ViewVC Help
View File | Revision Log | Show Annotations | View Changeset | Root Listing
root/OpenMD/branches/development/src/parallel/ForceMatrixDecomposition.cpp
(Generate patch)

Comparing branches/development/src/parallel/ForceMatrixDecomposition.cpp (file contents):
Revision 1586 by gezelter, Tue Jun 21 06:34:35 2011 UTC vs.
Revision 1849 by gezelter, Wed Feb 20 13:52:51 2013 UTC

# Line 36 | 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"
# Line 47 | Line 48 | namespace OpenMD {
48   using namespace std;
49   namespace OpenMD {
50  
51 +  ForceMatrixDecomposition::ForceMatrixDecomposition(SimInfo* info, InteractionManager* iMan) : ForceDecomposition(info, iMan) {
52 +
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 +    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 +
88 +
89    /**
90     * distributeInitialData is essentially a copy of the older fortran
91     * SimulationSetup
92     */
54  
93    void ForceMatrixDecomposition::distributeInitialData() {
94      snap_ = sman_->getCurrentSnapshot();
95      storageLayout_ = sman_->getStorageLayout();
96      ff_ = info_->getForceField();
97      nLocal_ = snap_->getNumberOfAtoms();
98 <
98 >  
99      nGroups_ = info_->getNLocalCutoffGroups();
100      // gather the information for atomtype IDs (atids):
101      idents = info_->getIdentArray();
# Line 67 | Line 105 | namespace OpenMD {
105  
106      massFactors = info_->getMassFactors();
107  
108 <    PairList excludes = info_->getExcludedInteractions();
109 <    PairList oneTwo = info_->getOneTwoInteractions();
110 <    PairList oneThree = info_->getOneThreeInteractions();
111 <    PairList oneFour = info_->getOneFourInteractions();
112 <
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  
121 <    AtomCommIntRow = new Communicator<Row,int>(nLocal_);
122 <    AtomCommRealRow = new Communicator<Row,RealType>(nLocal_);
79 <    AtomCommVectorRow = new Communicator<Row,Vector3d>(nLocal_);
80 <    AtomCommMatrixRow = new Communicator<Row,Mat3x3d>(nLocal_);
81 <    AtomCommPotRow = new Communicator<Row,potVec>(nLocal_);
121 >    MPI::Intracomm row = rowComm.getComm();
122 >    MPI::Intracomm col = colComm.getComm();
123  
124 <    AtomCommIntColumn = new Communicator<Column,int>(nLocal_);
125 <    AtomCommRealColumn = new Communicator<Column,RealType>(nLocal_);
126 <    AtomCommVectorColumn = new Communicator<Column,Vector3d>(nLocal_);
127 <    AtomCommMatrixColumn = new Communicator<Column,Mat3x3d>(nLocal_);
128 <    AtomCommPotColumn = new Communicator<Column,potVec>(nLocal_);
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 <    cgCommIntRow = new Communicator<Row,int>(nGroups_);
131 <    cgCommVectorRow = new Communicator<Row,Vector3d>(nGroups_);
132 <    cgCommIntColumn = new Communicator<Column,int>(nGroups_);
133 <    cgCommVectorColumn = new Communicator<Column,Vector3d>(nGroups_);
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 <    nAtomsInRow_ = AtomCommIntRow->getSize();
137 <    nAtomsInCol_ = AtomCommIntColumn->getSize();
138 <    nGroupsInRow_ = cgCommIntRow->getSize();
139 <    nGroupsInCol_ = cgCommIntColumn->getSize();
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 +    nAtomsInRow_ = AtomPlanIntRow->getSize();
142 +    nAtomsInCol_ = AtomPlanIntColumn->getSize();
143 +    nGroupsInRow_ = cgPlanIntRow->getSize();
144 +    nGroupsInCol_ = cgPlanIntColumn->getSize();
145 +
146      // Modify the data storage objects with the correct layouts and sizes:
147      atomRowData.resize(nAtomsInRow_);
148      atomRowData.setStorageLayout(storageLayout_);
# Line 104 | Line 151 | namespace OpenMD {
151      cgRowData.resize(nGroupsInRow_);
152      cgRowData.setStorageLayout(DataStorage::dslPosition);
153      cgColData.resize(nGroupsInCol_);
154 <    cgColData.setStorageLayout(DataStorage::dslPosition);
155 <        
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 <    AtomCommIntRow->gather(idents, identsRow);
165 <    AtomCommIntColumn->gather(idents, identsCol);
164 >    AtomPlanIntRow->gather(idents, identsRow);
165 >    AtomPlanIntColumn->gather(idents, identsCol);
166      
167 <    AtomCommIntRow->gather(AtomLocalToGlobal, AtomRowToGlobal);
168 <    AtomCommIntColumn->gather(AtomLocalToGlobal, AtomColToGlobal);
169 <    
118 <    cgCommIntRow->gather(cgLocalToGlobal, cgRowToGlobal);
119 <    cgCommIntColumn->gather(cgLocalToGlobal, cgColToGlobal);
167 >    // allocate memory for the parallel objects
168 >    atypesRow.resize(nAtomsInRow_);
169 >    atypesCol.resize(nAtomsInCol_);
170  
171 <    AtomCommRealRow->gather(massFactors, massFactorsRow);
172 <    AtomCommRealColumn->gather(massFactors, massFactorsCol);
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 +    expot_row.resize(nAtomsInRow_);
180 +    expot_col.resize(nAtomsInCol_);
181 +
182 +    AtomRowToGlobal.resize(nAtomsInRow_);
183 +    AtomColToGlobal.resize(nAtomsInCol_);
184 +    AtomPlanIntRow->gather(AtomLocalToGlobal, AtomRowToGlobal);
185 +    AtomPlanIntColumn->gather(AtomLocalToGlobal, AtomColToGlobal);
186 +
187 +    cgRowToGlobal.resize(nGroupsInRow_);
188 +    cgColToGlobal.resize(nGroupsInCol_);
189 +    cgPlanIntRow->gather(cgLocalToGlobal, cgRowToGlobal);
190 +    cgPlanIntColumn->gather(cgLocalToGlobal, cgColToGlobal);
191 +
192 +    massFactorsRow.resize(nAtomsInRow_);
193 +    massFactorsCol.resize(nAtomsInCol_);
194 +    AtomPlanRealRow->gather(massFactors, massFactorsRow);
195 +    AtomPlanRealColumn->gather(massFactors, massFactorsCol);
196 +
197      groupListRow_.clear();
198      groupListRow_.resize(nGroupsInRow_);
199      for (int i = 0; i < nGroupsInRow_; i++) {
# Line 143 | Line 216 | namespace OpenMD {
216        }      
217      }
218  
219 <    skipsForAtom.clear();
220 <    skipsForAtom.resize(nAtomsInRow_);
219 >    excludesForAtom.clear();
220 >    excludesForAtom.resize(nAtomsInRow_);
221      toposForAtom.clear();
222      toposForAtom.resize(nAtomsInRow_);
223      topoDist.clear();
# Line 155 | Line 228 | namespace OpenMD {
228        for (int j = 0; j < nAtomsInCol_; j++) {
229          int jglob = AtomColToGlobal[j];
230  
231 <        if (excludes.hasPair(iglob, jglob))
232 <          skipsForAtom[i].push_back(j);      
231 >        if (excludes->hasPair(iglob, jglob))
232 >          excludesForAtom[i].push_back(j);      
233          
234 <        if (oneTwo.hasPair(iglob, jglob)) {
234 >        if (oneTwo->hasPair(iglob, jglob)) {
235            toposForAtom[i].push_back(j);
236            topoDist[i].push_back(1);
237          } else {
238 <          if (oneThree.hasPair(iglob, jglob)) {
238 >          if (oneThree->hasPair(iglob, jglob)) {
239              toposForAtom[i].push_back(j);
240              topoDist[i].push_back(2);
241            } else {
242 <            if (oneFour.hasPair(iglob, jglob)) {
242 >            if (oneFour->hasPair(iglob, jglob)) {
243                toposForAtom[i].push_back(j);
244                topoDist[i].push_back(3);
245              }
# Line 175 | Line 248 | namespace OpenMD {
248        }      
249      }
250  
251 < #endif
252 <
253 <    groupList_.clear();
181 <    groupList_.resize(nGroups_);
182 <    for (int i = 0; i < nGroups_; i++) {
183 <      int gid = cgLocalToGlobal[i];
184 <      for (int j = 0; j < nLocal_; j++) {
185 <        int aid = AtomLocalToGlobal[j];
186 <        if (globalGroupMembership[aid] == gid) {
