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 1575 by gezelter, Fri Jun 3 21:39:49 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 +    nGroups_ = snap_->getNumberOfCutoffGroups();
61  
62 < /*  -*- c++ -*-  */
63 < #include "config.h"
64 < #include <stdlib.h>
62 >    // gather the information for atomtype IDs (atids):
63 >    identsLocal = info_->getIdentArray();
64 >    AtomLocalToGlobal = info_->getGlobalAtomIndices();
65 >    cgLocalToGlobal = info_->getGlobalGroupIndices();
66 >    vector<int> globalGroupMembership = info_->getGlobalGroupMembership();
67 >    vector<RealType> massFactorsLocal = info_->getMassFactors();
68 >    PairList excludes = info_->getExcludedInteractions();
69 >    PairList oneTwo = info_->getOneTwoInteractions();
70 >    PairList oneThree = info_->getOneThreeInteractions();
71 >    PairList oneFour = info_->getOneFourInteractions();
72 >
73   #ifdef IS_MPI
74 < #include <mpi.h>
75 < #endif
74 >
75 >    AtomCommIntRow = new Communicator<Row,int>(nLocal_);
76 >    AtomCommRealRow = new Communicator<Row,RealType>(nLocal_);
77 >    AtomCommVectorRow = new Communicator<Row,Vector3d>(nLocal_);
78 >    AtomCommMatrixRow = new Communicator<Row,Mat3x3d>(nLocal_);
79 >    AtomCommPotRow = new Communicator<Row,potVec>(nLocal_);
80  
81 < #include <iostream>
82 < #include <vector>
83 < #include <algorithm>
84 < #include <cmath>
85 < #include "parallel/ForceDecomposition.hpp"
81 >    AtomCommIntColumn = new Communicator<Column,int>(nLocal_);
82 >    AtomCommRealColumn = new Communicator<Column,RealType>(nLocal_);
83 >    AtomCommVectorColumn = new Communicator<Column,Vector3d>(nLocal_);
84 >    AtomCommMatrixColumn = new Communicator<Column,Mat3x3d>(nLocal_);
85 >    AtomCommPotColumn = new Communicator<Column,potVec>(nLocal_);
86  
87 +    cgCommIntRow = new Communicator<Row,int>(nGroups_);
88 +    cgCommVectorRow = new Communicator<Row,Vector3d>(nGroups_);
89 +    cgCommIntColumn = new Communicator<Column,int>(nGroups_);
90 +    cgCommVectorColumn = new Communicator<Column,Vector3d>(nGroups_);
91  
92 < using namespace std;
93 < using namespace OpenMD;
92 >    nAtomsInRow_ = AtomCommIntRow->getSize();
93 >    nAtomsInCol_ = AtomCommIntColumn->getSize();
94 >    nGroupsInRow_ = cgCommIntRow->getSize();
95 >    nGroupsInCol_ = cgCommIntColumn->getSize();
96  
97 < //__static
98 < #ifdef IS_MPI
99 < static vector<MPI:Comm> communictors;
100 < #endif
97 >    // Modify the data storage objects with the correct layouts and sizes:
98 >    atomRowData.resize(nAtomsInRow_);
99 >    atomRowData.setStorageLayout(storageLayout_);
100 >    atomColData.resize(nAtomsInCol_);
101 >    atomColData.setStorageLayout(storageLayout_);
102 >    cgRowData.resize(nGroupsInRow_);
103 >    cgRowData.setStorageLayout(DataStorage::dslPosition);
104 >    cgColData.resize(nGroupsInCol_);
105 >    cgColData.setStorageLayout(DataStorage::dslPosition);
106 >        
107 >    identsRow.reserve(nAtomsInRow_);
108 >    identsCol.reserve(nAtomsInCol_);
109 >    
110 >    AtomCommIntRow->gather(identsLocal, identsRow);
111 >    AtomCommIntColumn->gather(identsLocal, identsCol);
112 >    
113 >    AtomCommIntRow->gather(AtomLocalToGlobal, AtomRowToGlobal);
114 >    AtomCommIntColumn->gather(AtomLocalToGlobal, AtomColToGlobal);
115 >    
116 >    cgCommIntRow->gather(cgLocalToGlobal, cgRowToGlobal);
117 >    cgCommIntColumn->gather(cgLocalToGlobal, cgColToGlobal);
118  
119 < //____ MPITypeTraits
120 < template<typename T>
75 < struct MPITypeTraits;
119 >    AtomCommRealRow->gather(massFactorsLocal, massFactorsRow);
120 >    AtomCommRealColumn->gather(massFactorsLocal, massFactorsCol);
121  
122 < #ifdef IS_MPI
123 < template<>
124 < struct MPITypeTraits<RealType> {
125 <  static const MPI::Datatype datatype;
126 < };
127 < const MPI_Datatype MPITypeTraits<RealType>::datatype = MY_MPI_REAL;
122 >    groupListRow_.clear();
123 >    groupListRow_.reserve(nGroupsInRow_);
124 >    for (int i = 0; i < nGroupsInRow_; i++) {
125 >      int gid = cgRowToGlobal[i];
126 >      for (int j = 0; j < nAtomsInRow_; j++) {
127 >        int aid = AtomRowToGlobal[j];
128 >        if (globalGroupMembership[aid] == gid)
129 >          groupListRow_[i].push_back(j);
130 >      }      
131 >    }
132  
133 < template<>
134 < struct MPITypeTraits<int> {
135 <  static const MPI::Datatype datatype;
136 < };
137 < const MPI::Datatype MPITypeTraits<int>::datatype = MPI_INT;
138 < #endif
133 >    groupListCol_.clear();
134 >    groupListCol_.reserve(nGroupsInCol_);
135 >    for (int i = 0; i < nGroupsInCol_; i++) {
136 >      int gid = cgColToGlobal[i];
137 >      for (int j = 0; j < nAtomsInCol_; j++) {
138 >        int aid = AtomColToGlobal[j];
139 >        if (globalGroupMembership[aid] == gid)
140 >          groupListCol_[i].push_back(j);
141 >      }      
142 >    }
143  
144 < /**
145 < * Constructor for ForceDecomposition Parallel Decomposition Method
146 < * Will try to construct a symmetric grid of processors. Ideally, the
147 < * number of processors will be a square ex: 4, 9, 16, 25.
