torque, trq_tmp);
689	>	for (int i = 0; i < nt; i++) {
690	>	snap_->atomData.torque[i] += trq_tmp[i];
691	>	trq_tmp[i] = 0.0;
692	>	}
693	>
694	>	AtomPlanVectorColumn->scatter(atomColData.torque, trq_tmp);
695	>	for (int i = 0; i < nt; i++)
696	>	snap_->atomData.torque[i] += trq_tmp[i];
697	>	}
698	>
699	>	if (storageLayout_ & DataStorage::dslSkippedCharge) {
700	>
701	>	int ns = snap_->atomData.skippedCharge.size();
702	>	vector<RealType> skch_tmp(ns, 0.0);
703	>
704	>	AtomPlanRealRow->scatter(atomRowData.skippedCharge, skch_tmp);
705	>	for (int i = 0; i < ns; i++) {
706	>	snap_->atomData.skippedCharge[i] += skch_tmp[i];
707	>	skch_tmp[i] = 0.0;
708	>	}
709	>
710	>	AtomPlanRealColumn->scatter(atomColData.skippedCharge, skch_tmp);
711	>	for (int i = 0; i < ns; i++)
712	>	snap_->atomData.skippedCharge[i] += skch_tmp[i];
713	>	}
714	>
715	>	nLocal_ = snap_->getNumberOfAtoms();
716	>
717	>	vector<potVec> pot_temp(nLocal_,
718	>	Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
719	>
720	>	// scatter/gather pot_row into the members of my column
721	>
722	>	AtomPlanPotRow->scatter(pot_row, pot_temp);
723	>
724	>	for (int ii = 0;  ii < pot_temp.size(); ii++ )
725	>	pairwisePot += pot_temp[ii];
726	>
727	>	fill(pot_temp.begin(), pot_temp.end(),
728	>	Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
729	>
730	>	AtomPlanPotColumn->scatter(pot_col, pot_temp);
731	>
732	>	for (int ii = 0;  ii < pot_temp.size(); ii++ )
733	>	pairwisePot += pot_temp[ii];
734	>	#endif
735	>
736	>	cerr << "pairwisePot = " <<  pairwisePot << "\n";
737	>	}
738	>
739	>	int ForceMatrixDecomposition::getNAtomsInRow() {
740	>	#ifdef IS_MPI
741	>	return nAtomsInRow_;
742	>	#else
743	>	return nLocal_;
744	>	#endif
745	>	}
746	>
747	>	/**
748	>	* returns the list of atoms belonging to this group.
749	>	*/
750	>	vector<int> ForceMatrixDecomposition::getAtomsInGroupRow(int cg1){
751	>	#ifdef IS_MPI
752	>	return groupListRow_[cg1];
753	>	#else
754	>	return groupList_[cg1];
755	>	#endif
756	>	}
757	>
758	>	vector<int> ForceMatrixDecomposition::getAtomsInGroupColumn(int cg2){
759	>	#ifdef IS_MPI
760	>	return groupListCol_[cg2];
761	>	#else
762	>	return groupList_[cg2];
763	>	#endif
764	>	}
765	>
766	>	Vector3d ForceMatrixDecomposition::getIntergroupVector(int cg1, int cg2){
767	>	Vector3d d;
768	>
769	>	#ifdef IS_MPI
770	>	d = cgColData.position[cg2] - cgRowData.position[cg1];
771	>	cerr << "cg1 = " << cg1 << "\tcg1p = " << cgRowData.position[cg1] << "\n";
772	>	cerr << "cg2 = " << cg2 << "\tcg2p = " << cgColData.position[cg2] << "\n";
773	>	#else
774	>	d = snap_->cgData.position[cg2] - snap_->cgData.position[cg1];
775	>	cerr << "cg1 = " << cg1 << "\tcg1p = " << snap_->cgData.position[cg1] << "\n";
776	>	cerr << "cg2 = " << cg2 << "\tcg2p = " << snap_->cgData.position[cg2] << "\n";
777	>	#endif
778	>
779	>	snap_->wrapVector(d);
780	>	return d;
781	>	}
782	>
783	>
784	>	Vector3d ForceMatrixDecomposition::getAtomToGroupVectorRow(int atom1, int cg1){
785	>
786	>	Vector3d d;
787	>
788	>	#ifdef IS_MPI
789	>	d = cgRowData.position[cg1] - atomRowData.position[atom1];
790	>	#else