187 <          groupList_[i].push_back(j);
188 <        }
189 <      }      
190 <    }
191 <
192 <    skipsForAtom.clear();
193 <    skipsForAtom.resize(nLocal_);
251 > #else
252 >    excludesForAtom.clear();
253 >    excludesForAtom.resize(nLocal_);
254      toposForAtom.clear();
255      toposForAtom.resize(nLocal_);
256      topoDist.clear();
# Line 202 | Line 262 | namespace OpenMD {
262        for (int j = 0; j < nLocal_; j++) {
263          int jglob = AtomLocalToGlobal[j];
264  
265 <        if (excludes.hasPair(iglob, jglob))
266 <          skipsForAtom[i].push_back(j);              
265 >        if (excludes->hasPair(iglob, jglob))
266 >          excludesForAtom[i].push_back(j);              
267          
268 <        if (oneTwo.hasPair(iglob, jglob)) {
268 >        if (oneTwo->hasPair(iglob, jglob)) {
269            toposForAtom[i].push_back(j);
270            topoDist[i].push_back(1);
271          } else {
272 <          if (oneThree.hasPair(iglob, jglob)) {
272 >          if (oneThree->hasPair(iglob, jglob)) {
273              toposForAtom[i].push_back(j);
274              topoDist[i].push_back(2);
275            } else {
276 <            if (oneFour.hasPair(iglob, jglob)) {
276 >            if (oneFour->hasPair(iglob, jglob)) {
277                toposForAtom[i].push_back(j);
278                topoDist[i].push_back(3);
279              }
# Line 221 | Line 281 | namespace OpenMD {
281          }
282        }      
283      }
284 <    
284 > #endif
285 >
286 >    // allocate memory for the parallel objects
287 >    atypesLocal.resize(nLocal_);
288 >
289 >    for (int i = 0; i < nLocal_; i++)
290 >      atypesLocal[i] = ff_->getAtomType(idents[i]);
291 >
292 >    groupList_.clear();
293 >    groupList_.resize(nGroups_);
294 >    for (int i = 0; i < nGroups_; i++) {
295 >      int gid = cgLocalToGlobal[i];
296 >      for (int j = 0; j < nLocal_; j++) {
297 >        int aid = AtomLocalToGlobal[j];
298 >        if (globalGroupMembership[aid] == gid) {
299 >          groupList_[i].push_back(j);
300 >        }
301 >      }      
302 >    }
303 >
304 >
305      createGtypeCutoffMap();
306 +
307    }
308    
309    void ForceMatrixDecomposition::createGtypeCutoffMap() {
310      
311      RealType tol = 1e-6;
312 <    RealType rc;
312 >    largestRcut_ = 0.0;
313      int atid;
314      set<AtomType*> atypes = info_->getSimulatedAtomTypes();
315 <    vector<RealType> atypeCutoff;
316 <    atypeCutoff.resize( atypes.size() );
315 >    
316 >    map<int, RealType> atypeCutoff;
317        
318      for (set<AtomType*>::iterator at = atypes.begin();
319           at != atypes.end(); ++at){
320        atid = (*at)->getIdent();
321 <
241 <      if (userChoseCutoff_)
321 >      if (userChoseCutoff_)
322          atypeCutoff[atid] = userCutoff_;
323        else
324          atypeCutoff[atid] = interactionMan_->getSuggestedCutoffRadius(*at);
325      }
326 <
326 >    
327      vector<RealType> gTypeCutoffs;
248
328      // first we do a single loop over the cutoff groups to find the
329      // largest cutoff for any atypes present in this group.
330   #ifdef IS_MPI
# Line 303 | Line 382 | namespace OpenMD {
382  
383      vector<RealType> groupCutoff(nGroups_, 0.0);
384      groupToGtype.resize(nGroups_);
306
385      for (int cg1 = 0; cg1 < nGroups_; cg1++) {
308
386        groupCutoff[cg1] = 0.0;
387        vector<int> atomList = getAtomsInGroupRow(cg1);
311
388        for (vector<int>::iterator ia = atomList.begin();
389             ia != atomList.end(); ++ia) {            
390          int atom1 = (*ia);
391          atid = idents[atom1];
392 <        if (atypeCutoff[atid] > groupCutoff[cg1]) {
392 >        if (atypeCutoff[atid] > groupCutoff[cg1])
393            groupCutoff[cg1] = atypeCutoff[atid];
318        }
394        }
395 <
395 >      
396        bool gTypeFound = false;
397 <      for (int gt = 0; gt < gTypeCutoffs.size(); gt++) {
397 >      for (unsigned int gt = 0; gt < gTypeCutoffs.size(); gt++) {
398          if (abs(groupCutoff[cg1] - gTypeCutoffs[gt]) < tol) {
399            groupToGtype[cg1] = gt;
400            gTypeFound = true;
401          }
402        }
403 <      if (!gTypeFound) {
403 >      if (!gTypeFound) {      
404          gTypeCutoffs.push_back( groupCutoff[cg1] );
405          groupToGtype[cg1] = gTypeCutoffs.size() - 1;
406        }      
# Line 334 | Line 409 | namespace OpenMD {
409  
410      // Now we find the maximum group cutoff value present in the simulation
411  
412 <    RealType groupMax = *max_element(gTypeCutoffs.begin(), gTypeCutoffs.end());
412 >    RealType groupMax = *max_element(gTypeCutoffs.begin(),
413 >                                     gTypeCutoffs.end());
414  
415   #ifdef IS_MPI
416 <    MPI::COMM_WORLD.Allreduce(&groupMax, &groupMax, 1, MPI::REALTYPE, MPI::MAX);
416 >    MPI::COMM_WORLD.Allreduce(&groupMax, &groupMax, 1, MPI::REALTYPE,
417 >                              MPI::MAX);
418   #endif
419      
420      RealType tradRcut = groupMax;
421  
422 <    for (int i = 0; i < gTypeCutoffs.size();  i++) {
423 <      for (int j = 0; j < gTypeCutoffs.size();  j++) {      
422 >    for (unsigned int i = 0; i < gTypeCutoffs.size();  i++) {
423 >      for (unsigned int j = 0; j < gTypeCutoffs.size();  j++) {      
424          RealType thisRcut;
425          switch(cutoffPolicy_) {
426          case TRADITIONAL:
# Line 367 | Line 444 | namespace OpenMD {
444  
445          pair<int,int> key = make_pair(i,j);
446          gTypeCutoffMap[key].first = thisRcut;
370
447          if (thisRcut > largestRcut_) largestRcut_ = thisRcut;
372
448          gTypeCutoffMap[key].second = thisRcut*thisRcut;
374        
449          gTypeCutoffMap[key].third = pow(thisRcut + skinThickness_, 2);
376
450          // sanity check
451          
452          if (userChoseCutoff_) {
# Line 390 | Line 463 | namespace OpenMD {
463      }
464    }
465  
393
466    groupCutoffs ForceMatrixDecomposition::getGroupCutoffs(int cg1, int cg2) {
467      int i, j;  
468   #ifdef IS_MPI
# Line 404 | Line 476 | namespace OpenMD {
476    }
477  
478    int ForceMatrixDecomposition::getTopologicalDistance(int atom1, int atom2) {
479 <    for (int j = 0; j < toposForAtom[atom1].size(); j++) {
479 >    for (unsigned int j = 0; j < toposForAtom[atom1].size(); j++) {
480        if (toposForAtom[atom1][j] == atom2)
481          return topoDist[atom1][j];
482      }
# Line 414 | Line 486 | namespace OpenMD {
486    void ForceMatrixDecomposition::zeroWorkArrays() {
487      pairwisePot = 0.0;
488      embeddingPot = 0.0;
489 +    excludedPot = 0.0;
490 +    excludedSelfPot = 0.0;
491  
492   #ifdef IS_MPI
493      if (storageLayout_ & DataStorage::dslForce) {
# Line 432 | Line 506 | namespace OpenMD {
506      fill(pot_col.begin(), pot_col.end(),
507           Vector<RealType, N_INTERACTION_FAMILIES> (0.0));  
508  
509 +    fill(expot_row.begin(), expot_row.end(),
510 +         Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
511 +
512 +    fill(expot_col.begin(), expot_col.end(),
513 +         Vector<RealType, N_INTERACTION_FAMILIES> (0.0));  
514 +
515      if (storageLayout_ & DataStorage::dslParticlePot) {    
516 <      fill(atomRowData.particlePot.begin(), atomRowData.particlePot.end(), 0.0);
517 <      fill(atomColData.particlePot.begin(), atomColData.particlePot.end(), 0.0);
