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

Comparing:
branches/development/src/parallel/ForceDecomposition.cpp (file contents), Revision 1538 by chuckv, Tue Jan 11 18:58:12 2011 UTC vs.
branches/development/src/parallel/ForceMatrixDecomposition.cpp (file contents), Revision 1590 by gezelter, Mon Jul 11 01:39:49 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 +  /**
51 +   * distributeInitialData is essentially a copy of the older fortran
52 +   * SimulationSetup
53 +   */
54 +  
55 +  void ForceMatrixDecomposition::distributeInitialData() {
56 +    snap_ = sman_->getCurrentSnapshot();
57 +    storageLayout_ = sman_->getStorageLayout();
58 +    ff_ = info_->getForceField();
59 +    nLocal_ = snap_->getNumberOfAtoms();
60 +    
61 +    nGroups_ = info_->getNLocalCutoffGroups();
62 +    // gather the information for atomtype IDs (atids):
63 +    idents = info_->getIdentArray();
64 +    AtomLocalToGlobal = info_->getGlobalAtomIndices();
65 +    cgLocalToGlobal = info_->getGlobalGroupIndices();
66 +    vector<int> globalGroupMembership = info_->getGlobalGroupMembership();
67  
68 < /*  -*- c++ -*-  */
69 < #include "config.h"
70 < #include <stdlib.h>
68 >    massFactors = info_->getMassFactors();
69 >
70 >    PairList* excludes = info_->getExcludedInteractions();
71 >    PairList* oneTwo = info_->getOneTwoInteractions();
72 >    PairList* oneThree = info_->getOneThreeInteractions();
73 >    PairList* oneFour = info_->getOneFourInteractions();
74 >
75   #ifdef IS_MPI
76 < #include <mpi.h>
77 < #endif
76 >
77 >    AtomCommIntRow = new Communicator<Row,int>(nLocal_);
78 >    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_);
82  
83 < #include <iostream>
84 < #include <vector>
85 < #include <algorithm>
86 < #include <cmath>
87 < #include "parallel/ForceDecomposition.hpp"
83 >    AtomCommIntColumn = new Communicator<Column,int>(nLocal_);
84 >    AtomCommRealColumn = new Communicator<Column,RealType>(nLocal_);
85 >    AtomCommVectorColumn = new Communicator<Column,Vector3d>(nLocal_);
86 >    AtomCommMatrixColumn = new Communicator<Column,Mat3x3d>(nLocal_);
87 >    AtomCommPotColumn = new Communicator<Column,potVec>(nLocal_);
88  
89 +    cgCommIntRow = new Communicator<Row,int>(nGroups_);
90 +    cgCommVectorRow = new Communicator<Row,Vector3d>(nGroups_);
91 +    cgCommIntColumn = new Communicator<Column,int>(nGroups_);
92 +    cgCommVectorColumn = new Communicator<Column,Vector3d>(nGroups_);
93  
94 < using namespace std;
95 < using namespace OpenMD;
94 >    nAtomsInRow_ = AtomCommIntRow->getSize();
95 >    nAtomsInCol_ = AtomCommIntColumn->getSize();
96 >    nGroupsInRow_ = cgCommIntRow->getSize();
97 >    nGroupsInCol_ = cgCommIntColumn->getSize();
98  
99 < //__static
100 < #ifdef IS_MPI
101 < static vector<MPI:Comm> communictors;
102 < #endif
99 >    // Modify the data storage objects with the correct layouts and sizes:
100 >    atomRowData.resize(nAtomsInRow_);
101 >    atomRowData.setStorageLayout(storageLayout_);
102 >    atomColData.resize(nAtomsInCol_);
103 >    atomColData.setStorageLayout(storageLayout_);
104 >    cgRowData.resize(nGroupsInRow_);
105 >    cgRowData.setStorageLayout(DataStorage::dslPosition);
106 >    cgColData.resize(nGroupsInCol_);
107 >    cgColData.setStorageLayout(DataStorage::dslPosition);
108 >        
109 >    identsRow.resize(nAtomsInRow_);
110 >    identsCol.resize(nAtomsInCol_);
111 >    
112 >    AtomCommIntRow->gather(idents, identsRow);
113 >    AtomCommIntColumn->gather(idents, identsCol);
114 >    
115 >    // allocate memory for the parallel objects
116 >    AtomRowToGlobal.resize(nAtomsInRow_);
117 >    AtomColToGlobal.resize(nAtomsInCol_);
118 >    cgRowToGlobal.resize(nGroupsInRow_);
119 >    cgColToGlobal.resize(nGroupsInCol_);
120 >    massFactorsRow.resize(nAtomsInRow_);
121 >    massFactorsCol.resize(nAtomsInCol_);
122 >    pot_row.resize(nAtomsInRow_);
123 >    pot_col.resize(nAtomsInCol_);
124  
125 < //____ MPITypeTraits
126 < template<typename T>
127 < struct MPITypeTraits;
125 >    AtomCommIntRow->gather(AtomLocalToGlobal, AtomRowToGlobal);
126 >    AtomCommIntColumn->gather(AtomLocalToGlobal, AtomColToGlobal);
127 >    
128 >    cgCommIntRow->gather(cgLocalToGlobal, cgRowToGlobal);
129 >    cgCommIntColumn->gather(cgLocalToGlobal, cgColToGlobal);
130  
131 < #ifdef IS_MPI
132 < template<>
79 < struct MPITypeTraits<RealType> {
80 <  static const MPI::Datatype datatype;
81 < };
82 < const MPI_Datatype MPITypeTraits<RealType>::datatype = MY_MPI_REAL;
131 >    AtomCommRealRow->gather(massFactors, massFactorsRow);
132 >    AtomCommRealColumn->gather(massFactors, massFactorsCol);
133  
134 < template<>
135 < struct MPITypeTraits<int> {
136 <  static const MPI::Datatype datatype;
137 < };
138 < const MPI::Datatype MPITypeTraits<int>::datatype = MPI_INT;
139 < #endif
134 >    groupListRow_.clear();
135 >    groupListRow_.resize(nGroupsInRow_);
136 >    for (int i = 0; i < nGroupsInRow_; i++) {
137 >      int gid = cgRowToGlobal[i];
138 >      for (int j = 0; j < nAtomsInRow_; j++) {
139 >        int aid = AtomRowToGlobal[j];
140 >        if (globalGroupMembership[aid] == gid)
141 >          groupListRow_[i].push_back(j);
142 >      }      
143 >    }
144  
145 < /**
146 < * Constructor for ForceDecomposition Parallel Decomposition Method
147 < * Will try to construct a symmetric grid of processors. Ideally, the
148 < * number of processors will be a square ex: 4, 9, 16, 25.