791	>	d = snap_->cgData.position[cg1] - snap_->atomData.position[atom1];
792	>	#endif
793	>
794	>	snap_->wrapVector(d);
795	>	return d;
796	>	}
797	>
798	>	Vector3d ForceMatrixDecomposition::getAtomToGroupVectorColumn(int atom2, int cg2){
799	>	Vector3d d;
800	>
801	>	#ifdef IS_MPI
802	>	d = cgColData.position[cg2] - atomColData.position[atom2];
803	>	#else
804	>	d = snap_->cgData.position[cg2] - snap_->atomData.position[atom2];
805	>	#endif
806	>
807	>	snap_->wrapVector(d);
808	>	return d;
809	>	}
810	>
811	>	RealType ForceMatrixDecomposition::getMassFactorRow(int atom1) {
812	>	#ifdef IS_MPI
813	>	return massFactorsRow[atom1];
814	>	#else
815	>	return massFactors[atom1];
816	>	#endif
817	>	}
818	>
819	>	RealType ForceMatrixDecomposition::getMassFactorColumn(int atom2) {
820	>	#ifdef IS_MPI
821	>	return massFactorsCol[atom2];
822	>	#else
823	>	return massFactors[atom2];
824	>	#endif
825	>
826	>	}
827	>
828	>	Vector3d ForceMatrixDecomposition::getInteratomicVector(int atom1, int atom2){
829	>	Vector3d d;
830	>
831	>	#ifdef IS_MPI
832	>	d = atomColData.position[atom2] - atomRowData.position[atom1];
833	>	#else
834	>	d = snap_->atomData.position[atom2] - snap_->atomData.position[atom1];
835	>	#endif
836	>
837	>	snap_->wrapVector(d);
838	>	return d;
839	>	}
840	>
841	>	vector<int> ForceMatrixDecomposition::getExcludesForAtom(int atom1) {
842	>	return excludesForAtom[atom1];
843	>	}
844	>
845	>	/**
846	>	* We need to exclude some overcounted interactions that result from
847	>	* the parallel decomposition.
848	>	*/
849	>	bool ForceMatrixDecomposition::skipAtomPair(int atom1, int atom2) {
850	>	int unique_id_1, unique_id_2;
851	>
852	>
853	>	cerr << "sap with atom1, atom2 =\t" << atom1 << "\t" << atom2 << "\n";
854	>	#ifdef IS_MPI
855	>	// in MPI, we have to look up the unique IDs for each atom
856	>	unique_id_1 = AtomRowToGlobal[atom1];
857	>	unique_id_2 = AtomColToGlobal[atom2];
858	>
859	>	cerr << "sap with uid1, uid2 =\t" << unique_id_1 << "\t" << unique_id_2 << "\n";
860	>	// this situation should only arise in MPI simulations
861	>	if (unique_id_1 == unique_id_2) return true;
862	>
863	>	// this prevents us from doing the pair on multiple processors
864	>	if (unique_id_1 < unique_id_2) {
865	>	if ((unique_id_1 + unique_id_2) % 2 == 0) return true;
866	>	} else {
867	>	if ((unique_id_1 + unique_id_2) % 2 == 1) return true;
868	>	}
869	>	#endif
870	>	return false;
871	>	}
872	>
873	>	/**
874	>	* We need to handle the interactions for atoms who are involved in
875	>	* the same rigid body as well as some short range interactions
876	>	* (bonds, bends, torsions) differently from other interactions.
877	>	* We'll still visit the pairwise routines, but with a flag that
878	>	* tells those routines to exclude the pair from direct long range
879	>	* interactions.  Some indirect interactions (notably reaction
880	>	* field) must still be handled for these pairs.
881	>	*/
882	>	bool ForceMatrixDecomposition::excludeAtomPair(int atom1, int atom2) {
883	>	int unique_id_2;
884	>	#ifdef IS_MPI
885	>	// in MPI, we have to look up the unique IDs for the row atom.