516 >      fill(atomRowData.particlePot.begin(), atomRowData.particlePot.end(),
517 >           0.0);
518 >      fill(atomColData.particlePot.begin(), atomColData.particlePot.end(),
519 >           0.0);
520      }
521  
522      if (storageLayout_ & DataStorage::dslDensity) {      
# Line 443 | Line 525 | namespace OpenMD {
525      }
526  
527      if (storageLayout_ & DataStorage::dslFunctional) {  
528 <      fill(atomRowData.functional.begin(), atomRowData.functional.end(), 0.0);
529 <      fill(atomColData.functional.begin(), atomColData.functional.end(), 0.0);
528 >      fill(atomRowData.functional.begin(), atomRowData.functional.end(),
529 >           0.0);
530 >      fill(atomColData.functional.begin(), atomColData.functional.end(),
531 >           0.0);
532      }
533  
534      if (storageLayout_ & DataStorage::dslFunctionalDerivative) {      
# Line 455 | Line 539 | namespace OpenMD {
539      }
540  
541      if (storageLayout_ & DataStorage::dslSkippedCharge) {      
542 <      fill(atomRowData.skippedCharge.begin(), atomRowData.skippedCharge.end(), 0.0);
543 <      fill(atomColData.skippedCharge.begin(), atomColData.skippedCharge.end(), 0.0);
542 >      fill(atomRowData.skippedCharge.begin(),
543 >           atomRowData.skippedCharge.end(), 0.0);
544 >      fill(atomColData.skippedCharge.begin(),
545 >           atomColData.skippedCharge.end(), 0.0);
546      }
547  
548 < #else
549 <    
548 >    if (storageLayout_ & DataStorage::dslFlucQForce) {      
549 >      fill(atomRowData.flucQFrc.begin(),
550 >           atomRowData.flucQFrc.end(), 0.0);
551 >      fill(atomColData.flucQFrc.begin(),
552 >           atomColData.flucQFrc.end(), 0.0);
553 >    }
554 >
555 >    if (storageLayout_ & DataStorage::dslElectricField) {    
556 >      fill(atomRowData.electricField.begin(),
557 >           atomRowData.electricField.end(), V3Zero);
558 >      fill(atomColData.electricField.begin(),
559 >           atomColData.electricField.end(), V3Zero);
560 >    }
561 >
562 > #endif
563 >    // even in parallel, we need to zero out the local arrays:
564 >
565      if (storageLayout_ & DataStorage::dslParticlePot) {      
566        fill(snap_->atomData.particlePot.begin(),
567             snap_->atomData.particlePot.end(), 0.0);
# Line 470 | Line 571 | namespace OpenMD {
571        fill(snap_->atomData.density.begin(),
572             snap_->atomData.density.end(), 0.0);
573      }
574 +
575      if (storageLayout_ & DataStorage::dslFunctional) {
576        fill(snap_->atomData.functional.begin(),
577             snap_->atomData.functional.end(), 0.0);
578      }
579 +
580      if (storageLayout_ & DataStorage::dslFunctionalDerivative) {      
581        fill(snap_->atomData.functionalDerivative.begin(),
582             snap_->atomData.functionalDerivative.end(), 0.0);
583      }
584 +
585      if (storageLayout_ & DataStorage::dslSkippedCharge) {      
586        fill(snap_->atomData.skippedCharge.begin(),
587             snap_->atomData.skippedCharge.end(), 0.0);
588      }
589 < #endif
590 <    
589 >
590 >    if (storageLayout_ & DataStorage::dslElectricField) {      
591 >      fill(snap_->atomData.electricField.begin(),
592 >           snap_->atomData.electricField.end(), V3Zero);
593 >    }
594    }
595  
596  
# Line 493 | Line 600 | namespace OpenMD {
600   #ifdef IS_MPI
601      
602      // gather up the atomic positions
603 <    AtomCommVectorRow->gather(snap_->atomData.position,
603 >    AtomPlanVectorRow->gather(snap_->atomData.position,
604                                atomRowData.position);
605 <    AtomCommVectorColumn->gather(snap_->atomData.position,
605 >    AtomPlanVectorColumn->gather(snap_->atomData.position,
606                                   atomColData.position);
607      
608      // gather up the cutoff group positions
609 <    cgCommVectorRow->gather(snap_->cgData.position,
609 >
610 >    cgPlanVectorRow->gather(snap_->cgData.position,
611                              cgRowData.position);
612 <    cgCommVectorColumn->gather(snap_->cgData.position,
612 >
613 >    cgPlanVectorColumn->gather(snap_->cgData.position,
614                                 cgColData.position);
615 +
616 +
617 +
618 +    if (needVelocities_) {
619 +      // gather up the atomic velocities
620 +      AtomPlanVectorColumn->gather(snap_->atomData.velocity,
621 +                                   atomColData.velocity);
622 +      
623 +      cgPlanVectorColumn->gather(snap_->cgData.velocity,
624 +                                 cgColData.velocity);
625 +    }
626 +
627      
628      // if needed, gather the atomic rotation matrices
629      if (storageLayout_ & DataStorage::dslAmat) {
630 <      AtomCommMatrixRow->gather(snap_->atomData.aMat,
630 >      AtomPlanMatrixRow->gather(snap_->atomData.aMat,
631                                  atomRowData.aMat);
632 <      AtomCommMatrixColumn->gather(snap_->atomData.aMat,
632 >      AtomPlanMatrixColumn->gather(snap_->atomData.aMat,
633                                     atomColData.aMat);
634      }
635 <    
636 <    // if needed, gather the atomic eletrostatic frames
637 <    if (storageLayout_ & DataStorage::dslElectroFrame) {
638 <      AtomCommMatrixRow->gather(snap_->atomData.electroFrame,
639 <                                atomRowData.electroFrame);
640 <      AtomCommMatrixColumn->gather(snap_->atomData.electroFrame,
641 <                                   atomColData.electroFrame);
635 >
636 >    // if needed, gather the atomic eletrostatic information
637 >    if (storageLayout_ & DataStorage::dslDipole) {
638 >      AtomPlanVectorRow->gather(snap_->atomData.dipole,
639 >                                atomRowData.dipole);
640 >      AtomPlanVectorColumn->gather(snap_->atomData.dipole,
641 >                                   atomColData.dipole);
642      }
643 +
644 +    if (storageLayout_ & DataStorage::dslQuadrupole) {
645 +      AtomPlanMatrixRow->gather(snap_->atomData.quadrupole,
646 +                                atomRowData.quadrupole);
647 +      AtomPlanMatrixColumn->gather(snap_->atomData.quadrupole,
648 +                                   atomColData.quadrupole);
649 +    }
650 +        
651 +    // if needed, gather the atomic fluctuating charge values
652 +    if (storageLayout_ & DataStorage::dslFlucQPosition) {
653 +      AtomPlanRealRow->gather(snap_->atomData.flucQPos,
654 +                              atomRowData.flucQPos);
655 +      AtomPlanRealColumn->gather(snap_->atomData.flucQPos,
656 +                                 atomColData.flucQPos);
657 +    }
658 +
659   #endif      
660    }
661    
# Line 532 | Line 669 | namespace OpenMD {
669      
670      if (storageLayout_ & DataStorage::dslDensity) {
671        
672 <      AtomCommRealRow->scatter(atomRowData.density,
672 >      AtomPlanRealRow->scatter(atomRowData.density,
673                                 snap_->atomData.density);
674        
675        int n = snap_->atomData.density.size();
676        vector<RealType> rho_tmp(n, 0.0);
677 <      AtomCommRealColumn->scatter(atomColData.density, rho_tmp);
677 >      AtomPlanRealColumn->scatter(atomColData.density, rho_tmp);
678        for (int i = 0; i < n; i++)
679          snap_->atomData.density[i] += rho_tmp[i];
680      }
681 +
682 +    // this isn't necessary if we don't have polarizable atoms, but
683 +    // we'll leave it here for now.