149 < *
150 < */
145 >    groupListCol_.clear();
146 >    groupListCol_.resize(nGroupsInCol_);
147 >    for (int i = 0; i < nGroupsInCol_; i++) {
148 >      int gid = cgColToGlobal[i];
149 >      for (int j = 0; j < nAtomsInCol_; j++) {
150 >        int aid = AtomColToGlobal[j];
151 >        if (globalGroupMembership[aid] == gid)
152 >          groupListCol_[i].push_back(j);
153 >      }      
154 >    }
155  
156 < ForceDecomposition::ForceDecomposition() {
156 >    excludesForAtom.clear();
157 >    excludesForAtom.resize(nAtomsInRow_);
158 >    toposForAtom.clear();
159 >    toposForAtom.resize(nAtomsInRow_);
160 >    topoDist.clear();
161 >    topoDist.resize(nAtomsInRow_);
162 >    for (int i = 0; i < nAtomsInRow_; i++) {
163 >      int iglob = AtomRowToGlobal[i];
164  
165 < #ifdef IS_MPI
166 <  int nProcs = MPI::COMM_WORLD.Get_size();
167 <  int worldRank = MPI::COMM_WORLD.Get_rank();
165 >      for (int j = 0; j < nAtomsInCol_; j++) {
166 >        int jglob = AtomColToGlobal[j];
167 >
168 >        if (excludes->hasPair(iglob, jglob))
169 >          excludesForAtom[i].push_back(j);      
170 >        
171 >        if (oneTwo->hasPair(iglob, jglob)) {
172 >          toposForAtom[i].push_back(j);
173 >          topoDist[i].push_back(1);
174 >        } else {
175 >          if (oneThree->hasPair(iglob, jglob)) {
176 >            toposForAtom[i].push_back(j);
177 >            topoDist[i].push_back(2);
178 >          } else {
179 >            if (oneFour->hasPair(iglob, jglob)) {
180 >              toposForAtom[i].push_back(j);
181 >              topoDist[i].push_back(3);
182 >            }
183 >          }
184 >        }
185 >      }      
186 >    }
187 >
188   #endif
189  
190 <  // First time through, construct column stride.
191 <  if (communicators.size() == 0)
192 <  {
193 <    int nColumnsMax = (int) round(sqrt((float) nProcs));
194 <    for (int i = 0; i < nProcs; ++i)
195 <    {
196 <      if (nProcs%i==0) nColumns=i;
190 >    groupList_.clear();
191 >    groupList_.resize(nGroups_);
192 >    for (int i = 0; i < nGroups_; i++) {
193 >      int gid = cgLocalToGlobal[i];
194 >      for (int j = 0; j < nLocal_; j++) {
195 >        int aid = AtomLocalToGlobal[j];
196 >        if (globalGroupMembership[aid] == gid) {
197 >          groupList_[i].push_back(j);
198 >        }
199 >      }      
200      }
201  
202 <    int nRows = nProcs/nColumns;    
203 <    myRank_ = (int) worldRank%nColumns;
204 <  }
205 <  else
206 <  {
207 <    myRank_ = myRank/nColumns;
120 <  }
121 <  MPI::Comm newComm = MPI:COMM_WORLD.Split(myRank_,0);
122 <  
123 <  isColumn_ = false;
124 <  
125 < }
202 >    excludesForAtom.clear();
203 >    excludesForAtom.resize(nLocal_);
204 >    toposForAtom.clear();
205 >    toposForAtom.resize(nLocal_);
206 >    topoDist.clear();
207 >    topoDist.resize(nLocal_);
208  
209 < ForceDecomposition::gather(sendbuf, receivebuf){
210 <  communicators(myIndex_).Allgatherv();
129 < }
209 >    for (int i = 0; i < nLocal_; i++) {
210 >      int iglob = AtomLocalToGlobal[i];
211  
212 +      for (int j = 0; j < nLocal_; j++) {
213 +        int jglob = AtomLocalToGlobal[j];
214  
215 +        if (excludes->hasPair(iglob, jglob))
216 +          excludesForAtom[i].push_back(j);              
217 +        
218 +        if (oneTwo->hasPair(iglob, jglob)) {
219 +          toposForAtom[i].push_back(j);
220 +          topoDist[i].push_back(1);
221 +        } else {
222 +          if (oneThree->hasPair(iglob, jglob)) {
223 +            toposForAtom[i].push_back(j);
224 +            topoDist[i].push_back(2);
225 +          } else {
226 +            if (oneFour->hasPair(iglob, jglob)) {
227 +              toposForAtom[i].push_back(j);
228 +              topoDist[i].push_back(3);
229 +            }
230 +          }
231 +        }
232 +      }      
233 +    }
234 +    
235 +    createGtypeCutoffMap();
236  
237 < ForceDecomposition::scatter(sbuffer, rbuffer){
238 <  communicators(myIndex_).Reduce_scatter(sbuffer, recevbuf. recvcounts, MPI::DOUBLE, MPI::SUM);
239 < }
237 >  }
238 >  
239 >  void ForceMatrixDecomposition::createGtypeCutoffMap() {
240 >    
241 >    RealType tol = 1e-6;
242 >    RealType rc;
243 >    int atid;
244 >    set<AtomType*> atypes = info_->getSimulatedAtomTypes();
245 >    map<int, RealType> atypeCutoff;
246 >      
247 >    for (set<AtomType*>::iterator at = atypes.begin();
248 >         at != atypes.end(); ++at){
249 >      atid = (*at)->getIdent();
250 >      if (userChoseCutoff_)
251 >        atypeCutoff[atid] = userCutoff_;
252 >      else
253 >        atypeCutoff[atid] = interactionMan_->getSuggestedCutoffRadius(*at);
254 >    }
255 >
256 >    vector<RealType> gTypeCutoffs;
257 >    // first we do a single loop over the cutoff groups to find the
258 >    // largest cutoff for any atypes present in this group.