886	>	unique_id_2 = AtomColToGlobal[atom2];
887	>	#else
888	>	// in the normal loop, the atom numbers are unique
889	>	unique_id_2 = atom2;
890	>	#endif
891	>
892	>	for (vector<int>::iterator i = excludesForAtom[atom1].begin();
893	>	i != excludesForAtom[atom1].end(); ++i) {
894	>	if ( (*i) == unique_id_2 ) return true;
895	>	}
896	>
897	>	return false;
898	>	}
899	>
900	>
901	>	void ForceMatrixDecomposition::addForceToAtomRow(int atom1, Vector3d fg){
902	>	#ifdef IS_MPI
903	>	atomRowData.force[atom1] += fg;
904	>	#else
905	>	snap_->atomData.force[atom1] += fg;
906	>	#endif
907	>	}
908	>
909	>	void ForceMatrixDecomposition::addForceToAtomColumn(int atom2, Vector3d fg){
910	>	#ifdef IS_MPI
911	>	atomColData.force[atom2] += fg;
912	>	#else
913	>	snap_->atomData.force[atom2] += fg;
914	>	#endif
915	>	}
916	>
917	>	// filling interaction blocks with pointers
918	>	void ForceMatrixDecomposition::fillInteractionData(InteractionData &idat,
919	>	int atom1, int atom2) {
920	>
921	>	idat.excluded = excludeAtomPair(atom1, atom2);
922	>
923	>	#ifdef IS_MPI
924	>	idat.atypes = make_pair( atypesRow[atom1], atypesCol[atom2]);
925	>	//idat.atypes = make_pair( ff_->getAtomType(identsRow[atom1]),
926	>	//                         ff_->getAtomType(identsCol[atom2]) );
927	>
928	>	if (storageLayout_ & DataStorage::dslAmat) {
929	>	idat.A1 = &(atomRowData.aMat[atom1]);
930	>	idat.A2 = &(atomColData.aMat[atom2]);
931	>	}
932	>
933	>	if (storageLayout_ & DataStorage::dslElectroFrame) {
934	>	idat.eFrame1 = &(atomRowData.electroFrame[atom1]);
935	>	idat.eFrame2 = &(atomColData.electroFrame[atom2]);
936	>	}
937	>
938	>	if (storageLayout_ & DataStorage::dslTorque) {
939	>	idat.t1 = &(atomRowData.torque[atom1]);
940	>	idat.t2 = &(atomColData.torque[atom2]);
941	>	}
942	>
943	>	if (storageLayout_ & DataStorage::dslDensity) {
944	>	idat.rho1 = &(atomRowData.density[atom1]);
945	>	idat.rho2 = &(atomColData.density[atom2]);
946	>	}
947	>
948	>	if (storageLayout_ & DataStorage::dslFunctional) {
949	>	idat.frho1 = &(atomRowData.functional[atom1]);
950	>	idat.frho2 = &(atomColData.functional[atom2]);
951	>	}
952	>
953	>	if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
954	>	idat.dfrho1 = &(atomRowData.functionalDerivative[atom1]);
955	>	idat.dfrho2 = &(atomColData.functionalDerivative[atom2]);
956	>	}
957	>
958	>	if (storageLayout_ & DataStorage::dslParticlePot) {
959	>	idat.particlePot1 = &(atomRowData.particlePot[atom1]);
960	>	idat.particlePot2 = &(atomColData.particlePot[atom2]);
961	>	}
962	>
963	>	if (storageLayout_ & DataStorage::dslSkippedCharge) {
964	>	idat.skippedCharge1 = &(atomRowData.skippedCharge[atom1]);
965	>	idat.skippedCharge2 = &(atomColData.skippedCharge[atom2]);
966	>	}
967	>
968	>	#else
969	>
970	>	idat.atypes = make_pair( atypesLocal[atom1], atypesLocal[atom2]);
971	>	//idat.atypes = make_pair( ff_->getAtomType(idents[atom1]),
972	>	//                         ff_->getAtomType(idents[atom2]) );
973	>
974	>	if (storageLayout_ & DataStorage::dslAmat) {
975	>	idat.A1 = &(snap_->atomData.aMat[atom1]);
976	>	idat.A2 = &(snap_->atomData.aMat[atom2]);
977	>	}
978	>
979	>	if (storageLayout_ & DataStorage::dslElectroFrame) {
980	>	idat.eFrame1 = &(snap_->atomData.electroFrame[atom1]);
981	>	idat.eFrame2 = &(snap_->atomData.electroFrame[atom2]);
982	>	}
983	>