684 +    if (storageLayout_ & DataStorage::dslElectricField) {
685 +      
686 +      AtomPlanVectorRow->scatter(atomRowData.electricField,
687 +                                 snap_->atomData.electricField);
688 +      
689 +      int n = snap_->atomData.electricField.size();
690 +      vector<Vector3d> field_tmp(n, V3Zero);
691 +      AtomPlanVectorColumn->scatter(atomColData.electricField,
692 +                                    field_tmp);
693 +      for (int i = 0; i < n; i++)
694 +        snap_->atomData.electricField[i] += field_tmp[i];
695 +    }
696   #endif
697    }
698  
# Line 553 | Line 705 | namespace OpenMD {
705      storageLayout_ = sman_->getStorageLayout();
706   #ifdef IS_MPI
707      if (storageLayout_ & DataStorage::dslFunctional) {
708 <      AtomCommRealRow->gather(snap_->atomData.functional,
708 >      AtomPlanRealRow->gather(snap_->atomData.functional,
709                                atomRowData.functional);
710 <      AtomCommRealColumn->gather(snap_->atomData.functional,
710 >      AtomPlanRealColumn->gather(snap_->atomData.functional,
711                                   atomColData.functional);
712      }
713      
714      if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
715 <      AtomCommRealRow->gather(snap_->atomData.functionalDerivative,
715 >      AtomPlanRealRow->gather(snap_->atomData.functionalDerivative,
716                                atomRowData.functionalDerivative);
717 <      AtomCommRealColumn->gather(snap_->atomData.functionalDerivative,
717 >      AtomPlanRealColumn->gather(snap_->atomData.functionalDerivative,
718                                   atomColData.functionalDerivative);
719      }
720   #endif
# Line 576 | Line 728 | namespace OpenMD {
728      int n = snap_->atomData.force.size();
729      vector<Vector3d> frc_tmp(n, V3Zero);
730      
731 <    AtomCommVectorRow->scatter(atomRowData.force, frc_tmp);
731 >    AtomPlanVectorRow->scatter(atomRowData.force, frc_tmp);
732      for (int i = 0; i < n; i++) {
733        snap_->atomData.force[i] += frc_tmp[i];
734        frc_tmp[i] = 0.0;
735      }
736      
737 <    AtomCommVectorColumn->scatter(atomColData.force, frc_tmp);
738 <    for (int i = 0; i < n; i++)
737 >    AtomPlanVectorColumn->scatter(atomColData.force, frc_tmp);
738 >    for (int i = 0; i < n; i++) {
739        snap_->atomData.force[i] += frc_tmp[i];
740 <    
741 <    
740 >    }
741 >        
742      if (storageLayout_ & DataStorage::dslTorque) {
743  
744 <      int nt = snap_->atomData.force.size();
744 >      int nt = snap_->atomData.torque.size();
745        vector<Vector3d> trq_tmp(nt, V3Zero);
746  
747 <      AtomCommVectorRow->scatter(atomRowData.torque, trq_tmp);
748 <      for (int i = 0; i < n; i++) {
747 >      AtomPlanVectorRow->scatter(atomRowData.torque, trq_tmp);
748 >      for (int i = 0; i < nt; i++) {
749          snap_->atomData.torque[i] += trq_tmp[i];
750          trq_tmp[i] = 0.0;
751        }
752        
753 <      AtomCommVectorColumn->scatter(atomColData.torque, trq_tmp);
754 <      for (int i = 0; i < n; i++)
753 >      AtomPlanVectorColumn->scatter(atomColData.torque, trq_tmp);
754 >      for (int i = 0; i < nt; i++)
755          snap_->atomData.torque[i] += trq_tmp[i];
756      }
757 +
758 +    if (storageLayout_ & DataStorage::dslSkippedCharge) {
759 +
760 +      int ns = snap_->atomData.skippedCharge.size();
761 +      vector<RealType> skch_tmp(ns, 0.0);
762 +
763 +      AtomPlanRealRow->scatter(atomRowData.skippedCharge, skch_tmp);
764 +      for (int i = 0; i < ns; i++) {
765 +        snap_->atomData.skippedCharge[i] += skch_tmp[i];
766 +        skch_tmp[i] = 0.0;
767 +      }
768 +      
769 +      AtomPlanRealColumn->scatter(atomColData.skippedCharge, skch_tmp);
770 +      for (int i = 0; i < ns; i++)
771 +        snap_->atomData.skippedCharge[i] += skch_tmp[i];
772 +            
773 +    }
774      
775 +    if (storageLayout_ & DataStorage::dslFlucQForce) {
776 +
777 +      int nq = snap_->atomData.flucQFrc.size();
778 +      vector<RealType> fqfrc_tmp(nq, 0.0);
779 +
780 +      AtomPlanRealRow->scatter(atomRowData.flucQFrc, fqfrc_tmp);
781 +      for (int i = 0; i < nq; i++) {
782 +        snap_->atomData.flucQFrc[i] += fqfrc_tmp[i];
783 +        fqfrc_tmp[i] = 0.0;
784 +      }
785 +      
786 +      AtomPlanRealColumn->scatter(atomColData.flucQFrc, fqfrc_tmp);
787 +      for (int i = 0; i < nq; i++)
788 +        snap_->atomData.flucQFrc[i] += fqfrc_tmp[i];
789 +            
790 +    }
791 +
792 +    if (storageLayout_ & DataStorage::dslElectricField) {
793 +
794 +      int nef = snap_->atomData.electricField.size();
795 +      vector<Vector3d> efield_tmp(nef, V3Zero);
796 +
797 +      AtomPlanVectorRow->scatter(atomRowData.electricField, efield_tmp);
798 +      for (int i = 0; i < nef; i++) {
799 +        snap_->atomData.electricField[i] += efield_tmp[i];
800 +        efield_tmp[i] = 0.0;
801 +      }
802 +      
803 +      AtomPlanVectorColumn->scatter(atomColData.electricField, efield_tmp);
804 +      for (int i = 0; i < nef; i++)
805 +        snap_->atomData.electricField[i] += efield_tmp[i];
806 +    }
807 +
808 +
809      nLocal_ = snap_->getNumberOfAtoms();
810  
811      vector<potVec> pot_temp(nLocal_,
812                              Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
813 +    vector<potVec> expot_temp(nLocal_,
814 +                              Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
815  
816      // scatter/gather pot_row into the members of my column
817            
818 <    AtomCommPotRow->scatter(pot_row, pot_temp);
818 >    AtomPlanPotRow->scatter(pot_row, pot_temp);
819 >    AtomPlanPotRow->scatter(expot_row, expot_temp);
820  
821 <    for (int ii = 0;  ii < pot_temp.size(); ii++ )
821 >    for (int ii = 0;  ii < pot_temp.size(); ii++ )
822        pairwisePot += pot_temp[ii];
823 <    
823 >
824 >    for (int ii = 0;  ii < expot_temp.size(); ii++ )
825 >      excludedPot += expot_temp[ii];
826 >        
827 >    if (storageLayout_ & DataStorage::dslParticlePot) {
828 >      // This is the pairwise contribution to the particle pot.  The
829 >      // embedding contribution is added in each of the low level
830 >      // non-bonded routines.  In single processor, this is done in
831 >      // unpackInteractionData, not in collectData.
832 >      for (int ii = 0; ii < N_INTERACTION_FAMILIES; ii++) {
833 >        for (int i = 0; i < nLocal_; i++) {
834 >          // factor of two is because the total potential terms are divided
835 >          // by 2 in parallel due to row/ column scatter      
836 >          snap_->atomData.particlePot[i] += 2.0 * pot_temp[i](ii);
837 >        }
838 >      }
839 >    }
840 >
841      fill(pot_temp.begin(), pot_temp.end(),
842           Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
843 +    fill(expot_temp.begin(), expot_temp.end(),
844 +         Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
845        
846 <    AtomCommPotColumn->scatter(pot_col, pot_temp);    
846 >    AtomPlanPotColumn->scatter(pot_col, pot_temp);    
847 >    AtomPlanPotColumn->scatter(expot_col, expot_temp);    
848      
849      for (int ii = 0;  ii < pot_temp.size(); ii++ )
850        pairwisePot += pot_temp[ii];    
851 +
852 +    for (int ii = 0;  ii < expot_temp.size(); ii++ )
853 +      excludedPot += expot_temp[ii];    
854 +
855 +    if (storageLayout_ & DataStorage::dslParticlePot) {
856 +      // This is the pairwise contribution to the particle pot.  The
857 +      // embedding contribution is added in each of the low level
858 +      // non-bonded routines.  In single processor, this is done in
859 +      // unpackInteractionData, not in collectData.
860 +      for (int ii = 0; ii < N_INTERACTION_FAMILIES; ii++) {
861 +        for (int i = 0; i < nLocal_; i++) {
862 +          // factor of two is because the total potential terms are divided
863 +          // by 2 in parallel due to row/ column scatter      
864 +          snap_->atomData.particlePot[i] += 2.0 * pot_temp[i](ii);
865 +        }
866 +      }
867 +    }
868 +    
869 +    if (storageLayout_ & DataStorage::dslParticlePot) {
870 +      int npp = snap_->atomData.particlePot.size();
871 +      vector<RealType> ppot_temp(npp, 0.0);
872 +
873 +      // This is the direct or embedding contribution to the particle
874 +      // pot.
875 +      
876 +      AtomPlanRealRow->scatter(atomRowData.particlePot, ppot_temp);
877 +      for (int i = 0; i < npp; i++) {
878 +        snap_->atomData.particlePot[i] += ppot_temp[i];
879 +      }
880 +
881 +      fill(ppot_temp.begin(), ppot_temp.end(), 0.0);
882 +      
883 +      AtomPlanRealColumn->scatter(atomColData.particlePot, ppot_temp);
884 +      for (int i = 0; i < npp; i++) {
885 +        snap_->atomData.particlePot[i] += ppot_temp[i];
886 +      }
887 +    }
888 +
889 +    for (int ii = 0; ii < N_INTERACTION_FAMILIES; ii++) {
890 +      RealType ploc1 = pairwisePot[ii];
891 +      RealType ploc2 = 0.0;
892 +      MPI::COMM_WORLD.Allreduce(&ploc1, &ploc2, 1, MPI::REALTYPE, MPI::SUM);
893 +      pairwisePot[ii] = ploc2;
894 +    }
895 +
896 +    for (int ii = 0; ii < N_INTERACTION_FAMILIES; ii++) {
897 +      RealType ploc1 = excludedPot[ii];
898 +      RealType ploc2 = 0.0;
899 +      MPI::COMM_WORLD.Allreduce(&ploc1, &ploc2, 1, MPI::REALTYPE, MPI::SUM);
900 +      excludedPot[ii] = ploc2;
901 +    }
902 +
903 +    // Here be dragons.