259 > #ifdef IS_MPI
260 >    vector<RealType> groupCutoffRow(nGroupsInRow_, 0.0);
261 >    groupRowToGtype.resize(nGroupsInRow_);
262 >    for (int cg1 = 0; cg1 < nGroupsInRow_; cg1++) {
263 >      vector<int> atomListRow = getAtomsInGroupRow(cg1);
264 >      for (vector<int>::iterator ia = atomListRow.begin();
265 >           ia != atomListRow.end(); ++ia) {            
266 >        int atom1 = (*ia);
267 >        atid = identsRow[atom1];
268 >        if (atypeCutoff[atid] > groupCutoffRow[cg1]) {
269 >          groupCutoffRow[cg1] = atypeCutoff[atid];
270 >        }
271 >      }
272 >
273 >      bool gTypeFound = false;
274 >      for (int gt = 0; gt < gTypeCutoffs.size(); gt++) {
275 >        if (abs(groupCutoffRow[cg1] - gTypeCutoffs[gt]) < tol) {
276 >          groupRowToGtype[cg1] = gt;
277 >          gTypeFound = true;
278 >        }
279 >      }
280 >      if (!gTypeFound) {
281 >        gTypeCutoffs.push_back( groupCutoffRow[cg1] );
282 >        groupRowToGtype[cg1] = gTypeCutoffs.size() - 1;
283 >      }
284 >      
285 >    }
286 >    vector<RealType> groupCutoffCol(nGroupsInCol_, 0.0);
287 >    groupColToGtype.resize(nGroupsInCol_);
288 >    for (int cg2 = 0; cg2 < nGroupsInCol_; cg2++) {
289 >      vector<int> atomListCol = getAtomsInGroupColumn(cg2);
290 >      for (vector<int>::iterator jb = atomListCol.begin();
291 >           jb != atomListCol.end(); ++jb) {            
292 >        int atom2 = (*jb);
293 >        atid = identsCol[atom2];
294 >        if (atypeCutoff[atid] > groupCutoffCol[cg2]) {
295 >          groupCutoffCol[cg2] = atypeCutoff[atid];
296 >        }
297 >      }
298 >      bool gTypeFound = false;
299 >      for (int gt = 0; gt < gTypeCutoffs.size(); gt++) {
300 >        if (abs(groupCutoffCol[cg2] - gTypeCutoffs[gt]) < tol) {
301 >          groupColToGtype[cg2] = gt;
302 >          gTypeFound = true;
303 >        }
304 >      }
305 >      if (!gTypeFound) {
306 >        gTypeCutoffs.push_back( groupCutoffCol[cg2] );
307 >        groupColToGtype[cg2] = gTypeCutoffs.size() - 1;
308 >      }
309 >    }
310 > #else
311  
312 +    vector<RealType> groupCutoff(nGroups_, 0.0);
313 +    groupToGtype.resize(nGroups_);
314 +    for (int cg1 = 0; cg1 < nGroups_; cg1++) {
315  
316 +      groupCutoff[cg1] = 0.0;
317 +      vector<int> atomList = getAtomsInGroupRow(cg1);
318 +
319 +      for (vector<int>::iterator ia = atomList.begin();
320 +           ia != atomList.end(); ++ia) {            
321 +        int atom1 = (*ia);
322 +        atid = idents[atom1];
323 +        if (atypeCutoff[atid] > groupCutoff[cg1]) {
324 +          groupCutoff[cg1] = atypeCutoff[atid];
325 +        }
326 +      }
327 +
328 +      bool gTypeFound = false;
329 +      for (int gt = 0; gt < gTypeCutoffs.size(); gt++) {
330 +        if (abs(groupCutoff[cg1] - gTypeCutoffs[gt]) < tol) {
331 +          groupToGtype[cg1] = gt;
332 +          gTypeFound = true;
333 +        }
334 +      }
335 +      if (!gTypeFound) {
336 +        gTypeCutoffs.push_back( groupCutoff[cg1] );
337 +        groupToGtype[cg1] = gTypeCutoffs.size() - 1;
338 +      }      
339 +    }
340 + #endif
341 +
342 +    // Now we find the maximum group cutoff value present in the simulation
343 +
344 +    RealType groupMax = *max_element(gTypeCutoffs.begin(),
345 +                                     gTypeCutoffs.end());
346 +
347 + #ifdef IS_MPI
348 +    MPI::COMM_WORLD.Allreduce(&groupMax, &groupMax, 1, MPI::REALTYPE,
349 +                              MPI::MAX);
350 + #endif
351 +    
352 +    RealType tradRcut = groupMax;
353 +
354 +    for (int i = 0; i < gTypeCutoffs.size();  i++) {
355 +      for (int j = 0; j < gTypeCutoffs.size();  j++) {      
356 +        RealType thisRcut;
357 +        switch(cutoffPolicy_) {
358 +        case TRADITIONAL:
359 +          thisRcut = tradRcut;
360 +          break;
361 +        case MIX:
362 +          thisRcut = 0.5 * (gTypeCutoffs[i] + gTypeCutoffs[j]);
363 +          break;
364 +        case MAX:
365 +          thisRcut = max(gTypeCutoffs[i], gTypeCutoffs[j]);
366 +          break;
367 +        default:
368 +          sprintf(painCave.errMsg,
369 +                  "ForceMatrixDecomposition::createGtypeCutoffMap "
370 +                  "hit an unknown cutoff policy!\n");
371 +          painCave.severity = OPENMD_ERROR;
372 +          painCave.isFatal = 1;
373 +          simError();
374 +          break;
375 +        }
376 +
377 +        pair<int,int> key = make_pair(i,j);
378 +        gTypeCutoffMap[key].first = thisRcut;
379 +
380 +        if (thisRcut > largestRcut_) largestRcut_ = thisRcut;
381 +
382 +        gTypeCutoffMap[key].second = thisRcut*thisRcut;
383 +        
384 +        gTypeCutoffMap[key].third = pow(thisRcut + skinThickness_, 2);
385 +
386 +        // sanity check
387 +        
388 +        if (userChoseCutoff_) {
389 +          if (abs(gTypeCutoffMap[key].first - userCutoff_) > 0.0001) {
390 +            sprintf(painCave.errMsg,
391 +                    "ForceMatrixDecomposition::createGtypeCutoffMap "
392 +                    "user-specified rCut (%lf) does not match computed group Cutoff\n", userCutoff_);
393 +            painCave.severity = OPENMD_ERROR;
394 +            painCave.isFatal = 1;
395 +            simError();            
396 +          }
397 +        }
398 +      }
399 +    }
400 +  }
401 +
402 +
403 +  groupCutoffs ForceMatrixDecomposition::getGroupCutoffs(int cg1, int cg2) {
404 +    int i, j;  
405 + #ifdef IS_MPI
406 +    i = groupRowToGtype[cg1];