984	>	if (storageLayout_ & DataStorage::dslTorque) {
985	>	idat.t1 = &(snap_->atomData.torque[atom1]);
986	>	idat.t2 = &(snap_->atomData.torque[atom2]);
987	>	}
988	>
989	>	if (storageLayout_ & DataStorage::dslDensity) {
990	>	idat.rho1 = &(snap_->atomData.density[atom1]);
991	>	idat.rho2 = &(snap_->atomData.density[atom2]);
992	>	}
993	>
994	>	if (storageLayout_ & DataStorage::dslFunctional) {
995	>	idat.frho1 = &(snap_->atomData.functional[atom1]);
996	>	idat.frho2 = &(snap_->atomData.functional[atom2]);
997	>	}
998	>
999	>	if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
1000	>	idat.dfrho1 = &(snap_->atomData.functionalDerivative[atom1]);
1001	>	idat.dfrho2 = &(snap_->atomData.functionalDerivative[atom2]);
1002	>	}
1003	>
1004	>	if (storageLayout_ & DataStorage::dslParticlePot) {
1005	>	idat.particlePot1 = &(snap_->atomData.particlePot[atom1]);
1006	>	idat.particlePot2 = &(snap_->atomData.particlePot[atom2]);
1007	>	}
1008	>
1009	>	if (storageLayout_ & DataStorage::dslSkippedCharge) {
1010	>	idat.skippedCharge1 = &(snap_->atomData.skippedCharge[atom1]);
1011	>	idat.skippedCharge2 = &(snap_->atomData.skippedCharge[atom2]);
1012	>	}
1013	>	#endif
1014	>	}
1015	>
1016	>
1017	>	void ForceMatrixDecomposition::unpackInteractionData(InteractionData &idat, int atom1, int atom2) {
1018	>	#ifdef IS_MPI
1019	>	pot_row[atom1] += 0.5 *  *(idat.pot);
1020	>	pot_col[atom2] += 0.5 *  *(idat.pot);
1021	>
1022	>	atomRowData.force[atom1] += *(idat.f1);
1023	>	atomColData.force[atom2] -= *(idat.f1);
1024	>	#else
1025	>	pairwisePot += *(idat.pot);
1026	>
1027	>	snap_->atomData.force[atom1] += *(idat.f1);
1028	>	snap_->atomData.force[atom2] -= *(idat.f1);
1029	>	#endif
1030	>
1031	>	}
1032	>
1033	>	vector<vector<int> > ForceMatrixDecomposition::buildLayerBasedNeighborList() {
1034	>	printf("buildLayerBasedNeighborList; nGroups:%d\n", nGroups_);
1035	>	// Na = nGroups_
1036	>	/* cell occupancy counter */
1037	>	vector<int> k_c;
1038	>	/* c_i - has cell containing atom i (size Na) */
1039	>	vector<int> c;
1040	>	/* l_i - layer containing atom i (size Na) */
1041	>	vector<int> l;
1042	>
1043	>	//      cellList_.clear();
1044	>
1045	>	RealType rList_ = (largestRcut_ + skinThickness_);
1046	>	Snapshot* snap_ = sman_->getCurrentSnapshot();
1047	>	Mat3x3d Hmat = snap_->getHmat();
1048	>	Vector3d Hx = Hmat.getColumn(0);
1049	>	Vector3d Hy = Hmat.getColumn(1);
1050	>	Vector3d Hz = Hmat.getColumn(2);
1051	>
1052	>	nCells_.x() = (int) (Hx.length()) / rList_;
1053	>	nCells_.y() = (int) (Hy.length()) / rList_;
1054	>	nCells_.z() = (int) (Hz.length()) / rList_;
1055	>
1056	>	Mat3x3d invHmat = snap_->getInvHmat();
1057	>	Vector3d rs, scaled, dr;
1058	>	Vector3i whichCell;
1059	>	int cellIndex;
1060	>	int nCtot = nCells_.x() * nCells_.y() * nCells_.z();
1061	>
1062	>	//      cellList_.resize(nCtot);
1063	>	k_c = vector<int>(nCtot, 0);
1064	>
1065	>	for (int i = 0; i < nGroups_; i++)
1066	>	{
1067	>	rs = snap_->cgData.position[i];
1068	>
1069	>	// scaled positions relative to the box vectors
1070	>	scaled = invHmat * rs;
1071	>
1072	>	// wrap the vector back into the unit box by subtracting integer box
1073	>	// numbers
1074	>	for (int j = 0; j < 3; j++)
1075	>	{
1076	>	scaled[j] -= roundMe(scaled[j]);
1077	>	scaled[j] += 0.5;
1078	>	}
1079	>
1080	>	// find xyz-indices of cell that cutoffGroup is in.