904 +    MPI::Intracomm col = colComm.getComm();
905 +
906 +    col.Allreduce(MPI::IN_PLACE,
907 +                  &snap_->frameData.conductiveHeatFlux[0], 3,
908 +                  MPI::REALTYPE, MPI::SUM);
909 +
910 +
911   #endif
912  
913 +  }
914 +
915 +  /**
916 +   * Collects information obtained during the post-pair (and embedding
917 +   * functional) loops onto local data structures.
918 +   */
919 +  void ForceMatrixDecomposition::collectSelfData() {
920 +    snap_ = sman_->getCurrentSnapshot();
921 +    storageLayout_ = sman_->getStorageLayout();
922 +
923 + #ifdef IS_MPI
924 +    for (int ii = 0; ii < N_INTERACTION_FAMILIES; ii++) {
925 +      RealType ploc1 = embeddingPot[ii];
926 +      RealType ploc2 = 0.0;
927 +      MPI::COMM_WORLD.Allreduce(&ploc1, &ploc2, 1, MPI::REALTYPE, MPI::SUM);
928 +      embeddingPot[ii] = ploc2;
929 +    }    
930 +    for (int ii = 0; ii < N_INTERACTION_FAMILIES; ii++) {
931 +      RealType ploc1 = excludedSelfPot[ii];
932 +      RealType ploc2 = 0.0;
933 +      MPI::COMM_WORLD.Allreduce(&ploc1, &ploc2, 1, MPI::REALTYPE, MPI::SUM);
934 +      excludedSelfPot[ii] = ploc2;
935 +    }    
936 + #endif
937 +    
938    }
939  
940 +
941 +
942    int ForceMatrixDecomposition::getNAtomsInRow() {  
943   #ifdef IS_MPI
944      return nAtomsInRow_;
# Line 666 | Line 979 | namespace OpenMD {
979      return d;    
980    }
981  
982 +  Vector3d ForceMatrixDecomposition::getGroupVelocityColumn(int cg2){
983 + #ifdef IS_MPI
984 +    return cgColData.velocity[cg2];
985 + #else
986 +    return snap_->cgData.velocity[cg2];
987 + #endif
988 +  }
989  
990 +  Vector3d ForceMatrixDecomposition::getAtomVelocityColumn(int atom2){
991 + #ifdef IS_MPI
992 +    return atomColData.velocity[atom2];
993 + #else
994 +    return snap_->atomData.velocity[atom2];
995 + #endif
996 +  }
997 +
998 +
999    Vector3d ForceMatrixDecomposition::getAtomToGroupVectorRow(int atom1, int cg1){
1000  
1001      Vector3d d;
# Line 724 | Line 1053 | namespace OpenMD {
1053      return d;    
1054    }
1055  
1056 <  vector<int> ForceMatrixDecomposition::getSkipsForAtom(int atom1) {
1057 <    return skipsForAtom[atom1];
1056 >  vector<int> ForceMatrixDecomposition::getExcludesForAtom(int atom1) {
1057 >    return excludesForAtom[atom1];
1058    }
1059  
1060    /**
1061 <   * There are a number of reasons to skip a pair or a
733 <   * particle. Mostly we do this to exclude atoms who are involved in
734 <   * short range interactions (bonds, bends, torsions), but we also
735 <   * need to exclude some overcounted interactions that result from
1061 >   * We need to exclude some overcounted interactions that result from
1062     * the parallel decomposition.
1063     */
1064 <  bool ForceMatrixDecomposition::skipAtomPair(int atom1, int atom2) {
1064 >  bool ForceMatrixDecomposition::skipAtomPair(int atom1, int atom2, int cg1, int cg2) {
1065      int unique_id_1, unique_id_2;
1066 <
1066 >        
1067   #ifdef IS_MPI
1068      // in MPI, we have to look up the unique IDs for each atom
1069      unique_id_1 = AtomRowToGlobal[atom1];
1070      unique_id_2 = AtomColToGlobal[atom2];
1071 +    // group1 = cgRowToGlobal[cg1];
1072 +    // group2 = cgColToGlobal[cg2];
1073 + #else
1074 +    unique_id_1 = AtomLocalToGlobal[atom1];
1075 +    unique_id_2 = AtomLocalToGlobal[atom2];
1076 +    int group1 = cgLocalToGlobal[cg1];
1077 +    int group2 = cgLocalToGlobal[cg2];
1078 + #endif  
1079  
746    // this situation should only arise in MPI simulations
1080      if (unique_id_1 == unique_id_2) return true;
1081 <    
1081 >
1082 > #ifdef IS_MPI
1083      // this prevents us from doing the pair on multiple processors
1084      if (unique_id_1 < unique_id_2) {
1085        if ((unique_id_1 + unique_id_2) % 2 == 0) return true;
1086      } else {
1087 <      if ((unique_id_1 + unique_id_2) % 2 == 1) return true;
1087 >      if ((unique_id_1 + unique_id_2) % 2 == 1) return true;
1088      }
1089 < #else
1090 <    // in the normal loop, the atom numbers are unique
1091 <    unique_id_1 = atom1;
1092 <    unique_id_2 = atom2;
1089 > #endif    
1090 >
1091 > #ifndef IS_MPI
1092 >    if (group1 == group2) {
1093 >      if (unique_id_1 < unique_id_2) return true;
1094 >    }
1095   #endif
1096      
1097 <    for (vector<int>::iterator i = skipsForAtom[atom1].begin();
1098 <         i != skipsForAtom[atom1].end(); ++i) {
1099 <      if ( (*i) == unique_id_2 ) return true;
1097 >    return false;
1098 >  }
1099 >
1100 >  /**
1101 >   * We need to handle the interactions for atoms who are involved in
1102 >   * the same rigid body as well as some short range interactions
1103 >   * (bonds, bends, torsions) differently from other interactions.
1104 >   * We'll still visit the pairwise routines, but with a flag that
1105 >   * tells those routines to exclude the pair from direct long range
1106 >   * interactions.  Some indirect interactions (notably reaction
1107 >   * field) must still be handled for these pairs.