407 +    j = groupColToGtype[cg2];
408 + #else
409 +    i = groupToGtype[cg1];
410 +    j = groupToGtype[cg2];
411 + #endif    
412 +    return gTypeCutoffMap[make_pair(i,j)];
413 +  }
414 +
415 +  int ForceMatrixDecomposition::getTopologicalDistance(int atom1, int atom2) {
416 +    for (int j = 0; j < toposForAtom[atom1].size(); j++) {
417 +      if (toposForAtom[atom1][j] == atom2)
418 +        return topoDist[atom1][j];
419 +    }
420 +    return 0;
421 +  }
422 +
423 +  void ForceMatrixDecomposition::zeroWorkArrays() {
424 +    pairwisePot = 0.0;
425 +    embeddingPot = 0.0;
426 +
427 + #ifdef IS_MPI
428 +    if (storageLayout_ & DataStorage::dslForce) {
429 +      fill(atomRowData.force.begin(), atomRowData.force.end(), V3Zero);
430 +      fill(atomColData.force.begin(), atomColData.force.end(), V3Zero);
431 +    }
432 +
433 +    if (storageLayout_ & DataStorage::dslTorque) {
434 +      fill(atomRowData.torque.begin(), atomRowData.torque.end(), V3Zero);
435 +      fill(atomColData.torque.begin(), atomColData.torque.end(), V3Zero);
436 +    }
437 +    
438 +    fill(pot_row.begin(), pot_row.end(),
439 +         Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
440 +
441 +    fill(pot_col.begin(), pot_col.end(),
442 +         Vector<RealType, N_INTERACTION_FAMILIES> (0.0));  
443 +
444 +    if (storageLayout_ & DataStorage::dslParticlePot) {    
445 +      fill(atomRowData.particlePot.begin(), atomRowData.particlePot.end(),
446 +           0.0);
447 +      fill(atomColData.particlePot.begin(), atomColData.particlePot.end(),
448 +           0.0);
449 +    }
450 +
451 +    if (storageLayout_ & DataStorage::dslDensity) {      
452 +      fill(atomRowData.density.begin(), atomRowData.density.end(), 0.0);
453 +      fill(atomColData.density.begin(), atomColData.density.end(), 0.0);
454 +    }
455 +
456 +    if (storageLayout_ & DataStorage::dslFunctional) {  
457 +      fill(atomRowData.functional.begin(), atomRowData.functional.end(),
458 +           0.0);
459 +      fill(atomColData.functional.begin(), atomColData.functional.end(),
460 +           0.0);
461 +    }
462 +
463 +    if (storageLayout_ & DataStorage::dslFunctionalDerivative) {      
464 +      fill(atomRowData.functionalDerivative.begin(),
465 +           atomRowData.functionalDerivative.end(), 0.0);
466 +      fill(atomColData.functionalDerivative.begin(),
467 +           atomColData.functionalDerivative.end(), 0.0);
468 +    }
469 +
470 +    if (storageLayout_ & DataStorage::dslSkippedCharge) {      
471 +      fill(atomRowData.skippedCharge.begin(),
472 +           atomRowData.skippedCharge.end(), 0.0);
473 +      fill(atomColData.skippedCharge.begin(),
474 +           atomColData.skippedCharge.end(), 0.0);
475 +    }
476 +
477 + #endif
478 +    // even in parallel, we need to zero out the local arrays:
479 +
480 +    if (storageLayout_ & DataStorage::dslParticlePot) {      
481 +      fill(snap_->atomData.particlePot.begin(),
482 +           snap_->atomData.particlePot.end(), 0.0);
483 +    }
484 +    
485 +    if (storageLayout_ & DataStorage::dslDensity) {      
486 +      fill(snap_->atomData.density.begin(),
487 +           snap_->atomData.density.end(), 0.0);
488 +    }
489 +    if (storageLayout_ & DataStorage::dslFunctional) {
490 +      fill(snap_->atomData.functional.begin(),
491 +           snap_->atomData.functional.end(), 0.0);
492 +    }
493 +    if (storageLayout_ & DataStorage::dslFunctionalDerivative) {      
494 +      fill(snap_->atomData.functionalDerivative.begin(),
495 +           snap_->atomData.functionalDerivative.end(), 0.0);
496 +    }
497 +    if (storageLayout_ & DataStorage::dslSkippedCharge) {      
498 +      fill(snap_->atomData.skippedCharge.begin(),
499 +           snap_->atomData.skippedCharge.end(), 0.0);
500 +    }
501 +    
502 +  }
503 +
504 +
505 +  void ForceMatrixDecomposition::distributeData()  {
506 +    snap_ = sman_->getCurrentSnapshot();
507 +    storageLayout_ = sman_->getStorageLayout();
508 + #ifdef IS_MPI
509 +    
510 +    // gather up the atomic positions
511 +    AtomCommVectorRow->gather(snap_->atomData.position,
512 +                              atomRowData.position);
513 +    AtomCommVectorColumn->gather(snap_->atomData.position,
514 +                                 atomColData.position);
515 +    
516 +    // gather up the cutoff group positions
517 +    cgCommVectorRow->gather(snap_->cgData.position,
518 +                            cgRowData.position);
519 +    cgCommVectorColumn->gather(snap_->cgData.position,
520 +                               cgColData.position);
521 +    
522 +    // if needed, gather the atomic rotation matrices
523 +    if (storageLayout_ & DataStorage::dslAmat) {
524 +      AtomCommMatrixRow->gather(snap_->atomData.aMat,
525 +                                atomRowData.aMat);
526 +      AtomCommMatrixColumn->gather(snap_->atomData.aMat,
527 +                                   atomColData.aMat);
528 +    }
529 +    
530 +    // if needed, gather the atomic eletrostatic frames
531 +    if (storageLayout_ & DataStorage::dslElectroFrame) {
532 +      AtomCommMatrixRow->gather(snap_->atomData.electroFrame,
533 +                                atomRowData.electroFrame);
534 +      AtomCommMatrixColumn->gather(snap_->atomData.electroFrame,
535 +                                   atomColData.electroFrame);
536 +    }
537 +
538 + #endif      
539 +  }
540 +  
541 +  /* collects information obtained during the pre-pair loop onto local
542 +   * data structures.