148 < *
149 < */
144 >    skipsForRowAtom.clear();
145 >    skipsForRowAtom.reserve(nAtomsInRow_);
146 >    for (int i = 0; i < nAtomsInRow_; i++) {
147 >      int iglob = AtomRowToGlobal[i];
148 >      for (int j = 0; j < nAtomsInCol_; j++) {
149 >        int jglob = AtomColToGlobal[j];        
150 >        if (excludes.hasPair(iglob, jglob))
151 >          skipsForRowAtom[i].push_back(j);      
152 >      }      
153 >    }
154  
155 < ForceDecomposition::ForceDecomposition() {
155 >    toposForRowAtom.clear();
156 >    toposForRowAtom.reserve(nAtomsInRow_);
157 >    for (int i = 0; i < nAtomsInRow_; i++) {
158 >      int iglob = AtomRowToGlobal[i];
159 >      int nTopos = 0;
160 >      for (int j = 0; j < nAtomsInCol_; j++) {
161 >        int jglob = AtomColToGlobal[j];        
162 >        if (oneTwo.hasPair(iglob, jglob)) {
163 >          toposForRowAtom[i].push_back(j);
164 >          topoDistRow[i][nTopos] = 1;
165 >          nTopos++;
166 >        }
167 >        if (oneThree.hasPair(iglob, jglob)) {
168 >          toposForRowAtom[i].push_back(j);
169 >          topoDistRow[i][nTopos] = 2;
170 >          nTopos++;
171 >        }
172 >        if (oneFour.hasPair(iglob, jglob)) {
173 >          toposForRowAtom[i].push_back(j);
174 >          topoDistRow[i][nTopos] = 3;
175 >          nTopos++;
176 >        }
177 >      }      
178 >    }
179  
100 – #ifdef IS_MPI
101 –  int nProcs = MPI::COMM_WORLD.Get_size();
102 –  int worldRank = MPI::COMM_WORLD.Get_rank();
180   #endif
181  
182 <  // First time through, construct column stride.
183 <  if (communicators.size() == 0)
184 <  {
185 <    int nColumnsMax = (int) round(sqrt((float) nProcs));
186 <    for (int i = 0; i < nProcs; ++i)
187 <    {
188 <      if (nProcs%i==0) nColumns=i;
182 >    groupList_.clear();
183 >    groupList_.reserve(nGroups_);
184 >    for (int i = 0; i < nGroups_; i++) {
185 >      int gid = cgLocalToGlobal[i];
186 >      for (int j = 0; j < nLocal_; j++) {
187 >        int aid = AtomLocalToGlobal[j];
188 >        if (globalGroupMembership[aid] == gid)
189 >          groupList_[i].push_back(j);
190 >      }      
191      }
192  
193 <    int nRows = nProcs/nColumns;    
194 <    myRank_ = (int) worldRank%nColumns;
116 <  }
117 <  else
118 <  {
119 <    myRank_ = myRank/nColumns;
120 <  }
121 <  MPI::Comm newComm = MPI:COMM_WORLD.Split(myRank_,0);
122 <  
123 <  isColumn_ = false;
124 <  
125 < }
193 >    skipsForLocalAtom.clear();
194 >    skipsForLocalAtom.reserve(nLocal_);
195  
196 < ForceDecomposition::gather(sendbuf, receivebuf){
197 <  communicators(myIndex_).Allgatherv();
198 < }
196 >    for (int i = 0; i < nLocal_; i++) {
197 >      int iglob = AtomLocalToGlobal[i];
198 >      for (int j = 0; j < nLocal_; j++) {
199 >        int jglob = AtomLocalToGlobal[j];        
200 >        if (excludes.hasPair(iglob, jglob))
201 >          skipsForLocalAtom[i].push_back(j);      
202 >      }      
203 >    }
204  
205 <
206 <
207 < ForceDecomposition::scatter(sbuffer, rbuffer){
208 <  communicators(myIndex_).Reduce_scatter(sbuffer, recevbuf. recvcounts, MPI::DOUBLE, MPI::SUM);
209 < }
205 >    toposForLocalAtom.clear();
206 >    toposForLocalAtom.reserve(nLocal_);
207 >    for (int i = 0; i < nLocal_; i++) {
208 >      int iglob = AtomLocalToGlobal[i];
209 >      int nTopos = 0;
210 >      for (int j = 0; j < nLocal_; j++) {
211 >        int jglob = AtomLocalToGlobal[j];        
212 >        if (oneTwo.hasPair(iglob, jglob)) {
213 >          toposForLocalAtom[i].push_back(j);
214 >          topoDistLocal[i][nTopos] = 1;
215 >          nTopos++;
216 >        }
217 >        if (oneThree.hasPair(iglob, jglob)) {
218 >          toposForLocalAtom[i].push_back(j);
219 >          topoDistLocal[i][nTopos] = 2;
220 >          nTopos++;
221 >        }
222 >        if (oneFour.hasPair(iglob, jglob)) {
223 >          toposForLocalAtom[i].push_back(j);
224 >          topoDistLocal[i][nTopos] = 3;
225 >          nTopos++;
226 >        }
227 >      }      
228 >    }
229 >  }
230 >  
231 >  void ForceMatrixDecomposition::zeroWorkArrays() {
232  
233 +    for (int j = 0; j < N_INTERACTION_FAMILIES; j++) {
234 +      longRangePot_[j] = 0.0;
235 +    }
236  
237 + #ifdef IS_MPI
238 +    if (storageLayout_ & DataStorage::dslForce) {