1108 >   */
1109 >  bool ForceMatrixDecomposition::excludeAtomPair(int atom1, int atom2) {
1110 >
1111 >    // excludesForAtom was constructed to use row/column indices in the MPI
1112 >    // version, and to use local IDs in the non-MPI version:
1113 >    
1114 >    for (vector<int>::iterator i = excludesForAtom[atom1].begin();
1115 >         i != excludesForAtom[atom1].end(); ++i) {
1116 >      if ( (*i) == atom2 ) return true;
1117      }
1118  
1119      return false;
# Line 785 | Line 1138 | namespace OpenMD {
1138  
1139      // filling interaction blocks with pointers
1140    void ForceMatrixDecomposition::fillInteractionData(InteractionData &idat,
1141 <                                                     int atom1, int atom2) {    
1141 >                                                     int atom1, int atom2) {
1142 >
1143 >    idat.excluded = excludeAtomPair(atom1, atom2);
1144 >  
1145   #ifdef IS_MPI
1146 +    idat.atypes = make_pair( atypesRow[atom1], atypesCol[atom2]);
1147 +    //idat.atypes = make_pair( ff_->getAtomType(identsRow[atom1]),
1148 +    //                         ff_->getAtomType(identsCol[atom2]) );
1149      
791    idat.atypes = make_pair( ff_->getAtomType(identsRow[atom1]),
792                             ff_->getAtomType(identsCol[atom2]) );
793    
1150      if (storageLayout_ & DataStorage::dslAmat) {
1151        idat.A1 = &(atomRowData.aMat[atom1]);
1152        idat.A2 = &(atomColData.aMat[atom2]);
1153      }
1154      
799    if (storageLayout_ & DataStorage::dslElectroFrame) {
800      idat.eFrame1 = &(atomRowData.electroFrame[atom1]);
801      idat.eFrame2 = &(atomColData.electroFrame[atom2]);
802    }
803
1155      if (storageLayout_ & DataStorage::dslTorque) {
1156        idat.t1 = &(atomRowData.torque[atom1]);
1157        idat.t2 = &(atomColData.torque[atom2]);
1158      }
1159  
1160 +    if (storageLayout_ & DataStorage::dslDipole) {
1161 +      idat.dipole1 = &(atomRowData.dipole[atom1]);
1162 +      idat.dipole2 = &(atomColData.dipole[atom2]);
1163 +    }
1164 +
1165 +    if (storageLayout_ & DataStorage::dslQuadrupole) {
1166 +      idat.quadrupole1 = &(atomRowData.quadrupole[atom1]);
1167 +      idat.quadrupole2 = &(atomColData.quadrupole[atom2]);
1168 +    }
1169 +
1170      if (storageLayout_ & DataStorage::dslDensity) {
1171        idat.rho1 = &(atomRowData.density[atom1]);
1172        idat.rho2 = &(atomColData.density[atom2]);
# Line 826 | Line 1187 | namespace OpenMD {
1187        idat.particlePot2 = &(atomColData.particlePot[atom2]);
1188      }
1189  
1190 < #else
1190 >    if (storageLayout_ & DataStorage::dslSkippedCharge) {              
1191 >      idat.skippedCharge1 = &(atomRowData.skippedCharge[atom1]);
1192 >      idat.skippedCharge2 = &(atomColData.skippedCharge[atom2]);
1193 >    }
1194  
1195 <    idat.atypes = make_pair( ff_->getAtomType(idents[atom1]),
1196 <                             ff_->getAtomType(idents[atom2]) );
1195 >    if (storageLayout_ & DataStorage::dslFlucQPosition) {              
1196 >      idat.flucQ1 = &(atomRowData.flucQPos[atom1]);
1197 >      idat.flucQ2 = &(atomColData.flucQPos[atom2]);
1198 >    }
1199  
1200 + #else
1201 +    
1202 +    idat.atypes = make_pair( atypesLocal[atom1], atypesLocal[atom2]);
1203 +
1204      if (storageLayout_ & DataStorage::dslAmat) {
1205        idat.A1 = &(snap_->atomData.aMat[atom1]);
1206        idat.A2 = &(snap_->atomData.aMat[atom2]);
1207      }
1208  
1209 <    if (storageLayout_ & DataStorage::dslElectroFrame) {
840 <      idat.eFrame1 = &(snap_->atomData.electroFrame[atom1]);
841 <      idat.eFrame2 = &(snap_->atomData.electroFrame[atom2]);
842 <    }
1209 >    RealType ct = dot(idat.A1->getColumn(2), idat.A2->getColumn(2));
1210  
1211      if (storageLayout_ & DataStorage::dslTorque) {
1212        idat.t1 = &(snap_->atomData.torque[atom1]);
1213        idat.t2 = &(snap_->atomData.torque[atom2]);
1214 +    }
1215 +
1216 +    if (storageLayout_ & DataStorage::dslDipole) {
1217 +      idat.dipole1 = &(snap_->atomData.dipole[atom1]);
1218 +      idat.dipole2 = &(snap_->atomData.dipole[atom2]);
1219 +    }
1220 +
1221 +    if (storageLayout_ & DataStorage::dslQuadrupole) {
1222 +      idat.quadrupole1 = &(snap_->atomData.quadrupole[atom1]);
1223 +      idat.quadrupole2 = &(snap_->atomData.quadrupole[atom2]);
1224      }
1225  
1226      if (storageLayout_ & DataStorage::dslDensity) {    
# Line 866 | Line 1243 | namespace OpenMD {
1243        idat.particlePot2 = &(snap_->atomData.particlePot[atom2]);
1244      }
1245  
1246 +    if (storageLayout_ & DataStorage::dslSkippedCharge) {
1247 +      idat.skippedCharge1 = &(snap_->atomData.skippedCharge[atom1]);
1248 +      idat.skippedCharge2 = &(snap_->atomData.skippedCharge[atom2]);
1249 +    }
1250 +
1251 +    if (storageLayout_ & DataStorage::dslFlucQPosition) {              
1252 +      idat.flucQ1 = &(snap_->atomData.flucQPos[atom1]);
1253 +      idat.flucQ2 = &(snap_->atomData.flucQPos[atom2]);
1254 +    }
1255 +
1256   #endif
1257    }
1258  
1259    
1260    void ForceMatrixDecomposition::unpackInteractionData(InteractionData &idat, int atom1, int atom2) {    
1261   #ifdef IS_MPI
1262 <    pot_row[atom1] += 0.5 *  *(idat.pot);
1263 <    pot_col[atom2] += 0.5 *  *(idat.pot);
1262 >    pot_row[atom1] += RealType(0.5) *  *(idat.pot);
1263 >    pot_col[atom2] += RealType(0.5) *  *(idat.pot);
1264 >    expot_row[atom1] += RealType(0.5) *  *(idat.excludedPot);
1265 >    expot_col[atom2] += RealType(0.5) *  *(idat.excludedPot);
1266  
1267      atomRowData.force[atom1] += *(idat.f1);
1268      atomColData.force[atom2] -= *(idat.f1);
880 #else
881    pairwisePot += *(idat.pot);
1269  
1270 <    snap_->atomData.force[atom1] += *(idat.f1);
1271 <    snap_->atomData.force[atom2] -= *(idat.f1);
1272 < #endif
886 <    
887 <  }
888 <
889 <
890 <  void ForceMatrixDecomposition::fillSkipData(InteractionData &idat,
891 <                                              int atom1, int atom2) {
892 < #ifdef IS_MPI
893 <    idat.atypes = make_pair( ff_->getAtomType(identsRow[atom1]),
894 <                             ff_->getAtomType(identsCol[atom2]) );
895 <
896 <    if (storageLayout_ & DataStorage::dslElectroFrame) {
897 <      idat.eFrame1 = &(atomRowData.electroFrame[atom1]);
898 <      idat.eFrame2 = &(atomColData.electroFrame[atom2]);
1270 >    if (storageLayout_ & DataStorage::dslFlucQForce) {              
1271 >      atomRowData.flucQFrc[atom1] -= *(idat.dVdFQ1);
1272 >      atomColData.flucQFrc[atom2] -= *(idat.dVdFQ2);
1273      }
1274  
1275 <    if (storageLayout_ & DataStorage::dslTorque) {
1276 <      idat.t1 = &(atomRowData.torque[atom1]);
1277 <      idat.t2 = &(atomColData.torque[atom2]);
1275 >    if (storageLayout_ & DataStorage::dslElectricField) {              
1276 >      atomRowData.electricField[atom1] += *(idat.eField1);
1277 >      atomColData.electricField[atom2] += *(idat.eField2);
1278      }
1279  
906    if (storageLayout_ & DataStorage::dslSkippedCharge) {
907      idat.skippedCharge1 = &(atomRowData.skippedCharge[atom1]);
908      idat.skippedCharge2 = &(atomColData.skippedCharge[atom2]);
909    }
1280   #else
1281 <    idat.atypes = make_pair( ff_->getAtomType(idents[atom1]),
1282 <                             ff_->getAtomType(idents[atom2]) );
1281 >    pairwisePot += *(idat.pot);
1282 >    excludedPot += *(idat.excludedPot);
1283  
1284 <    if (storageLayout_ & DataStorage::dslElectroFrame) {
1285 <      idat.eFrame1 = &(snap_->atomData.electroFrame[atom1]);
916 <      idat.eFrame2 = &(snap_->atomData.electroFrame[atom2]);
917 <    }
1284 >    snap_->atomData.force[atom1] += *(idat.f1);
1285 >    snap_->atomData.force[atom2] -= *(idat.f1);
1286  
1287 <    if (storageLayout_ & DataStorage::dslTorque) {
1288 <      idat.t1 = &(snap_->atomData.torque[atom1]);
1289 <      idat.t2 = &(snap_->atomData.torque[atom2]);
1287 >    if (idat.doParticlePot) {
1288 >      // This is the pairwise contribution to the particle pot.  The
1289 >      // embedding contribution is added in each of the low level
1290 >      // non-bonded routines.  In parallel, this calculation is done
1291 >      // in collectData, not in unpackInteractionData.