543 +   */
544 +  void ForceMatrixDecomposition::collectIntermediateData() {
545 +    snap_ = sman_->getCurrentSnapshot();
546 +    storageLayout_ = sman_->getStorageLayout();
547 + #ifdef IS_MPI
548 +    
549 +    if (storageLayout_ & DataStorage::dslDensity) {
550 +      
551 +      AtomCommRealRow->scatter(atomRowData.density,
552 +                               snap_->atomData.density);
553 +      
554 +      int n = snap_->atomData.density.size();
555 +      vector<RealType> rho_tmp(n, 0.0);
556 +      AtomCommRealColumn->scatter(atomColData.density, rho_tmp);
557 +      for (int i = 0; i < n; i++)
558 +        snap_->atomData.density[i] += rho_tmp[i];
559 +    }
560 + #endif
561 +  }
562 +
563 +  /*
564 +   * redistributes information obtained during the pre-pair loop out to
565 +   * row and column-indexed data structures
566 +   */
567 +  void ForceMatrixDecomposition::distributeIntermediateData() {
568 +    snap_ = sman_->getCurrentSnapshot();
569 +    storageLayout_ = sman_->getStorageLayout();
570 + #ifdef IS_MPI
571 +    if (storageLayout_ & DataStorage::dslFunctional) {
572 +      AtomCommRealRow->gather(snap_->atomData.functional,
573 +                              atomRowData.functional);
574 +      AtomCommRealColumn->gather(snap_->atomData.functional,
575 +                                 atomColData.functional);
576 +    }
577 +    
578 +    if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
579 +      AtomCommRealRow->gather(snap_->atomData.functionalDerivative,
580 +                              atomRowData.functionalDerivative);
581 +      AtomCommRealColumn->gather(snap_->atomData.functionalDerivative,
582 +                                 atomColData.functionalDerivative);
583 +    }
584 + #endif
585 +  }
586 +  
587 +  
588 +  void ForceMatrixDecomposition::collectData() {
589 +    snap_ = sman_->getCurrentSnapshot();
590 +    storageLayout_ = sman_->getStorageLayout();
591 + #ifdef IS_MPI    
592 +    int n = snap_->atomData.force.size();
593 +    vector<Vector3d> frc_tmp(n, V3Zero);
594 +    
595 +    AtomCommVectorRow->scatter(atomRowData.force, frc_tmp);
596 +    for (int i = 0; i < n; i++) {
597 +      snap_->atomData.force[i] += frc_tmp[i];
598 +      frc_tmp[i] = 0.0;
599 +    }
600 +    
601 +    AtomCommVectorColumn->scatter(atomColData.force, frc_tmp);
602 +    for (int i = 0; i < n; i++)
603 +      snap_->atomData.force[i] += frc_tmp[i];
604 +        
605 +    if (storageLayout_ & DataStorage::dslTorque) {
606 +
607 +      int nt = snap_->atomData.torque.size();
608 +      vector<Vector3d> trq_tmp(nt, V3Zero);
609 +
610 +      AtomCommVectorRow->scatter(atomRowData.torque, trq_tmp);
611 +      for (int i = 0; i < nt; i++) {
612 +        snap_->atomData.torque[i] += trq_tmp[i];
613 +        trq_tmp[i] = 0.0;
614 +      }
615 +      
616 +      AtomCommVectorColumn->scatter(atomColData.torque, trq_tmp);
617 +      for (int i = 0; i < nt; i++)
618 +        snap_->atomData.torque[i] += trq_tmp[i];
619 +    }
620 +
621 +    if (storageLayout_ & DataStorage::dslSkippedCharge) {
622 +
623 +      int ns = snap_->atomData.skippedCharge.size();
624 +      vector<RealType> skch_tmp(ns, 0.0);
625 +
626 +      AtomCommRealRow->scatter(atomRowData.skippedCharge, skch_tmp);
627 +      for (int i = 0; i < ns; i++) {
628 +        snap_->atomData.skippedCharge[i] += skch_tmp[i];
629 +        skch_tmp[i] = 0.0;
630 +      }
631 +      
632 +      AtomCommRealColumn->scatter(atomColData.skippedCharge, skch_tmp);
633 +      for (int i = 0; i < ns; i++)
634 +        snap_->atomData.skippedCharge[i] += skch_tmp[i];
635 +    }
636 +    
637 +    nLocal_ = snap_->getNumberOfAtoms();
638 +
639 +    vector<potVec> pot_temp(nLocal_,
640 +                            Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
641 +
642 +    // scatter/gather pot_row into the members of my column
643 +          
644 +    AtomCommPotRow->scatter(pot_row, pot_temp);
645 +
646 +    for (int ii = 0;  ii < pot_temp.size(); ii++ )
647 +      pairwisePot += pot_temp[ii];
648 +    
649 +    fill(pot_temp.begin(), pot_temp.end(),
650 +         Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
651 +      
652 +    AtomCommPotColumn->scatter(pot_col, pot_temp);    
653 +    
654 +    for (int ii = 0;  ii < pot_temp.size(); ii++ )
655 +      pairwisePot += pot_temp[ii];    
656 + #endif
657 +
658 +  }
659 +
660 +  int ForceMatrixDecomposition::getNAtomsInRow() {  
661 + #ifdef IS_MPI
662 +    return nAtomsInRow_;
663 + #else
664 +    return nLocal_;
665 + #endif
666 +  }
667 +
668 +  /**
669 +   * returns the list of atoms belonging to this group.  
670 +   */
671 +  vector<int> ForceMatrixDecomposition::getAtomsInGroupRow(int cg1){
672 + #ifdef IS_MPI
673 +    return groupListRow_[cg1];
674 + #else
675 +    return groupList_[cg1];
676 + #endif
677 +  }
678 +
679 +  vector<int> ForceMatrixDecomposition::getAtomsInGroupColumn(int cg2){
680 + #ifdef IS_MPI
681 +    return groupListCol_[cg2];
682 + #else
683 +    return groupList_[cg2];
684 + #endif
685 +  }
686 +  
687 +  Vector3d ForceMatrixDecomposition::getIntergroupVector(int cg1, int cg2){
688 +    Vector3d d;
689 +    
690 + #ifdef IS_MPI
691 +    d = cgColData.position[cg2] - cgRowData.position[cg1];
692 + #else
693 +    d = snap_->cgData.position[cg2] - snap_->cgData.position[cg1];
694 + #endif
695 +    
696 +    snap_->wrapVector(d);
697 +    return d;    
698 +  }
699 +
700 +
701 +  Vector3d ForceMatrixDecomposition::getAtomToGroupVectorRow(int atom1, int cg1){
702 +
703 +    Vector3d d;
704 +    
705 + #ifdef IS_MPI
706 +    d = cgRowData.position[cg1] - atomRowData.position[atom1];
707 + #else
708 +    d = snap_->cgData.position[cg1] - snap_->atomData.position[atom1];
709 + #endif
710 +
711 +    snap_->wrapVector(d);
712 +    return d;    
713 +  }
714 +  
715 +  Vector3d ForceMatrixDecomposition::getAtomToGroupVectorColumn(int atom2, int cg2){
716 +    Vector3d d;
717 +    
718 + #ifdef IS_MPI
719 +    d = cgColData.position[cg2] - atomColData.position[atom2];
720 + #else
721 +    d = snap_->cgData.position[cg2] - snap_->atomData.position[atom2];
722 + #endif
723 +    
724 +    snap_->wrapVector(d);
725 +    return d;    
726 +  }
727 +
728 +  RealType ForceMatrixDecomposition::getMassFactorRow(int atom1) {
729 + #ifdef IS_MPI
730 +    return massFactorsRow[atom1];
731 + #else
732 +    return massFactors[atom1];
733 + #endif
734 +  }
735 +
736 +  RealType ForceMatrixDecomposition::getMassFactorColumn(int atom2) {
737 + #ifdef IS_MPI
738 +    return massFactorsCol[atom2];
739 + #else
740 +    return massFactors[atom2];
741 + #endif
742 +
743 +  }
744 +    
745 +  Vector3d ForceMatrixDecomposition::getInteratomicVector(int atom1, int atom2){
746 +    Vector3d d;
747 +    
748 + #ifdef IS_MPI
749 +    d = atomColData.position[atom2] - atomRowData.position[atom1];
750 + #else
751 +    d = snap_->atomData.position[atom2] - snap_->atomData.position[atom1];
752 + #endif
753 +
754 +    snap_->wrapVector(d);
755 +    return d;    
756 +  }
757 +
758 +  vector<int> ForceMatrixDecomposition::getExcludesForAtom(int atom1) {
759 +    return excludesForAtom[atom1];
760 +  }
761 +
762 +  /**
763 +   * We need to exclude some overcounted interactions that result from
764 +   * the parallel decomposition.
765 +   */
766 +  bool ForceMatrixDecomposition::skipAtomPair(int atom1, int atom2) {
767 +    int unique_id_1, unique_id_2;
768 +
769 + #ifdef IS_MPI
770 +    // in MPI, we have to look up the unique IDs for each atom
771 +    unique_id_1 = AtomRowToGlobal[atom1];
772 +    unique_id_2 = AtomColToGlobal[atom2];
773 +
774 +    // this situation should only arise in MPI simulations
775 +    if (unique_id_1 == unique_id_2) return true;
776 +    
777 +    // this prevents us from doing the pair on multiple processors
778 +    if (unique_id_1 < unique_id_2) {
779 +      if ((unique_id_1 + unique_id_2) % 2 == 0) return true;
780 +    } else {
781 +      if ((unique_id_1 + unique_id_2) % 2 == 1) return true;
782 +    }
783 + #endif
784 +    return false;
785 +  }
786 +
787 +  /**
788 +   * We need to handle the interactions for atoms who are involved in
789 +   * the same rigid body as well as some short range interactions
790 +   * (bonds, bends, torsions) differently from other interactions.
791 +   * We'll still visit the pairwise routines, but with a flag that
792 +   * tells those routines to exclude the pair from direct long range
793 +   * interactions.  Some indirect interactions (notably reaction
794 +   * field) must still be handled for these pairs.
795 +   */
796 +  bool ForceMatrixDecomposition::excludeAtomPair(int atom1, int atom2) {
797 +    int unique_id_2;
798 +    
799 + #ifdef IS_MPI
800 +    // in MPI, we have to look up the unique IDs for the row atom.
801 +    unique_id_2 = AtomColToGlobal[atom2];
802 + #else
803 +    // in the normal loop, the atom numbers are unique
804 +    unique_id_2 = atom2;
805 + #endif
806 +    
807 +    for (vector<int>::iterator i = excludesForAtom[atom1].begin();
808 +         i != excludesForAtom[atom1].end(); ++i) {
809 +      if ( (*i) == unique_id_2 ) return true;
810 +    }
811 +
812 +    return false;
813 +  }
814 +
815 +
816 +  void ForceMatrixDecomposition::addForceToAtomRow(int atom1, Vector3d fg){
817 + #ifdef IS_MPI
818 +    atomRowData.force[atom1] += fg;
819 + #else
820 +    snap_->atomData.force[atom1] += fg;
821 + #endif
822 +  }
823 +
824 +  void ForceMatrixDecomposition::addForceToAtomColumn(int atom2, Vector3d fg){
825 + #ifdef IS_MPI
826 +    atomColData.force[atom2] += fg;
827 + #else
828 +    snap_->atomData.force[atom2] += fg;
829 + #endif
830 +  }
831 +
832 +    // filling interaction blocks with pointers
833 +  void ForceMatrixDecomposition::fillInteractionData(InteractionData &idat,
834 +                                                     int atom1, int atom2) {
835 +
836 +    idat.excluded = excludeAtomPair(atom1, atom2);
837 +  
838 + #ifdef IS_MPI
839 +    
840 +    idat.atypes = make_pair( ff_->getAtomType(identsRow[atom1]),
841 +                             ff_->getAtomType(identsCol[atom2]) );
842 +    
843 +    if (storageLayout_ & DataStorage::dslAmat) {
844 +      idat.A1 = &(atomRowData.aMat[atom1]);
845 +      idat.A2 = &(atomColData.aMat[atom2]);
846 +    }
847 +    
848 +    if (storageLayout_ & DataStorage::dslElectroFrame) {
849 +      idat.eFrame1 = &(atomRowData.electroFrame[atom1]);
850 +      idat.eFrame2 = &(atomColData.electroFrame[atom2]);
851 +    }
852 +
853 +    if (storageLayout_ & DataStorage::dslTorque) {
854 +      idat.t1 = &(atomRowData.torque[atom1]);
855 +      idat.t2 = &(atomColData.torque[atom2]);
856 +    }
857 +
858 +    if (storageLayout_ & DataStorage::dslDensity) {
859 +      idat.rho1 = &(atomRowData.density[atom1]);
860 +      idat.rho2 = &(atomColData.density[atom2]);
861 +    }
862 +
863 +    if (storageLayout_ & DataStorage::dslFunctional) {
864 +      idat.frho1 = &(atomRowData.functional[atom1]);
865 +      idat.frho2 = &(atomColData.functional[atom2]);
866 +    }
867 +
868 +    if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
869 +      idat.dfrho1 = &(atomRowData.functionalDerivative[atom1]);
870 +      idat.dfrho2 = &(atomColData.functionalDerivative[atom2]);
871 +    }
872 +
873 +    if (storageLayout_ & DataStorage::dslParticlePot) {
874 +      idat.particlePot1 = &(atomRowData.particlePot[atom1]);
875 +      idat.particlePot2 = &(atomColData.particlePot[atom2]);
876 +    }
877 +
878 +    if (storageLayout_ & DataStorage::dslSkippedCharge) {              
879 +      idat.skippedCharge1 = &(atomRowData.skippedCharge[atom1]);
880 +      idat.skippedCharge2 = &(atomColData.skippedCharge[atom2]);
881 +    }
882 +
883 + #else
884 +
885 +    idat.atypes = make_pair( ff_->getAtomType(idents[atom1]),
886 +                             ff_->getAtomType(idents[atom2]) );
887 +
888 +    if (storageLayout_ & DataStorage::dslAmat) {
889 +      idat.A1 = &(snap_->atomData.aMat[atom1]);
890 +      idat.A2 = &(snap_->atomData.aMat[atom2]);
891 +    }
892 +
893 +    if (storageLayout_ & DataStorage::dslElectroFrame) {
894 +      idat.eFrame1 = &(snap_->atomData.electroFrame[atom1]);
895 +      idat.eFrame2 = &(snap_->atomData.electroFrame[atom2]);
896 +    }
897 +
898 +    if (storageLayout_ & DataStorage::dslTorque) {