1081	>	whichCell.x() = nCells_.x() * scaled.x();
1082	>	whichCell.y() = nCells_.y() * scaled.y();
1083	>	whichCell.z() = nCells_.z() * scaled.z();
1084	>
1085	>	// find single index of this cell:
1086	>	cellIndex = Vlinear(whichCell, nCells_);
1087	>
1088	>	c.push_back(cellIndex);
1089	>
1090	>	//              // add this cutoff group to the list of groups in this cell;
1091	>	//              cellList_[cellIndex].push_back(i);
1092	>	}
1093	>
1094	>	int k_c_curr;
1095	>	int k_c_max = 0;
1096	>	/* the cell-layer occupancy matrix */
1097	>	vector<vector<int> > H_c_l = vector<vector<int> >(nCtot);
1098	>
1099	>	for(int i = 0; i < nGroups_; ++i)
1100	>	{
1101	>	k_c_curr = ++k_c[c[i]];
1102	>	l.push_back(k_c_curr);
1103	>
1104	>	/* determines the number of layers in use */
1105	>	if(k_c_max < k_c_curr)
1106	>	{
1107	>	k_c_max = k_c_curr;
1108	>	}
1109	>
1110	>	H_c_l[c[i]].push_back(/*l[*/i/*]*/);
1111	>	}
1112	>
1113	>	int m;
1114	>	/* the neighbor matrix */
1115	>	vector<vector<int> >neighborMatW = vector<vector<int> >(nGroups_);
1116	>
1117	>	//      vector<pair<int, int> > neighborList;
1118	>	groupCutoffs cuts;
1119	>
1120	>	/* loops over objects(atoms, rigidBodies, cutoffGroups, etc.) */
1121	>	for(int i = 0; i < nGroups_; ++i)
1122	>	{
1123	>	m = 0;
1124	>	/* c' */
1125	>	int c1 = c[i];
1126	>	Vector3i c1v = idxToV(c1, nCells_);
1127	>
1128	>	/* loops over the neighboring cells c'' */
1129	>	for (vector<Vector3i>::iterator os = cellOffsets_.begin(); os != cellOffsets_.end(); ++os)
1130	>	{
1131	>	Vector3i c2v = c1v + (*os);
1132	>
1133	>	if (c2v.x() >= nCells_.x())
1134	>	{
1135	>	c2v.x() = 0;
1136	>	} else if (c2v.x() < 0)
1137	>	{
1138	>	c2v.x() = nCells_.x() - 1;
1139	>	}
1140	>
1141	>	if (c2v.y() >= nCells_.y())
1142	>	{
1143	>	c2v.y() = 0;
1144	>	} else if (c2v.y() < 0)
1145	>	{
1146	>	c2v.y() = nCells_.y() - 1;
1147	>	}
1148	>
1149	>	if (c2v.z() >= nCells_.z())
1150	>	{
1151	>	c2v.z() = 0;
1152	>	} else if (c2v.z() < 0)
1153	>	{
1154	>	c2v.z() = nCells_.z() - 1;
1155	>	}
1156	>
1157	>	int c2 = Vlinear(c2v, nCells_);
1158	>	/* loops over layers l to access the neighbor atoms */
1159	>	for (vector<int>::iterator j = H_c_l[c2].begin(); j != H_c_l[c2].end(); ++j)
1160	>	{
1161	>	//                              if i'' = 0 then break // doesn't apply to vector implementation of matrix
1162	>	//                              if(i != *j)
1163	>	if (c2 != c1 || (*j) < (i))
1164	>	{
1165	>	dr = snap_->cgData.position[(*j)] - snap_->cgData.position[(i)];
1166	>	snap_->wrapVector(dr);
1167	>	cuts = getGroupCutoffs((i), (*j));
1168	>	if (dr.lengthSquare() < cuts.third)
1169	>	{
1170	>	++m;