1292 >      snap_->atomData.particlePot[atom1] += *(idat.vpair) * *(idat.sw);
1293 >      snap_->atomData.particlePot[atom2] += *(idat.vpair) * *(idat.sw);
1294      }
1295 +    
1296 +    if (storageLayout_ & DataStorage::dslFlucQForce) {              
1297 +      snap_->atomData.flucQFrc[atom1] -= *(idat.dVdFQ1);
1298 +      snap_->atomData.flucQFrc[atom2] -= *(idat.dVdFQ2);
1299 +    }
1300  
1301 <    if (storageLayout_ & DataStorage::dslSkippedCharge) {
1302 <      idat.skippedCharge1 = &(snap_->atomData.skippedCharge[atom1]);
1303 <      idat.skippedCharge2 = &(snap_->atomData.skippedCharge[atom2]);
1301 >    if (storageLayout_ & DataStorage::dslElectricField) {              
1302 >      snap_->atomData.electricField[atom1] += *(idat.eField1);
1303 >      snap_->atomData.electricField[atom2] += *(idat.eField2);
1304      }
928 #endif    
929  }
1305  
931
932  void ForceMatrixDecomposition::unpackSkipData(InteractionData &idat, int atom1, int atom2) {    
933 #ifdef IS_MPI
934    pot_row[atom1] += 0.5 *  *(idat.pot);
935    pot_col[atom2] += 0.5 *  *(idat.pot);
936 #else
937    pairwisePot += *(idat.pot);  
1306   #endif
1307 <
1307 >    
1308    }
1309  
942
1310    /*
1311     * buildNeighborList
1312     *
# Line 950 | Line 1317 | namespace OpenMD {
1317        
1318      vector<pair<int, int> > neighborList;
1319      groupCutoffs cuts;
1320 +    bool doAllPairs = false;
1321 +
1322   #ifdef IS_MPI
1323      cellListRow_.clear();
1324      cellListCol_.clear();
# Line 958 | Line 1327 | namespace OpenMD {
1327   #endif
1328  
1329      RealType rList_ = (largestRcut_ + skinThickness_);
961    RealType rl2 = rList_ * rList_;
1330      Snapshot* snap_ = sman_->getCurrentSnapshot();
1331      Mat3x3d Hmat = snap_->getHmat();
1332      Vector3d Hx = Hmat.getColumn(0);
# Line 969 | Line 1337 | namespace OpenMD {
1337      nCells_.y() = (int) ( Hy.length() )/ rList_;
1338      nCells_.z() = (int) ( Hz.length() )/ rList_;
1339  
1340 +    // handle small boxes where the cell offsets can end up repeating cells
1341 +    
1342 +    if (nCells_.x() < 3) doAllPairs = true;
1343 +    if (nCells_.y() < 3) doAllPairs = true;
1344 +    if (nCells_.z() < 3) doAllPairs = true;
1345 +
1346      Mat3x3d invHmat = snap_->getInvHmat();
1347      Vector3d rs, scaled, dr;
1348      Vector3i whichCell;
# Line 982 | Line 1356 | namespace OpenMD {
1356      cellList_.resize(nCtot);
1357   #endif
1358  
1359 +    if (!doAllPairs) {
1360   #ifdef IS_MPI
986    for (int i = 0; i < nGroupsInRow_; i++) {
987      rs = cgRowData.position[i];
1361  
1362 <      // scaled positions relative to the box vectors
1363 <      scaled = invHmat * rs;
1364 <
1365 <      // wrap the vector back into the unit box by subtracting integer box
1366 <      // numbers
1367 <      for (int j = 0; j < 3; j++) {
1368 <        scaled[j] -= roundMe(scaled[j]);
1369 <        scaled[j] += 0.5;
1362 >      for (int i = 0; i < nGroupsInRow_; i++) {
1363 >        rs = cgRowData.position[i];
1364 >        
1365 >        // scaled positions relative to the box vectors
1366 >        scaled = invHmat * rs;
1367 >        
1368 >        // wrap the vector back into the unit box by subtracting integer box
1369 >        // numbers
1370 >        for (int j = 0; j < 3; j++) {
1371 >          scaled[j] -= roundMe(scaled[j]);
1372 >          scaled[j] += 0.5;
1373 >          // Handle the special case when an object is exactly on the
1374 >          // boundary (a scaled coordinate of 1.0 is the same as
1375 >          // scaled coordinate of 0.0)
1376 >          if (scaled[j] >= 1.0) scaled[j] -= 1.0;
1377 >        }
1378 >        
1379 >        // find xyz-indices of cell that cutoffGroup is in.
1380 >        whichCell.x() = nCells_.x() * scaled.x();
1381 >        whichCell.y() = nCells_.y() * scaled.y();
1382 >        whichCell.z() = nCells_.z() * scaled.z();
1383 >        
1384 >        // find single index of this cell:
1385 >        cellIndex = Vlinear(whichCell, nCells_);
1386 >        
1387 >        // add this cutoff group to the list of groups in this cell;
1388 >        cellListRow_[cellIndex].push_back(i);
1389        }
1390 <    
1391 <      // find xyz-indices of cell that cutoffGroup is in.
1392 <      whichCell.x() = nCells_.x() * scaled.x();
1393 <      whichCell.y() = nCells_.y() * scaled.y();
1394 <      whichCell.z() = nCells_.z() * scaled.z();
1395 <
1396 <      // find single index of this cell:
1397 <      cellIndex = Vlinear(whichCell, nCells_);
1398 <
1399 <      // add this cutoff group to the list of groups in this cell;
1400 <      cellListRow_[cellIndex].push_back(i);
1401 <    }
1402 <
1403 <    for (int i = 0; i < nGroupsInCol_; i++) {
1404 <      rs = cgColData.position[i];
1405 <
1406 <      // scaled positions relative to the box vectors
1407 <      scaled = invHmat * rs;
1408 <
1409 <      // wrap the vector back into the unit box by subtracting integer box
1410 <      // numbers
1411 <      for (int j = 0; j < 3; j++) {
1412 <        scaled[j] -= roundMe(scaled[j]);
1413 <        scaled[j] += 0.5;
1390 >      for (int i = 0; i < nGroupsInCol_; i++) {
1391 >        rs = cgColData.position[i];
1392 >        
1393 >        // scaled positions relative to the box vectors
1394 >        scaled = invHmat * rs;
1395 >        
1396 >        // wrap the vector back into the unit box by subtracting integer box
1397 >        // numbers
1398 >        for (int j = 0; j < 3; j++) {
1399 >          scaled[j] -= roundMe(scaled[j]);
1400 >          scaled[j] += 0.5;
1401 >          // Handle the special case when an object is exactly on the
1402 >          // boundary (a scaled coordinate of 1.0 is the same as
1403 >          // scaled coordinate of 0.0)
1404 >          if (scaled[j] >= 1.0) scaled[j] -= 1.0;
1405 >        }
1406 >        
1407 >        // find xyz-indices of cell that cutoffGroup is in.
1408 >        whichCell.x() = nCells_.x() * scaled.x();
1409 >        whichCell.y() = nCells_.y() * scaled.y();
1410 >        whichCell.z() = nCells_.z() * scaled.z();
1411 >        
1412 >        // find single index of this cell:
1413 >        cellIndex = Vlinear(whichCell, nCells_);
1414 >        
1415 >        // add this cutoff group to the list of groups in this cell;
1416 >        cellListCol_[cellIndex].push_back(i);
1417        }
1418 <
1024 <      // find xyz-indices of cell that cutoffGroup is in.
1025 <      whichCell.x() = nCells_.x() * scaled.x();
1026 <      whichCell.y() = nCells_.y() * scaled.y();
1027 <      whichCell.z() = nCells_.z() * scaled.z();
1028 <
1029 <      // find single index of this cell:
1030 <      cellIndex = Vlinear(whichCell, nCells_);
1031 <
1032 <      // add this cutoff group to the list of groups in this cell;
1033 <      cellListCol_[cellIndex].push_back(i);
1034 <    }
1418 >    
1419   #else
1420 <    for (int i = 0; i < nGroups_; i++) {
1421 <      rs = snap_->cgData.position[i];
1422 <
1423 <      // scaled positions relative to the box vectors
1424 <      scaled = invHmat * rs;
1425 <
1426 <      // wrap the vector back into the unit box by subtracting integer box
1427 <      // numbers
1428 <      for (int j = 0; j < 3; j++) {
1429 <        scaled[j] -= roundMe(scaled[j]);
1430 <        scaled[j] += 0.5;
1420 >      for (int i = 0; i < nGroups_; i++) {
1421 >        rs = snap_->cgData.position[i];
1422 >        
1423 >        // scaled positions relative to the box vectors
1424 >        scaled = invHmat * rs;
1425 >        
1426 >        // wrap the vector back into the unit box by subtracting integer box
1427 >        // numbers
1428 >        for (int j = 0; j < 3; j++) {
1429 >          scaled[j] -= roundMe(scaled[j]);
1430 >          scaled[j] += 0.5;
1431 >          // Handle the special case when an object is exactly on the
1432 >          // boundary (a scaled coordinate of 1.0 is the same as
1433 >          // scaled coordinate of 0.0)
1434 >          if (scaled[j] >= 1.0) scaled[j] -= 1.0;
1435 >        }
1436 >        
1437 >        // find xyz-indices of cell that cutoffGroup is in.
1438 >        whichCell.x() = nCells_.x() * scaled.x();
1439 >        whichCell.y() = nCells_.y() * scaled.y();
1440 >        whichCell.z() = nCells_.z() * scaled.z();
1441 >        
1442 >        // find single index of this cell:
1443 >        cellIndex = Vlinear(whichCell, nCells_);
1444 >        
1445 >        // add this cutoff group to the list of groups in this cell;
1446 >        cellList_[cellIndex].push_back(i);
1447        }
1448  
1049      // find xyz-indices of cell that cutoffGroup is in.