899 +      idat.t1 = &(snap_->atomData.torque[atom1]);
900 +      idat.t2 = &(snap_->atomData.torque[atom2]);
901 +    }
902 +
903 +    if (storageLayout_ & DataStorage::dslDensity) {    
904 +      idat.rho1 = &(snap_->atomData.density[atom1]);
905 +      idat.rho2 = &(snap_->atomData.density[atom2]);
906 +    }
907 +
908 +    if (storageLayout_ & DataStorage::dslFunctional) {
909 +      idat.frho1 = &(snap_->atomData.functional[atom1]);
910 +      idat.frho2 = &(snap_->atomData.functional[atom2]);
911 +    }
912 +
913 +    if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
914 +      idat.dfrho1 = &(snap_->atomData.functionalDerivative[atom1]);
915 +      idat.dfrho2 = &(snap_->atomData.functionalDerivative[atom2]);
916 +    }
917 +
918 +    if (storageLayout_ & DataStorage::dslParticlePot) {
919 +      idat.particlePot1 = &(snap_->atomData.particlePot[atom1]);
920 +      idat.particlePot2 = &(snap_->atomData.particlePot[atom2]);
921 +    }
922 +
923 +    if (storageLayout_ & DataStorage::dslSkippedCharge) {
924 +      idat.skippedCharge1 = &(snap_->atomData.skippedCharge[atom1]);
925 +      idat.skippedCharge2 = &(snap_->atomData.skippedCharge[atom2]);
926 +    }
927 + #endif
928 +  }
929 +
930 +  
931 +  void ForceMatrixDecomposition::unpackInteractionData(InteractionData &idat, int atom1, int atom2) {    
932 + #ifdef IS_MPI
933 +    pot_row[atom1] += 0.5 *  *(idat.pot);
934 +    pot_col[atom2] += 0.5 *  *(idat.pot);
935 +
936 +    atomRowData.force[atom1] += *(idat.f1);
937 +    atomColData.force[atom2] -= *(idat.f1);
938 + #else
939 +    pairwisePot += *(idat.pot);
940 +
941 +    snap_->atomData.force[atom1] += *(idat.f1);
942 +    snap_->atomData.force[atom2] -= *(idat.f1);
943 + #endif
944 +    
945 +  }
946 +
947 +  /*
948 +   * buildNeighborList
949 +   *
950 +   * first element of pair is row-indexed CutoffGroup
951 +   * second element of pair is column-indexed CutoffGroup
952 +   */
953 +  vector<pair<int, int> > ForceMatrixDecomposition::buildNeighborList() {
954 +      
955 +    vector<pair<int, int> > neighborList;
956 +    groupCutoffs cuts;
957 +    bool doAllPairs = false;
958 +
959 + #ifdef IS_MPI
960 +    cellListRow_.clear();
961 +    cellListCol_.clear();
962 + #else
963 +    cellList_.clear();
964 + #endif
965 +
966 +    RealType rList_ = (largestRcut_ + skinThickness_);
967 +    RealType rl2 = rList_ * rList_;
968 +    Snapshot* snap_ = sman_->getCurrentSnapshot();
969 +    Mat3x3d Hmat = snap_->getHmat();
970 +    Vector3d Hx = Hmat.getColumn(0);
971 +    Vector3d Hy = Hmat.getColumn(1);
972 +    Vector3d Hz = Hmat.getColumn(2);
973 +
974 +    nCells_.x() = (int) ( Hx.length() )/ rList_;
975 +    nCells_.y() = (int) ( Hy.length() )/ rList_;
976 +    nCells_.z() = (int) ( Hz.length() )/ rList_;
977 +
978 +    // handle small boxes where the cell offsets can end up repeating cells
979 +    
980 +    if (nCells_.x() < 3) doAllPairs = true;
981 +    if (nCells_.y() < 3) doAllPairs = true;
982 +    if (nCells_.z() < 3) doAllPairs = true;
983 +
984 +    Mat3x3d invHmat = snap_->getInvHmat();
985 +    Vector3d rs, scaled, dr;
986 +    Vector3i whichCell;
987 +    int cellIndex;
988 +    int nCtot = nCells_.x() * nCells_.y() * nCells_.z();
989 +
990 + #ifdef IS_MPI
991 +    cellListRow_.resize(nCtot);
992 +    cellListCol_.resize(nCtot);
993 + #else
994 +    cellList_.resize(nCtot);
995 + #endif
996 +
997 +    if (!doAllPairs) {
998 + #ifdef IS_MPI
999 +
1000 +      for (int i = 0; i < nGroupsInRow_; i++) {
1001 +        rs = cgRowData.position[i];
1002 +        
1003 +        // scaled positions relative to the box vectors
1004 +        scaled = invHmat * rs;
1005 +        
1006 +        // wrap the vector back into the unit box by subtracting integer box
1007 +        // numbers
1008 +        for (int j = 0; j < 3; j++) {
1009 +          scaled[j] -= roundMe(scaled[j]);
1010 +          scaled[j] += 0.5;
1011 +        }
1012 +        
1013 +        // find xyz-indices of cell that cutoffGroup is in.
1014 +        whichCell.x() = nCells_.x() * scaled.x();
1015 +        whichCell.y() = nCells_.y() * scaled.y();
1016 +        whichCell.z() = nCells_.z() * scaled.z();
1017 +        
1018 +        // find single index of this cell:
1019 +        cellIndex = Vlinear(whichCell, nCells_);
1020 +        
1021 +        // add this cutoff group to the list of groups in this cell;
1022 +        cellListRow_[cellIndex].push_back(i);
1023 +      }
1024 +      
1025 +      for (int i = 0; i < nGroupsInCol_; i++) {
1026 +        rs = cgColData.position[i];
1027 +        
1028 +        // scaled positions relative to the box vectors
1029 +        scaled = invHmat * rs;
1030 +        
1031 +        // wrap the vector back into the unit box by subtracting integer box
1032 +        // numbers
1033 +        for (int j = 0; j < 3; j++) {
1034 +          scaled[j] -= roundMe(scaled[j]);
1035 +          scaled[j] += 0.5;
1036 +        }
1037 +        
1038 +        // find xyz-indices of cell that cutoffGroup is in.