1171	>	/* transposed version of Rapaport W mat, to occupy successive memory locations on CPU */
1172	>	neighborMatW[i].push_back(*j);
1173	>	//                                              neighborList.push_back(make_pair((i), (*j)));
1174	>	}
1175	>	}
1176	>	}
1177	>	}
1178	>	}
1179	>
1180	>	// save the local cutoff group positions for the check that is
1181	>	// done on each loop:
1182	>	saved_CG_positions_.clear();
1183	>	for (int i = 0; i < nGroups_; i++)
1184	>	saved_CG_positions_.push_back(snap_->cgData.position[i]);
1185	>
1186	>	return neighborMatW;
1187		}
1188
1189	+	/*
1190	+	* buildNeighborList
1191	+	*
1192	+	* first element of pair is row-indexed CutoffGroup
1193	+	* second element of pair is column-indexed CutoffGroup
1194	+	*/
1195	+	vector<pair<int, int> > ForceMatrixDecomposition::buildNeighborList() {
1196	+
1197	+	vector<pair<int, int> > neighborList;
1198	+	groupCutoffs cuts;
1199	+	bool doAllPairs = false;
1200
1201	+	#ifdef IS_MPI
1202	+	cellListRow_.clear();
1203	+	cellListCol_.clear();
1204	+	#else
1205	+	cellList_.clear();
1206	+	#endif
1207	+
1208	+	RealType rList_ = (largestRcut_ + skinThickness_);
1209	+	RealType rl2 = rList_ * rList_;
1210	+	Snapshot* snap_ = sman_->getCurrentSnapshot();
1211	+	Mat3x3d Hmat = snap_->getHmat();
1212	+	Vector3d Hx = Hmat.getColumn(0);
1213	+	Vector3d Hy = Hmat.getColumn(1);
1214	+	Vector3d Hz = Hmat.getColumn(2);
1215	+
1216	+	nCells_.x() = (int) ( Hx.length() )/ rList_;
1217	+	nCells_.y() = (int) ( Hy.length() )/ rList_;
1218	+	nCells_.z() = (int) ( Hz.length() )/ rList_;
1219	+
1220	+	// handle small boxes where the cell offsets can end up repeating cells
1221	+
1222	+	if (nCells_.x() < 3) doAllPairs = true;
1223	+	if (nCells_.y() < 3) doAllPairs = true;
1224	+	if (nCells_.z() < 3) doAllPairs = true;
1225	+
1226	+	Mat3x3d invHmat = snap_->getInvHmat();
1227	+	Vector3d rs, scaled, dr;
1228	+	Vector3i whichCell;
1229	+	int cellIndex;
1230	+	int nCtot = nCells_.x() * nCells_.y() * nCells_.z();
1231	+
1232	+	#ifdef IS_MPI
1233	+	cellListRow_.resize(nCtot);
1234	+	cellListCol_.resize(nCtot);
1235	+	#else
1236	+	cellList_.resize(nCtot);
1237	+	#endif
1238	+
1239	+	if (!doAllPairs) {
1240	+	#ifdef IS_MPI
1241	+
1242	+	for (int i = 0; i < nGroupsInRow_; i++) {
1243	+	rs = cgRowData.position[i];
1244	+
1245	+	// scaled positions relative to the box vectors
1246	+	scaled = invHmat * rs;
1247	+
1248	+	// wrap the vector back into the unit box by subtracting integer box
1249	+	// numbers
1250	+	for (int j = 0; j < 3; j++) {
1251	+	scaled[j] -= roundMe(scaled[j]);
1252	+	scaled[j] += 0.5;
1253	+	}
1254	+
1255	+	// find xyz-indices of cell that cutoffGroup is in.