239 +      fill(atomRowData.force.begin(), atomRowData.force.end(), V3Zero);
240 +      fill(atomColData.force.begin(), atomColData.force.end(), V3Zero);
241 +    }
242 +
243 +    if (storageLayout_ & DataStorage::dslTorque) {
244 +      fill(atomRowData.torque.begin(), atomRowData.torque.end(), V3Zero);
245 +      fill(atomColData.torque.begin(), atomColData.torque.end(), V3Zero);
246 +    }
247 +    
248 +    fill(pot_row.begin(), pot_row.end(),
249 +         Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
250 +
251 +    fill(pot_col.begin(), pot_col.end(),
252 +         Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
253 +    
254 +    pot_local = Vector<RealType, N_INTERACTION_FAMILIES>(0.0);
255 +
256 +    if (storageLayout_ & DataStorage::dslParticlePot) {    
257 +      fill(atomRowData.particlePot.begin(), atomRowData.particlePot.end(), 0.0);
258 +      fill(atomColData.particlePot.begin(), atomColData.particlePot.end(), 0.0);
259 +    }
260 +
261 +    if (storageLayout_ & DataStorage::dslDensity) {      
262 +      fill(atomRowData.density.begin(), atomRowData.density.end(), 0.0);
263 +      fill(atomColData.density.begin(), atomColData.density.end(), 0.0);
264 +    }
265 +
266 +    if (storageLayout_ & DataStorage::dslFunctional) {  
267 +      fill(atomRowData.functional.begin(), atomRowData.functional.end(), 0.0);
268 +      fill(atomColData.functional.begin(), atomColData.functional.end(), 0.0);
269 +    }
270 +
271 +    if (storageLayout_ & DataStorage::dslFunctionalDerivative) {      
272 +      fill(atomRowData.functionalDerivative.begin(),
273 +           atomRowData.functionalDerivative.end(), 0.0);
274 +      fill(atomColData.functionalDerivative.begin(),
275 +           atomColData.functionalDerivative.end(), 0.0);
276 +    }
277 +
278 + #else
279 +    
280 +    if (storageLayout_ & DataStorage::dslParticlePot) {      
281 +      fill(snap_->atomData.particlePot.begin(),
282 +           snap_->atomData.particlePot.end(), 0.0);
283 +    }
284 +    
285 +    if (storageLayout_ & DataStorage::dslDensity) {      
286 +      fill(snap_->atomData.density.begin(),
287 +           snap_->atomData.density.end(), 0.0);
288 +    }
289 +    if (storageLayout_ & DataStorage::dslFunctional) {
290 +      fill(snap_->atomData.functional.begin(),
291 +           snap_->atomData.functional.end(), 0.0);
292 +    }
293 +    if (storageLayout_ & DataStorage::dslFunctionalDerivative) {      
294 +      fill(snap_->atomData.functionalDerivative.begin(),
295 +           snap_->atomData.functionalDerivative.end(), 0.0);
296 +    }
297 + #endif
298 +    
299 +  }
300 +
301 +
302 +  void ForceMatrixDecomposition::distributeData()  {
303 +    snap_ = sman_->getCurrentSnapshot();
304 +    storageLayout_ = sman_->getStorageLayout();
305 + #ifdef IS_MPI
306 +    
307 +    // gather up the atomic positions
308 +    AtomCommVectorRow->gather(snap_->atomData.position,
309 +                              atomRowData.position);
310 +    AtomCommVectorColumn->gather(snap_->atomData.position,
311 +                                 atomColData.position);
312 +    
313 +    // gather up the cutoff group positions
314 +    cgCommVectorRow->gather(snap_->cgData.position,
315 +                            cgRowData.position);
316 +    cgCommVectorColumn->gather(snap_->cgData.position,
317 +                               cgColData.position);
318 +    
319 +    // if needed, gather the atomic rotation matrices
320 +    if (storageLayout_ & DataStorage::dslAmat) {
321 +      AtomCommMatrixRow->gather(snap_->atomData.aMat,
322 +                                atomRowData.aMat);
323 +      AtomCommMatrixColumn->gather(snap_->atomData.aMat,
324 +                                   atomColData.aMat);
325 +    }
326 +    
327 +    // if needed, gather the atomic eletrostatic frames
328 +    if (storageLayout_ & DataStorage::dslElectroFrame) {
329 +      AtomCommMatrixRow->gather(snap_->atomData.electroFrame,
330 +                                atomRowData.electroFrame);
331 +      AtomCommMatrixColumn->gather(snap_->atomData.electroFrame,
332 +                                   atomColData.electroFrame);
333 +    }
334 + #endif      
335 +  }
336 +  
337 +  /* collects information obtained during the pre-pair loop onto local
338 +   * data structures.