1039 +        whichCell.x() = nCells_.x() * scaled.x();
1040 +        whichCell.y() = nCells_.y() * scaled.y();
1041 +        whichCell.z() = nCells_.z() * scaled.z();
1042 +        
1043 +        // find single index of this cell:
1044 +        cellIndex = Vlinear(whichCell, nCells_);
1045 +        
1046 +        // add this cutoff group to the list of groups in this cell;
1047 +        cellListCol_[cellIndex].push_back(i);
1048 +      }
1049 + #else
1050 +      for (int i = 0; i < nGroups_; i++) {
1051 +        rs = snap_->cgData.position[i];
1052 +        
1053 +        // scaled positions relative to the box vectors
1054 +        scaled = invHmat * rs;
1055 +        
1056 +        // wrap the vector back into the unit box by subtracting integer box
1057 +        // numbers
1058 +        for (int j = 0; j < 3; j++) {
1059 +          scaled[j] -= roundMe(scaled[j]);
1060 +          scaled[j] += 0.5;
1061 +        }
1062 +        
1063 +        // find xyz-indices of cell that cutoffGroup is in.
1064 +        whichCell.x() = nCells_.x() * scaled.x();
1065 +        whichCell.y() = nCells_.y() * scaled.y();
1066 +        whichCell.z() = nCells_.z() * scaled.z();
1067 +        
1068 +        // find single index of this cell:
1069 +        cellIndex = Vlinear(whichCell, nCells_);      
1070 +        
1071 +        // add this cutoff group to the list of groups in this cell;
1072 +        cellList_[cellIndex].push_back(i);
1073 +      }
1074 + #endif
1075 +
1076 +      for (int m1z = 0; m1z < nCells_.z(); m1z++) {
1077 +        for (int m1y = 0; m1y < nCells_.y(); m1y++) {
1078 +          for (int m1x = 0; m1x < nCells_.x(); m1x++) {
1079 +            Vector3i m1v(m1x, m1y, m1z);
1080 +            int m1 = Vlinear(m1v, nCells_);
1081 +            
1082 +            for (vector<Vector3i>::iterator os = cellOffsets_.begin();
1083 +                 os != cellOffsets_.end(); ++os) {
1084 +              
1085 +              Vector3i m2v = m1v + (*os);
1086 +              
1087 +              if (m2v.x() >= nCells_.x()) {
1088 +                m2v.x() = 0;          
1089 +              } else if (m2v.x() < 0) {
1090 +                m2v.x() = nCells_.x() - 1;
1091 +              }
1092 +              
1093 +              if (m2v.y() >= nCells_.y()) {
1094 +                m2v.y() = 0;          
1095 +              } else if (m2v.y() < 0) {
1096 +                m2v.y() = nCells_.y() - 1;
1097 +              }
1098 +              
1099 +              if (m2v.z() >= nCells_.z()) {
1100 +                m2v.z() = 0;          
1101 +              } else if (m2v.z() < 0) {
1102 +                m2v.z() = nCells_.z() - 1;
1103 +              }
1104 +              
1105 +              int m2 = Vlinear (m2v, nCells_);
1106 +              
1107 + #ifdef IS_MPI
1108 +              for (vector<int>::iterator j1 = cellListRow_[m1].begin();
1109 +                   j1 != cellListRow_[m1].end(); ++j1) {
1110 +                for (vector<int>::iterator j2 = cellListCol_[m2].begin();
1111 +                     j2 != cellListCol_[m2].end(); ++j2) {
1112 +                  
1113 +                  // Always do this if we're in different cells or if
1114 +                  // we're in the same cell and the global index of the
1115 +                  // j2 cutoff group is less than the j1 cutoff group
1116 +                  
1117 +                  if (m2 != m1 || cgColToGlobal[(*j2)] < cgRowToGlobal[(*j1)]) {
1118 +                    dr = cgColData.position[(*j2)] - cgRowData.position[(*j1)];
1119 +                    snap_->wrapVector(dr);
1120 +                    cuts = getGroupCutoffs( (*j1), (*j2) );
1121 +                    if (dr.lengthSquare() < cuts.third) {
1122 +                      neighborList.push_back(make_pair((*j1), (*j2)));
1123 +                    }
1124 +                  }
1125 +                }
1126 +              }
1127 + #else
1128 +              
1129 +              for (vector<int>::iterator j1 = cellList_[m1].begin();
1130 +                   j1 != cellList_[m1].end(); ++j1) {
1131 +                for (vector<int>::iterator j2 = cellList_[m2].begin();
1132 +                     j2 != cellList_[m2].end(); ++j2) {
1133 +                  
1134 +                  // Always do this if we're in different cells or if
1135 +                  // we're in the same cell and the global index of the
1136 +                  // j2 cutoff group is less than the j1 cutoff group
1137 +                  
1138 +                  if (m2 != m1 || (*j2) < (*j1)) {
1139 +                    dr = snap_->cgData.position[(*j2)] - snap_->cgData.position[(*j1)];
1140 +                    snap_->wrapVector(dr);
1141 +                    cuts = getGroupCutoffs( (*j1), (*j2) );
1142 +                    if (dr.lengthSquare() < cuts.third) {
1143 +                      neighborList.push_back(make_pair((*j1), (*j2)));
1144 +                    }
1145 +                  }
1146 +                }
1147 +              }
1148 + #endif
1149 +            }
1150 +          }
1151 +        }
1152 +      }
1153 +    } else {
1154 +      // branch to do all cutoff group pairs
1155 + #ifdef IS_MPI
1156 +      for (int j1 = 0; j1 < nGroupsInRow_; j1++) {
1157 +        for (int j2 = 0; j2 < nGroupsInCol_; j2++) {      
1158 +          dr = cgColData.position[j2] - cgRowData.position[j1];
1159 +          snap_->wrapVector(dr);
1160 +          cuts = getGroupCutoffs( j1, j2 );
1161 +          if (dr.lengthSquare() < cuts.third) {
1162 +            neighborList.push_back(make_pair(j1, j2));
1163 +          }
1164 +        }
1165 +      }
1166 + #else
1167 +      for (int j1 = 0; j1 < nGroups_ - 1; j1++) {
1168 +        for (int j2 = j1 + 1; j2 < nGroups_; j2++) {
1169 +          dr = snap_->cgData.position[j2] - snap_->cgData.position[j1];
1170 +          snap_->wrapVector(dr);
1171 +          cuts = getGroupCutoffs( j1, j2 );
1172 +          if (dr.lengthSquare() < cuts.third) {
1173 +            neighborList.push_back(make_pair(j1, j2));
1174 +          }
1175 +        }
1176 +      }        
1177 + #endif
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 neighborList;
1187 +  }
1188 + } //end namespace OpenMD

Diff Legend

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