1256	+	whichCell.x() = nCells_.x() * scaled.x();
1257	+	whichCell.y() = nCells_.y() * scaled.y();
1258	+	whichCell.z() = nCells_.z() * scaled.z();
1259	+
1260	+	// find single index of this cell:
1261	+	cellIndex = Vlinear(whichCell, nCells_);
1262	+
1263	+	// add this cutoff group to the list of groups in this cell;
1264	+	cellListRow_[cellIndex].push_back(i);
1265	+	}
1266	+	for (int i = 0; i < nGroupsInCol_; i++) {
1267	+	rs = cgColData.position[i];
1268	+
1269	+	// scaled positions relative to the box vectors
1270	+	scaled = invHmat * rs;
1271	+
1272	+	// wrap the vector back into the unit box by subtracting integer box
1273	+	// numbers
1274	+	for (int j = 0; j < 3; j++) {
1275	+	scaled[j] -= roundMe(scaled[j]);
1276	+	scaled[j] += 0.5;
1277	+	}
1278	+
1279	+	// find xyz-indices of cell that cutoffGroup is in.
1280	+	whichCell.x() = nCells_.x() * scaled.x();
1281	+	whichCell.y() = nCells_.y() * scaled.y();
1282	+	whichCell.z() = nCells_.z() * scaled.z();
1283	+
1284	+	// find single index of this cell:
1285	+	cellIndex = Vlinear(whichCell, nCells_);
1286	+
1287	+	// add this cutoff group to the list of groups in this cell;
1288	+	cellListCol_[cellIndex].push_back(i);
1289	+	}
1290	+	#else
1291	+	for (int i = 0; i < nGroups_; i++) {
1292	+	rs = snap_->cgData.position[i];
1293	+
1294	+	// scaled positions relative to the box vectors
1295	+	scaled = invHmat * rs;
1296	+
1297	+	// wrap the vector back into the unit box by subtracting integer box
1298	+	// numbers
1299	+	for (int j = 0; j < 3; j++) {
1300	+	scaled[j] -= roundMe(scaled[j]);
1301	+	scaled[j] += 0.5;
1302	+	}
1303	+
1304	+	// find xyz-indices of cell that cutoffGroup is in.
1305	+	whichCell.x() = nCells_.x() * scaled.x();
1306	+	whichCell.y() = nCells_.y() * scaled.y();
1307	+	whichCell.z() = nCells_.z() * scaled.z();
1308	+
1309	+	// find single index of this cell:
1310	+	cellIndex = Vlinear(whichCell, nCells_);
1311	+
1312	+	// add this cutoff group to the list of groups in this cell;
1313	+	cellList_[cellIndex].push_back(i);
1314	+	}
1315	+	#endif
1316	+
1317	+	for (int m1z = 0; m1z < nCells_.z(); m1z++) {
1318	+	for (int m1y = 0; m1y < nCells_.y(); m1y++) {
1319	+	for (int m1x = 0; m1x < nCells_.x(); m1x++) {
1320	+	Vector3i m1v(m1x, m1y, m1z);
1321	+	int m1 = Vlinear(m1v, nCells_);
1322	+
1323	+	for (vector<Vector3i>::iterator os = cellOffsets_.begin();
1324	+	os != cellOffsets_.end(); ++os) {
1325	+
1326	+	Vector3i m2v = m1v + (*os);
1327	+
1328	+	if (m2v.x() >= nCells_.x()) {
1329	+	m2v.x() = 0;
1330	+	} else if (m2v.x() < 0) {
1331	+	m2v.x() = nCells_.x() - 1;
1332	+	}
1333	+
1334	+	if (m2v.y() >= nCells_.y()) {
1335	+	m2v.y() = 0;
1336	+	} else if (m2v.y() < 0) {
1337	+	m2v.y() = nCells_.y() - 1;
1338	+	}
1339	+
1340	+	if (m2v.z() >= nCells_.z()) {
1341	+	m2v.z() = 0;
1342	+	} else if (m2v.z() < 0) {
1343	+	m2v.z() = nCells_.z() - 1;
1344	+	}
1345	+
1346	+	int m2 = Vlinear (m2v, nCells_);
1347	+
1348	+	#ifdef IS_MPI
1349	+	for (vector<int>::iterator j1 = cellListRow_[m1].begin();
1350	+	j1 != cellListRow_[m1].end(); ++j1) {
1351	+	for (vector<int>::iterator j2 = cellListCol_[m2].begin();
1352	+	j2 != cellListCol_[m2].end(); ++j2) {
1353	+
1354	+	// In parallel, we need to visit *all* pairs of row &
1355	+	// column indicies and will truncate later on.
1356	+	dr = cgColData.position[(*j2)] - cgRowData.position[(*j1)];
1357	+	snap_->wrapVector(dr);
1358	+	cuts = getGroupCutoffs( (*j1), (*j2) );
1359	+	if (dr.lengthSquare() < cuts.third) {
1360	+	neighborList.push_back(make_pair((*j1), (*j2)));
1361	+	}
1362	+	}
1363	+	}
1364	+	#else
1365	+
1366	+	for (vector<int>::iterator j1 = cellList_[m1].begin();
1367	+	j1 != cellList_[m1].end(); ++j1) {
1368	+	for (vector<int>::iterator j2 = cellList_[m2].begin();
1369	+	j2 != cellList_[m2].end(); ++j2) {
1370	+
1371	+	// Always do this if we're in different cells or if
1372	+	// we're in the same cell and the global index of the
1373	+	// j2 cutoff group is less than the j1 cutoff group
1374	+
1375	+	if (m2 != m1 || (*j2) < (*j1)) {
1376	+	dr = snap_->cgData.position[(*j2)] - snap_->cgData.position[(*j1)];
1377	+	snap_->wrapVector(dr);
1378	+	cuts = getGroupCutoffs( (*j1), (*j2) );
1379	+	if (dr.lengthSquare() < cuts.third) {
1380	+	neighborList.push_back(make_pair((*j1), (*j2)));
1381	+	}
1382	+	}
1383	+	}
1384	+	}
1385	+	#endif
1386	+	}
1387	+	}
1388	+	}
1389	+	}
1390	+	} else {
1391	+	// branch to do all cutoff group pairs
1392	+	#ifdef IS_MPI
1393	+	for (int j1 = 0; j1 < nGroupsInRow_; j1++) {
1394	+	for (int j2 = 0; j2 < nGroupsInCol_; j2++) {
1395	+	dr = cgColData.position[j2] - cgRowData.position[j1];
1396	+	snap_->wrapVector(dr);
1397	+	cuts = getGroupCutoffs( j1, j2 );
1398	+	if (dr.lengthSquare() < cuts.third) {
1399	+	neighborList.push_back(make_pair(j1, j2));
1400	+	}
1401	+	}
1402	+	}
1403	+	#else
1404	+	for (int j1 = 0; j1 < nGroups_ - 1; j1++) {
1405	+	for (int j2 = j1 + 1; j2 < nGroups_; j2++) {
1406	+	dr = snap_->cgData.position[j2] - snap_->cgData.position[j1];
1407	+	snap_->wrapVector(dr);
1408	+	cuts = getGroupCutoffs( j1, j2 );
1409	+	if (dr.lengthSquare() < cuts.third) {
1410	+	neighborList.push_back(make_pair(j1, j2));
1411	+	}
1412	+	}
1413	+	}
1414	+	#endif
1415	+	}
1416	+
1417	+	// save the local cutoff group positions for the check that is
1418	+	// done on each loop:
1419	+	saved_CG_positions_.clear();
1420	+	for (int i = 0; i < nGroups_; i++)
1421	+	saved_CG_positions_.push_back(snap_->cgData.position[i]);
1422	+
1423	+	return neighborList;
1424	+	}
1425	+	} //end namespace OpenMD

Diff Legend

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