339 +   */
340 +  void ForceMatrixDecomposition::collectIntermediateData() {
341 +    snap_ = sman_->getCurrentSnapshot();
342 +    storageLayout_ = sman_->getStorageLayout();
343 + #ifdef IS_MPI
344 +    
345 +    if (storageLayout_ & DataStorage::dslDensity) {
346 +      
347 +      AtomCommRealRow->scatter(atomRowData.density,
348 +                               snap_->atomData.density);
349 +      
350 +      int n = snap_->atomData.density.size();
351 +      vector<RealType> rho_tmp(n, 0.0);
352 +      AtomCommRealColumn->scatter(atomColData.density, rho_tmp);
353 +      for (int i = 0; i < n; i++)
354 +        snap_->atomData.density[i] += rho_tmp[i];
355 +    }
356 + #endif
357 +  }
358 +
359 +  /*
360 +   * redistributes information obtained during the pre-pair loop out to
361 +   * row and column-indexed data structures
362 +   */
363 +  void ForceMatrixDecomposition::distributeIntermediateData() {
364 +    snap_ = sman_->getCurrentSnapshot();
365 +    storageLayout_ = sman_->getStorageLayout();
366 + #ifdef IS_MPI
367 +    if (storageLayout_ & DataStorage::dslFunctional) {
368 +      AtomCommRealRow->gather(snap_->atomData.functional,
369 +                              atomRowData.functional);
370 +      AtomCommRealColumn->gather(snap_->atomData.functional,
371 +                                 atomColData.functional);
372 +    }
373 +    
374 +    if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
375 +      AtomCommRealRow->gather(snap_->atomData.functionalDerivative,
376 +                              atomRowData.functionalDerivative);
377 +      AtomCommRealColumn->gather(snap_->atomData.functionalDerivative,
378 +                                 atomColData.functionalDerivative);
379 +    }
380 + #endif
381 +  }
382 +  
383 +  
384 +  void ForceMatrixDecomposition::collectData() {
385 +    snap_ = sman_->getCurrentSnapshot();
386 +    storageLayout_ = sman_->getStorageLayout();
387 + #ifdef IS_MPI    
388 +    int n = snap_->atomData.force.size();
389 +    vector<Vector3d> frc_tmp(n, V3Zero);
390 +    
391 +    AtomCommVectorRow->scatter(atomRowData.force, frc_tmp);
392 +    for (int i = 0; i < n; i++) {
393 +      snap_->atomData.force[i] += frc_tmp[i];
394 +      frc_tmp[i] = 0.0;
395 +    }
396 +    
397 +    AtomCommVectorColumn->scatter(atomColData.force, frc_tmp);
398 +    for (int i = 0; i < n; i++)
399 +      snap_->atomData.force[i] += frc_tmp[i];
400 +    
401 +    
402 +    if (storageLayout_ & DataStorage::dslTorque) {
403 +
404 +      int nt = snap_->atomData.force.size();
405 +      vector<Vector3d> trq_tmp(nt, V3Zero);
406 +
407 +      AtomCommVectorRow->scatter(atomRowData.torque, trq_tmp);
408 +      for (int i = 0; i < n; i++) {
409 +        snap_->atomData.torque[i] += trq_tmp[i];
410 +        trq_tmp[i] = 0.0;
411 +      }
412 +      
413 +      AtomCommVectorColumn->scatter(atomColData.torque, trq_tmp);
414 +      for (int i = 0; i < n; i++)
415 +        snap_->atomData.torque[i] += trq_tmp[i];
416 +    }
417 +    
418 +    nLocal_ = snap_->getNumberOfAtoms();
419 +
420 +    vector<potVec> pot_temp(nLocal_,
421 +                            Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
422 +
423 +    // scatter/gather pot_row into the members of my column
424 +          
425 +    AtomCommPotRow->scatter(pot_row, pot_temp);
426 +
427 +    for (int ii = 0;  ii < pot_temp.size(); ii++ )
428 +      pot_local += pot_temp[ii];
429 +    
430 +    fill(pot_temp.begin(), pot_temp.end(),
431 +         Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
432 +      
433 +    AtomCommPotColumn->scatter(pot_col, pot_temp);    
434 +    
435 +    for (int ii = 0;  ii < pot_temp.size(); ii++ )
436 +      pot_local += pot_temp[ii];
437 +    
438 + #endif
439 +  }
440 +
441 +  int ForceMatrixDecomposition::getNAtomsInRow() {  
442 + #ifdef IS_MPI
443 +    return nAtomsInRow_;
444 + #else
445 +    return nLocal_;
446 + #endif
447 +  }
448 +
449 +  /**
450 +   * returns the list of atoms belonging to this group.  
451 +   */
452 +  vector<int> ForceMatrixDecomposition::getAtomsInGroupRow(int cg1){
453 + #ifdef IS_MPI
454 +    return groupListRow_[cg1];
455 + #else
456 +    return groupList_[cg1];
457 + #endif
458 +  }
459 +
460 +  vector<int> ForceMatrixDecomposition::getAtomsInGroupColumn(int cg2){
461 + #ifdef IS_MPI
462 +    return groupListCol_[cg2];
463 + #else
464 +    return groupList_[cg2];
465 + #endif
466 +  }
467 +  
468 +  Vector3d ForceMatrixDecomposition::getIntergroupVector(int cg1, int cg2){
469 +    Vector3d d;
470 +    
471 + #ifdef IS_MPI
472 +    d = cgColData.position[cg2] - cgRowData.position[cg1];
473 + #else
474 +    d = snap_->cgData.position[cg2] - snap_->cgData.position[cg1];
475 + #endif
476 +    
477 +    snap_->wrapVector(d);
478 +    return d;    
479 +  }
480 +
481 +
482 +  Vector3d ForceMatrixDecomposition::getAtomToGroupVectorRow(int atom1, int cg1){
483 +
484 +    Vector3d d;
485 +    
486 + #ifdef IS_MPI
487 +    d = cgRowData.position[cg1] - atomRowData.position[atom1];
488 + #else
489 +    d = snap_->cgData.position[cg1] - snap_->atomData.position[atom1];
490 + #endif
491 +
492 +    snap_->wrapVector(d);
493 +    return d;    
494 +  }
495 +  
496 +  Vector3d ForceMatrixDecomposition::getAtomToGroupVectorColumn(int atom2, int cg2){
497 +    Vector3d d;
498 +    
499 + #ifdef IS_MPI
500 +    d = cgColData.position[cg2] - atomColData.position[atom2];
501 + #else
502 +    d = snap_->cgData.position[cg2] - snap_->atomData.position[atom2];
503 + #endif
504 +    
505 +    snap_->wrapVector(d);
506 +    return d;    
507 +  }
508 +
509 +  RealType ForceMatrixDecomposition::getMassFactorRow(int atom1) {
510 + #ifdef IS_MPI
511 +    return massFactorsRow[atom1];
512 + #else
513 +    return massFactorsLocal[atom1];
514 + #endif
515 +  }
516 +
517 +  RealType ForceMatrixDecomposition::getMassFactorColumn(int atom2) {
518 + #ifdef IS_MPI
519 +    return massFactorsCol[atom2];
520 + #else
521 +    return massFactorsLocal[atom2];
522 + #endif
523 +
524 +  }
525 +    
526 +  Vector3d ForceMatrixDecomposition::getInteratomicVector(int atom1, int atom2){
527 +    Vector3d d;
528 +    
529 + #ifdef IS_MPI
530 +    d = atomColData.position[atom2] - atomRowData.position[atom1];
531 + #else
532 +    d = snap_->atomData.position[atom2] - snap_->atomData.position[atom1];
533 + #endif
534 +
535 +    snap_->wrapVector(d);
536 +    return d;    
537 +  }
538 +
539 +  vector<int> ForceMatrixDecomposition::getSkipsForRowAtom(int atom1) {
540 + #ifdef IS_MPI
541 +    return skipsForRowAtom[atom1];
542 + #else
543 +    return skipsForLocalAtom[atom1];
544 + #endif
545 +  }
546 +
547 +  /**
548 +   * There are a number of reasons to skip a pair or a
549 +   * particle. Mostly we do this to exclude atoms who are involved in
550 +   * short range interactions (bonds, bends, torsions), but we also
551 +   * need to exclude some overcounted interactions that result from
552 +   * the parallel decomposition.
553 +   */
554 +  bool ForceMatrixDecomposition::skipAtomPair(int atom1, int atom2) {
555 +    int unique_id_1, unique_id_2;
556 +
557 + #ifdef IS_MPI
558 +    // in MPI, we have to look up the unique IDs for each atom
559 +    unique_id_1 = AtomRowToGlobal[atom1];
560 +    unique_id_2 = AtomColToGlobal[atom2];
561 +
562 +    // this situation should only arise in MPI simulations
563 +    if (unique_id_1 == unique_id_2) return true;
564 +    
565 +    // this prevents us from doing the pair on multiple processors
566 +    if (unique_id_1 < unique_id_2) {
567 +      if ((unique_id_1 + unique_id_2) % 2 == 0) return true;
568 +    } else {
569 +      if ((unique_id_1 + unique_id_2) % 2 == 1) return true;
570 +    }
571 + #else
572 +    // in the normal loop, the atom numbers are unique
573 +    unique_id_1 = atom1;
574 +    unique_id_2 = atom2;
575 + #endif
576 +    
577 + #ifdef IS_MPI
578 +    for (vector<int>::iterator i = skipsForRowAtom[atom1].begin();
579 +         i != skipsForRowAtom[atom1].end(); ++i) {
580 +      if ( (*i) == unique_id_2 ) return true;
581 +    }    
582 + #else
583 +    for (vector<int>::iterator i = skipsForLocalAtom[atom1].begin();
584 +         i != skipsForLocalAtom[atom1].end(); ++i) {
585 +      if ( (*i) == unique_id_2 ) return true;
586 +    }    
587 + #endif
588 +  }
589 +
590 +  int ForceMatrixDecomposition::getTopoDistance(int atom1, int atom2) {
591 +    
592 + #ifdef IS_MPI
593 +    for (int i = 0; i < toposForRowAtom[atom1].size(); i++) {
594 +      if ( toposForRowAtom[atom1][i] == atom2 ) return topoDistRow[atom1][i];
595 +    }
596 + #else
597 +    for (int i = 0; i < toposForLocalAtom[atom1].size(); i++) {
598 +      if ( toposForLocalAtom[atom1][i] == atom2 ) return topoDistLocal[atom1][i];
599 +    }
600 + #endif
601 +
602 +    // zero is default for unconnected (i.e. normal) pair interactions
603 +    return 0;
604 +  }
605 +
606 +  void ForceMatrixDecomposition::addForceToAtomRow(int atom1, Vector3d fg){
607 + #ifdef IS_MPI
608 +    atomRowData.force[atom1] += fg;
609 + #else
610 +    snap_->atomData.force[atom1] += fg;
611 + #endif
612 +  }
613 +
614 +  void ForceMatrixDecomposition::addForceToAtomColumn(int atom2, Vector3d fg){
615 + #ifdef IS_MPI
616 +    atomColData.force[atom2] += fg;
617 + #else
618 +    snap_->atomData.force[atom2] += fg;
619 + #endif
620 +  }
621 +
622 +    // filling interaction blocks with pointers
623 +  InteractionData ForceMatrixDecomposition::fillInteractionData(int atom1, int atom2) {    
624 +    InteractionData idat;
625 +
626 + #ifdef IS_MPI
627 +    
628 +    idat.atypes = make_pair( ff_->getAtomType(identsRow[atom1]),
629 +                             ff_->getAtomType(identsCol[atom2]) );
630 +
631 +    
632 +    if (storageLayout_ & DataStorage::dslAmat) {
633 +      idat.A1 = &(atomRowData.aMat[atom1]);
634 +      idat.A2 = &(atomColData.aMat[atom2]);
635 +    }
636 +    
637 +    if (storageLayout_ & DataStorage::dslElectroFrame) {
638 +      idat.eFrame1 = &(atomRowData.electroFrame[atom1]);
639 +      idat.eFrame2 = &(atomColData.electroFrame[atom2]);
640 +    }
641 +
642 +    if (storageLayout_ & DataStorage::dslTorque) {
643 +      idat.t1 = &(atomRowData.torque[atom1]);
644 +      idat.t2 = &(atomColData.torque[atom2]);
645 +    }
646 +
647 +    if (storageLayout_ & DataStorage::dslDensity) {
648 +      idat.rho1 = &(atomRowData.density[atom1]);
649 +      idat.rho2 = &(atomColData.density[atom2]);
650 +    }
651 +
652 +    if (storageLayout_ & DataStorage::dslFunctional) {
653 +      idat.frho1 = &(atomRowData.functional[atom1]);
654 +      idat.frho2 = &(atomColData.functional[atom2]);
655 +    }
656 +
657 +    if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
658 +      idat.dfrho1 = &(atomRowData.functionalDerivative[atom1]);
659 +      idat.dfrho2 = &(atomColData.functionalDerivative[atom2]);
660 +    }
661 +
662 +    if (storageLayout_ & DataStorage::dslParticlePot) {
663 +      idat.particlePot1 = &(atomRowData.particlePot[atom1]);
664 +      idat.particlePot2 = &(atomColData.particlePot[atom2]);
665 +    }
666 +
667 + #else
668 +
669 +    idat.atypes = make_pair( ff_->getAtomType(identsLocal[atom1]),
670 +                             ff_->getAtomType(identsLocal[atom2]) );
671 +
672 +    if (storageLayout_ & DataStorage::dslAmat) {
673 +      idat.A1 = &(snap_->atomData.aMat[atom1]);
674 +      idat.A2 = &(snap_->atomData.aMat[atom2]);
675 +    }
676 +
677 +    if (storageLayout_ & DataStorage::dslElectroFrame) {
678 +      idat.eFrame1 = &(snap_->atomData.electroFrame[atom1]);
679 +      idat.eFrame2 = &(snap_->atomData.electroFrame[atom2]);
680 +    }
681 +
682 +    if (storageLayout_ & DataStorage::dslTorque) {
683 +      idat.t1 = &(snap_->atomData.torque[atom1]);
684 +      idat.t2 = &(snap_->atomData.torque[atom2]);
685 +    }
686 +
687 +    if (storageLayout_ & DataStorage::dslDensity) {
688 +      idat.rho1 = &(snap_->atomData.density[atom1]);
689 +      idat.rho2 = &(snap_->atomData.density[atom2]);
690 +    }
691 +
692 +    if (storageLayout_ & DataStorage::dslFunctional) {
693 +      idat.frho1 = &(snap_->atomData.functional[atom1]);
694 +      idat.frho2 = &(snap_->atomData.functional[atom2]);
695 +    }
696 +
697 +    if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
698 +      idat.dfrho1 = &(snap_->atomData.functionalDerivative[atom1]);
699 +      idat.dfrho2 = &(snap_->atomData.functionalDerivative[atom2]);
700 +    }
701 +
702 +    if (storageLayout_ & DataStorage::dslParticlePot) {
703 +      idat.particlePot1 = &(snap_->atomData.particlePot[atom1]);
704 +      idat.particlePot2 = &(snap_->atomData.particlePot[atom2]);
705 +    }
706 +
707 + #endif
708 +    return idat;
709 +  }
710 +
711 +  
712 +  void ForceMatrixDecomposition::unpackInteractionData(InteractionData idat, int atom1, int atom2) {    
713 + #ifdef IS_MPI
714 +    pot_row[atom1] += 0.5 *  *(idat.pot);
715 +    pot_col[atom2] += 0.5 *  *(idat.pot);
716 +
717 +    atomRowData.force[atom1] += *(idat.f1);
718 +    atomColData.force[atom2] -= *(idat.f1);
719 + #else
720 +    longRangePot_ += *(idat.pot);
721 +    
722 +    snap_->atomData.force[atom1] += *(idat.f1);
723 +    snap_->atomData.force[atom2] -= *(idat.f1);
724 + #endif
725 +
726 +  }
727 +
728 +
729 +  InteractionData ForceMatrixDecomposition::fillSkipData(int atom1, int atom2){
730 +
731 +    InteractionData idat;
732 + #ifdef IS_MPI
733 +    idat.atypes = make_pair( ff_->getAtomType(identsRow[atom1]),
734 +                             ff_->getAtomType(identsCol[atom2]) );
735 +
736 +    if (storageLayout_ & DataStorage::dslElectroFrame) {
737 +      idat.eFrame1 = &(atomRowData.electroFrame[atom1]);
738 +      idat.eFrame2 = &(atomColData.electroFrame[atom2]);
739 +    }
740 +    if (storageLayout_ & DataStorage::dslTorque) {
741 +      idat.t1 = &(atomRowData.torque[atom1]);
742 +      idat.t2 = &(atomColData.torque[atom2]);
743 +    }
744 + #else
745 +    idat.atypes = make_pair( ff_->getAtomType(identsLocal[atom1]),
746 +                             ff_->getAtomType(identsLocal[atom2]) );
747 +
748 +    if (storageLayout_ & DataStorage::dslElectroFrame) {
749 +      idat.eFrame1 = &(snap_->atomData.electroFrame[atom1]);
750 +      idat.eFrame2 = &(snap_->atomData.electroFrame[atom2]);
751 +    }
752 +    if (storageLayout_ & DataStorage::dslTorque) {
753 +      idat.t1 = &(snap_->atomData.torque[atom1]);
754 +      idat.t2 = &(snap_->atomData.torque[atom2]);
755 +    }
756 + #endif    
757 +  }
758 +
759 +  /*
760 +   * buildNeighborList
761 +   *
762 +   * first element of pair is row-indexed CutoffGroup
763 +   * second element of pair is column-indexed CutoffGroup
764 +   */
765 +  vector<pair<int, int> > ForceMatrixDecomposition::buildNeighborList() {
766 +      
767 +    vector<pair<int, int> > neighborList;
768 + #ifdef IS_MPI
769 +    cellListRow_.clear();
770 +    cellListCol_.clear();
771 + #else
772 +    cellList_.clear();
773 + #endif
774 +
775 +    // dangerous to not do error checking.
776 +    RealType rCut_;
777 +
778 +    RealType rList_ = (rCut_ + skinThickness_);
779 +    RealType rl2 = rList_ * rList_;
780 +    Snapshot* snap_ = sman_->getCurrentSnapshot();
781 +    Mat3x3d Hmat = snap_->getHmat();
782 +    Vector3d Hx = Hmat.getColumn(0);
783 +    Vector3d Hy = Hmat.getColumn(1);
784 +    Vector3d Hz = Hmat.getColumn(2);
785 +
786 +    nCells_.x() = (int) ( Hx.length() )/ rList_;
787 +    nCells_.y() = (int) ( Hy.length() )/ rList_;
788 +    nCells_.z() = (int) ( Hz.length() )/ rList_;
789 +
790 +    Mat3x3d invHmat = snap_->getInvHmat();
791 +    Vector3d rs, scaled, dr;
792 +    Vector3i whichCell;
793 +    int cellIndex;
794 +
795 + #ifdef IS_MPI
796 +    for (int i = 0; i < nGroupsInRow_; i++) {
797 +      rs = cgRowData.position[i];
798 +      // scaled positions relative to the box vectors
799 +      scaled = invHmat * rs;
800 +      // wrap the vector back into the unit box by subtracting integer box
801 +      // numbers
802 +      for (int j = 0; j < 3; j++)
803 +        scaled[j] -= roundMe(scaled[j]);
804 +    
805 +      // find xyz-indices of cell that cutoffGroup is in.
806 +      whichCell.x() = nCells_.x() * scaled.x();
807 +      whichCell.y() = nCells_.y() * scaled.y();
808 +      whichCell.z() = nCells_.z() * scaled.z();
809 +
810 +      // find single index of this cell:
811 +      cellIndex = Vlinear(whichCell, nCells_);
812 +      // add this cutoff group to the list of groups in this cell;
813 +      cellListRow_[cellIndex].push_back(i);
814 +    }
815 +
816 +    for (int i = 0; i < nGroupsInCol_; i++) {
817 +      rs = cgColData.position[i];
818 +      // scaled positions relative to the box vectors
819 +      scaled = invHmat * rs;
820 +      // wrap the vector back into the unit box by subtracting integer box
821 +      // numbers
822 +      for (int j = 0; j < 3; j++)
823 +        scaled[j] -= roundMe(scaled[j]);
824 +
825 +      // find xyz-indices of cell that cutoffGroup is in.
826 +      whichCell.x() = nCells_.x() * scaled.x();
827 +      whichCell.y() = nCells_.y() * scaled.y();
828 +      whichCell.z() = nCells_.z() * scaled.z();
829 +
830 +      // find single index of this cell:
831 +      cellIndex = Vlinear(whichCell, nCells_);
832 +      // add this cutoff group to the list of groups in this cell;
833 +      cellListCol_[cellIndex].push_back(i);
834 +    }
835 + #else
836 +    for (int i = 0; i < nGroups_; i++) {
837 +      rs = snap_->cgData.position[i];
838 +      // scaled positions relative to the box vectors
839 +      scaled = invHmat * rs;
840 +      // wrap the vector back into the unit box by subtracting integer box
841 +      // numbers
842 +      for (int j = 0; j < 3; j++)
843 +        scaled[j] -= roundMe(scaled[j]);
844 +
845 +      // find xyz-indices of cell that cutoffGroup is in.
846 +      whichCell.x() = nCells_.x() * scaled.x();
847 +      whichCell.y() = nCells_.y() * scaled.y();
848 +      whichCell.z() = nCells_.z() * scaled.z();
849 +
850 +      // find single index of this cell:
851 +      cellIndex = Vlinear(whichCell, nCells_);
852 +      // add this cutoff group to the list of groups in this cell;
853 +      cellList_[cellIndex].push_back(i);
854 +    }
855 + #endif
856 +
857 +    for (int m1z = 0; m1z < nCells_.z(); m1z++) {
858 +      for (int m1y = 0; m1y < nCells_.y(); m1y++) {
859 +        for (int m1x = 0; m1x < nCells_.x(); m1x++) {
860 +          Vector3i m1v(m1x, m1y, m1z);
861 +          int m1 = Vlinear(m1v, nCells_);
862 +
863 +          for (vector<Vector3i>::iterator os = cellOffsets_.begin();
864 +               os != cellOffsets_.end(); ++os) {
865 +            
866 +            Vector3i m2v = m1v + (*os);
867 +            
868 +            if (m2v.x() >= nCells_.x()) {
869 +              m2v.x() = 0;          
870 +            } else if (m2v.x() < 0) {
871 +              m2v.x() = nCells_.x() - 1;
872 +            }
873 +            
874 +            if (m2v.y() >= nCells_.y()) {
875 +              m2v.y() = 0;          
876 +            } else if (m2v.y() < 0) {
877 +              m2v.y() = nCells_.y() - 1;
878 +            }
879 +            
880 +            if (m2v.z() >= nCells_.z()) {
881 +              m2v.z() = 0;          
882 +            } else if (m2v.z() < 0) {
883 +              m2v.z() = nCells_.z() - 1;
884 +            }
885 +            
886 +            int m2 = Vlinear (m2v, nCells_);
887 +
888 + #ifdef IS_MPI
889 +            for (vector<int>::iterator j1 = cellListRow_[m1].begin();
890 +                 j1 != cellListRow_[m1].end(); ++j1) {
891 +              for (vector<int>::iterator j2 = cellListCol_[m2].begin();
892 +                   j2 != cellListCol_[m2].end(); ++j2) {
893 +                              
894 +                // Always do this if we're in different cells or if
895 +                // we're in the same cell and the global index of the
896 +                // j2 cutoff group is less than the j1 cutoff group
897 +
898 +                if (m2 != m1 || cgColToGlobal[(*j2)] < cgRowToGlobal[(*j1)]) {
899 +                  dr = cgColData.position[(*j2)] - cgRowData.position[(*j1)];
900 +                  snap_->wrapVector(dr);
901 +                  if (dr.lengthSquare() < rl2) {
902 +                    neighborList.push_back(make_pair((*j1), (*j2)));
903 +                  }
904 +                }
905 +              }
906 +            }
907 + #else
908 +            for (vector<int>::iterator j1 = cellList_[m1].begin();
909 +                 j1 != cellList_[m1].end(); ++j1) {
910 +              for (vector<int>::iterator j2 = cellList_[m2].begin();
911 +                   j2 != cellList_[m2].end(); ++j2) {
912 +                              
913 +                // Always do this if we're in different cells or if
914 +                // we're in the same cell and the global index of the
915 +                // j2 cutoff group is less than the j1 cutoff group
916 +
917 +                if (m2 != m1 || (*j2) < (*j1)) {
918 +                  dr = snap_->cgData.position[(*j2)] - snap_->cgData.position[(*j1)];
919 +                  snap_->wrapVector(dr);
920 +                  if (dr.lengthSquare() < rl2) {
921 +                    neighborList.push_back(make_pair((*j1), (*j2)));
922 +                  }
923 +                }
924 +              }
925 +            }
926 + #endif
927 +          }
928 +        }
929 +      }
930 +    }
931 +
932 +    // save the local cutoff group positions for the check that is
933 +    // done on each loop:
934 +    saved_CG_positions_.clear();
935 +    for (int i = 0; i < nGroups_; i++)
936 +      saved_CG_positions_.push_back(snap_->cgData.position[i]);
937 +
938 +    return neighborList;
939 +  }
940 + } //end namespace OpenMD

Diff Legend

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