1050      whichCell.x() = nCells_.x() * scaled.x();
1051      whichCell.y() = nCells_.y() * scaled.y();
1052      whichCell.z() = nCells_.z() * scaled.z();
1053
1054      // find single index of this cell:
1055      cellIndex = Vlinear(whichCell, nCells_);      
1056
1057      // add this cutoff group to the list of groups in this cell;
1058      cellList_[cellIndex].push_back(i);
1059    }
1449   #endif
1450  
1451 <    for (int m1z = 0; m1z < nCells_.z(); m1z++) {
1452 <      for (int m1y = 0; m1y < nCells_.y(); m1y++) {
1453 <        for (int m1x = 0; m1x < nCells_.x(); m1x++) {
1454 <          Vector3i m1v(m1x, m1y, m1z);
1455 <          int m1 = Vlinear(m1v, nCells_);
1067 <
1068 <          for (vector<Vector3i>::iterator os = cellOffsets_.begin();
1069 <               os != cellOffsets_.end(); ++os) {
1451 >      for (int m1z = 0; m1z < nCells_.z(); m1z++) {
1452 >        for (int m1y = 0; m1y < nCells_.y(); m1y++) {
1453 >          for (int m1x = 0; m1x < nCells_.x(); m1x++) {
1454 >            Vector3i m1v(m1x, m1y, m1z);
1455 >            int m1 = Vlinear(m1v, nCells_);
1456              
1457 <            Vector3i m2v = m1v + (*os);
1458 <            
1459 <            if (m2v.x() >= nCells_.x()) {
1460 <              m2v.x() = 0;          
1461 <            } else if (m2v.x() < 0) {
1076 <              m2v.x() = nCells_.x() - 1;
1077 <            }
1078 <            
1079 <            if (m2v.y() >= nCells_.y()) {
1080 <              m2v.y() = 0;          
1081 <            } else if (m2v.y() < 0) {
1082 <              m2v.y() = nCells_.y() - 1;
1083 <            }
1084 <            
1085 <            if (m2v.z() >= nCells_.z()) {
1086 <              m2v.z() = 0;          
1087 <            } else if (m2v.z() < 0) {
1088 <              m2v.z() = nCells_.z() - 1;
1089 <            }
1090 <            
1091 <            int m2 = Vlinear (m2v, nCells_);
1457 >            for (vector<Vector3i>::iterator os = cellOffsets_.begin();
1458 >                 os != cellOffsets_.end(); ++os) {
1459 >              
1460 >              Vector3i m2v = m1v + (*os);
1461 >            
1462  
1463 < #ifdef IS_MPI
1464 <            for (vector<int>::iterator j1 = cellListRow_[m1].begin();
1465 <                 j1 != cellListRow_[m1].end(); ++j1) {
1466 <              for (vector<int>::iterator j2 = cellListCol_[m2].begin();
1467 <                   j2 != cellListCol_[m2].end(); ++j2) {
1468 <                              
1469 <                // Always do this if we're in different cells or if
1470 <                // we're in the same cell and the global index of the
1471 <                // j2 cutoff group is less than the j1 cutoff group
1463 >              if (m2v.x() >= nCells_.x()) {
1464 >                m2v.x() = 0;          
1465 >              } else if (m2v.x() < 0) {
1466 >                m2v.x() = nCells_.x() - 1;
1467 >              }
1468 >              
1469 >              if (m2v.y() >= nCells_.y()) {
1470 >                m2v.y() = 0;          
1471 >              } else if (m2v.y() < 0) {
1472 >                m2v.y() = nCells_.y() - 1;
1473 >              }
1474 >              
1475 >              if (m2v.z() >= nCells_.z()) {
1476 >                m2v.z() = 0;          
1477 >              } else if (m2v.z() < 0) {
1478 >                m2v.z() = nCells_.z() - 1;
1479 >              }
1480  
1481 <                if (m2 != m1 || cgColToGlobal[(*j2)] < cgRowToGlobal[(*j1)]) {
1481 >              int m2 = Vlinear (m2v, nCells_);
1482 >              
1483 > #ifdef IS_MPI
1484 >              for (vector<int>::iterator j1 = cellListRow_[m1].begin();
1485 >                   j1 != cellListRow_[m1].end(); ++j1) {
1486 >                for (vector<int>::iterator j2 = cellListCol_[m2].begin();
1487 >                     j2 != cellListCol_[m2].end(); ++j2) {
1488 >                  
1489 >                  // In parallel, we need to visit *all* pairs of row
1490 >                  // & column indicies and will divide labor in the
1491 >                  // force evaluation later.
1492                    dr = cgColData.position[(*j2)] - cgRowData.position[(*j1)];
1493                    snap_->wrapVector(dr);
1494                    cuts = getGroupCutoffs( (*j1), (*j2) );
1495                    if (dr.lengthSquare() < cuts.third) {
1496                      neighborList.push_back(make_pair((*j1), (*j2)));
1497 <                  }
1497 >                  }                  
1498                  }
1499                }
1112            }
1500   #else
1501 +              for (vector<int>::iterator j1 = cellList_[m1].begin();
1502 +                   j1 != cellList_[m1].end(); ++j1) {
1503 +                for (vector<int>::iterator j2 = cellList_[m2].begin();
1504 +                     j2 != cellList_[m2].end(); ++j2) {
1505 +    
1506 +                  // Always do this if we're in different cells or if
1507 +                  // we're in the same cell and the global index of
1508 +                  // the j2 cutoff group is greater than or equal to
1509 +                  // the j1 cutoff group.  Note that Rappaport's code
1510 +                  // has a "less than" conditional here, but that
1511 +                  // deals with atom-by-atom computation.  OpenMD
1512 +                  // allows atoms within a single cutoff group to
1513 +                  // interact with each other.
1514  
1115            for (vector<int>::iterator j1 = cellList_[m1].begin();
1116                 j1 != cellList_[m1].end(); ++j1) {
1117              for (vector<int>::iterator j2 = cellList_[m2].begin();
1118                   j2 != cellList_[m2].end(); ++j2) {
1515  
1120                // Always do this if we're in different cells or if
1121                // we're in the same cell and the global index of the
1122                // j2 cutoff group is less than the j1 cutoff group
1516  
1517 <                if (m2 != m1 || (*j2) < (*j1)) {
1518 <                  dr = snap_->cgData.position[(*j2)] - snap_->cgData.position[(*j1)];
1519 <                  snap_->wrapVector(dr);
1520 <                  cuts = getGroupCutoffs( (*j1), (*j2) );
1521 <                  if (dr.lengthSquare() < cuts.third) {
1522 <                    neighborList.push_back(make_pair((*j1), (*j2)));
1517 >                  if (m2 != m1 || (*j2) >= (*j1) ) {
1518 >
1519 >                    dr = snap_->cgData.position[(*j2)] - snap_->cgData.position[(*j1)];
1520 >                    snap_->wrapVector(dr);
1521 >                    cuts = getGroupCutoffs( (*j1), (*j2) );
1522 >                    if (dr.lengthSquare() < cuts.third) {
1523 >                      neighborList.push_back(make_pair((*j1), (*j2)));
1524 >                    }
1525                    }
1526                  }
1527                }
1133            }
1528   #endif
1529 +            }
1530            }
1531          }
1532        }
1533 +    } else {
1534 +      // branch to do all cutoff group pairs
1535 + #ifdef IS_MPI
1536 +      for (int j1 = 0; j1 < nGroupsInRow_; j1++) {
1537 +        for (int j2 = 0; j2 < nGroupsInCol_; j2++) {    
1538 +          dr = cgColData.position[j2] - cgRowData.position[j1];
1539 +          snap_->wrapVector(dr);
1540 +          cuts = getGroupCutoffs( j1, j2 );
1541 +          if (dr.lengthSquare() < cuts.third) {
1542 +            neighborList.push_back(make_pair(j1, j2));
1543 +          }
1544 +        }
1545 +      }      
1546 + #else
1547 +      // include all groups here.
1548 +      for (int j1 = 0; j1 < nGroups_; j1++) {
1549 +        // include self group interactions j2 == j1
1550 +        for (int j2 = j1; j2 < nGroups_; j2++) {
1551 +          dr = snap_->cgData.position[j2] - snap_->cgData.position[j1];
1552 +          snap_->wrapVector(dr);
1553 +          cuts = getGroupCutoffs( j1, j2 );
1554 +          if (dr.lengthSquare() < cuts.third) {
1555 +            neighborList.push_back(make_pair(j1, j2));
1556 +          }
1557 +        }    
1558 +      }
1559 + #endif
1560      }
1561 <    
1561 >      
1562      // save the local cutoff group positions for the check that is
1563      // done on each loop:
1564      saved_CG_positions_.clear();
1565      for (int i = 0; i < nGroups_; i++)
1566        saved_CG_positions_.push_back(snap_->cgData.position[i]);
1567 <  
1567 >    
1568      return neighborList;
1569    }
1570   } //end namespace OpenMD

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines