[ViewVC] Diff of: OpenMD/trunk/src/parallel/ForceMatrixDecomposition.cpp

Comparing:
branches/development/src/parallel/ForceDecomposition.cpp (file contents), Revision 1540 by gezelter, Mon Jan 17 21:34:36 2011 UTC vs.
branches/development/src/parallel/ForceMatrixDecomposition.cpp (file contents), Revision 1612 by gezelter, Fri Aug 12 19:59:56 2011 UTC

+ * [3]  Sun, Lin & Gezelter, J. Chem. Phys. 128, 24107 (2008).
+ * [4]  Vardeman & Gezelter, in progress (2009).
+ */
-<
+#include "parallel/ForceDecomposition.hpp"
-<
+#include "parallel/Communicator.hpp"
->
+#include "parallel/ForceMatrixDecomposition.hpp"
+#include "math/SquareMatrix3.hpp"
-+
+#include "nonbonded/NonBondedInteraction.hpp"
-+
+#include "brains/SnapshotManager.hpp"
-+
+#include "brains/PairList.hpp"
-+
+using namespace std;
+namespace OpenMD {
-<
+  void ForceDecomposition::distributeInitialData() {
->
+  ForceMatrixDecomposition::ForceMatrixDecomposition(SimInfo* info, InteractionManager* iMan) : ForceDecomposition(info, iMan) {
->
->
+    // In a parallel computation, row and colum scans must visit all
->
+    // surrounding cells (not just the 14 upper triangular blocks that
->
+    // are used when the processor can see all pairs)
+#ifdef IS_MPI
-+
+    cellOffsets_.clear();
-+
+    cellOffsets_.push_back( Vector3i(-1,-1,-1) );
-+
+    cellOffsets_.push_back( Vector3i( 0,-1,-1) );
-+
+    cellOffsets_.push_back( Vector3i( 1,-1,-1) );
-+
+    cellOffsets_.push_back( Vector3i(-1, 0,-1) );
-+
+    cellOffsets_.push_back( Vector3i( 0, 0,-1) );
-+
+    cellOffsets_.push_back( Vector3i( 1, 0,-1) );
-+
+    cellOffsets_.push_back( Vector3i(-1, 1,-1) );
-+
+    cellOffsets_.push_back( Vector3i( 0, 1,-1) );
-+
+    cellOffsets_.push_back( Vector3i( 1, 1,-1) );
-+
+    cellOffsets_.push_back( Vector3i(-1,-1, 0) );
-+
+    cellOffsets_.push_back( Vector3i( 0,-1, 0) );
-+
+    cellOffsets_.push_back( Vector3i( 1,-1, 0) );
-+
+    cellOffsets_.push_back( Vector3i(-1, 0, 0) );
-+
+    cellOffsets_.push_back( Vector3i( 0, 0, 0) );
-+
+    cellOffsets_.push_back( Vector3i( 1, 0, 0) );
-+
+    cellOffsets_.push_back( Vector3i(-1, 1, 0) );
-+
+    cellOffsets_.push_back( Vector3i( 0, 1, 0) );
-+
+    cellOffsets_.push_back( Vector3i( 1, 1, 0) );
-+
+    cellOffsets_.push_back( Vector3i(-1,-1, 1) );
-+
+    cellOffsets_.push_back( Vector3i( 0,-1, 1) );
-+
+    cellOffsets_.push_back( Vector3i( 1,-1, 1) );
-+
+    cellOffsets_.push_back( Vector3i(-1, 0, 1) );
-+
+    cellOffsets_.push_back( Vector3i( 0, 0, 1) );
-+
+    cellOffsets_.push_back( Vector3i( 1, 0, 1) );
-+
+    cellOffsets_.push_back( Vector3i(-1, 1, 1) );
-+
+    cellOffsets_.push_back( Vector3i( 0, 1, 1) );
-+
+    cellOffsets_.push_back( Vector3i( 1, 1, 1) );
-+
+#endif
-+
+  }
-–
+    int nAtoms;
-–
+    int nGroups;
-<
+    AtomCommRealI = new Communicator<Row,RealType>(nAtoms);
-<
+    AtomCommVectorI = new Communicator<Row,Vector3d>(nAtoms);
-<
+    AtomCommMatrixI = new Communicator<Row,Mat3x3d>(nAtoms);
->
+  /**
->
+   * distributeInitialData is essentially a copy of the older fortran
->
+   * SimulationSetup
->
+   */
->
+  void ForceMatrixDecomposition::distributeInitialData() {
->
+    snap_ = sman_->getCurrentSnapshot();
->
+    storageLayout_ = sman_->getStorageLayout();
->
+    ff_ = info_->getForceField();
->
+    nLocal_ = snap_->getNumberOfAtoms();
->
->
+    nGroups_ = info_->getNLocalCutoffGroups();
->
+    // gather the information for atomtype IDs (atids):
->
+    idents = info_->getIdentArray();
->
+    AtomLocalToGlobal = info_->getGlobalAtomIndices();
->
+    cgLocalToGlobal = info_->getGlobalGroupIndices();
->
+    vector<int> globalGroupMembership = info_->getGlobalGroupMembership();
-<
+    AtomCommRealJ = new Communicator<Column,RealType>(nAtoms);
-<
+    AtomCommVectorJ = new Communicator<Column,Vector3d>(nAtoms);
-<
+    AtomCommMatrixJ = new Communicator<Column,Mat3x3d>(nAtoms);
->
+    massFactors = info_->getMassFactors();
-<
+    cgCommVectorI = new Communicator<Row,Vector3d>(nGroups);
-<
+    cgCommVectorJ = new Communicator<Column,Vector3d>(nGroups);
-<
+    // more to come
->
+    PairList* excludes = info_->getExcludedInteractions();
->
+    PairList* oneTwo = info_->getOneTwoInteractions();
->
+    PairList* oneThree = info_->getOneThreeInteractions();
->
+    PairList* oneFour = info_->getOneFourInteractions();
->
->
+#ifdef IS_MPI
->
->
+    MPI::Intracomm row = rowComm.getComm();
->
+    MPI::Intracomm col = colComm.getComm();
->
->
+    AtomPlanIntRow = new Plan<int>(row, nLocal_);
->
+    AtomPlanRealRow = new Plan<RealType>(row, nLocal_);
->
+    AtomPlanVectorRow = new Plan<Vector3d>(row, nLocal_);
->
+    AtomPlanMatrixRow = new Plan<Mat3x3d>(row, nLocal_);
->
+    AtomPlanPotRow = new Plan<potVec>(row, nLocal_);
->
->
+    AtomPlanIntColumn = new Plan<int>(col, nLocal_);
->
+    AtomPlanRealColumn = new Plan<RealType>(col, nLocal_);
->
+    AtomPlanVectorColumn = new Plan<Vector3d>(col, nLocal_);
->
+    AtomPlanMatrixColumn = new Plan<Mat3x3d>(col, nLocal_);
->
+    AtomPlanPotColumn = new Plan<potVec>(col, nLocal_);
->
->
+    cgPlanIntRow = new Plan<int>(row, nGroups_);
->
+    cgPlanVectorRow = new Plan<Vector3d>(row, nGroups_);
->
+    cgPlanIntColumn = new Plan<int>(col, nGroups_);
->
+    cgPlanVectorColumn = new Plan<Vector3d>(col, nGroups_);
->
->
+    nAtomsInRow_ = AtomPlanIntRow->getSize();
->
+    nAtomsInCol_ = AtomPlanIntColumn->getSize();
->
+    nGroupsInRow_ = cgPlanIntRow->getSize();
->
+    nGroupsInCol_ = cgPlanIntColumn->getSize();
->
->
+    // Modify the data storage objects with the correct layouts and sizes:
->
+    atomRowData.resize(nAtomsInRow_);
->
+    atomRowData.setStorageLayout(storageLayout_);
->
+    atomColData.resize(nAtomsInCol_);
->
+    atomColData.setStorageLayout(storageLayout_);
->
+    cgRowData.resize(nGroupsInRow_);
->
+    cgRowData.setStorageLayout(DataStorage::dslPosition);
->
+    cgColData.resize(nGroupsInCol_);
->
+    cgColData.setStorageLayout(DataStorage::dslPosition);
->
->
+    identsRow.resize(nAtomsInRow_);
->
+    identsCol.resize(nAtomsInCol_);
->
->
+    AtomPlanIntRow->gather(idents, identsRow);
->
+    AtomPlanIntColumn->gather(idents, identsCol);
->
->
+    // allocate memory for the parallel objects
->
+    atypesRow.resize(nAtomsInRow_);
->
+    atypesCol.resize(nAtomsInCol_);
->
->
+    for (int i = 0; i < nAtomsInRow_; i++)
->
+      atypesRow[i] = ff_->getAtomType(identsRow[i]);
->
+    for (int i = 0; i < nAtomsInCol_; i++)
->
+      atypesCol[i] = ff_->getAtomType(identsCol[i]);
->
->
+    pot_row.resize(nAtomsInRow_);
->
+    pot_col.resize(nAtomsInCol_);
->
->
+    AtomRowToGlobal.resize(nAtomsInRow_);
->
+    AtomColToGlobal.resize(nAtomsInCol_);
->
+    AtomPlanIntRow->gather(AtomLocalToGlobal, AtomRowToGlobal);
->
+    AtomPlanIntColumn->gather(AtomLocalToGlobal, AtomColToGlobal);
->
->
+    cgRowToGlobal.resize(nGroupsInRow_);
->
+    cgColToGlobal.resize(nGroupsInCol_);
->
+    cgPlanIntRow->gather(cgLocalToGlobal, cgRowToGlobal);
->
+    cgPlanIntColumn->gather(cgLocalToGlobal, cgColToGlobal);
->
->
+    massFactorsRow.resize(nAtomsInRow_);
->
+    massFactorsCol.resize(nAtomsInCol_);
->
+    AtomPlanRealRow->gather(massFactors, massFactorsRow);
->
+    AtomPlanRealColumn->gather(massFactors, massFactorsCol);
->
->
+    groupListRow_.clear();
->
+    groupListRow_.resize(nGroupsInRow_);
->
+    for (int i = 0; i < nGroupsInRow_; i++) {
->
+      int gid = cgRowToGlobal[i];
->
+      for (int j = 0; j < nAtomsInRow_; j++) {
->
+        int aid = AtomRowToGlobal[j];
->
+        if (globalGroupMembership[aid] == gid)
->
+          groupListRow_[i].push_back(j);
->
+      }
->
+    }
->
->
+    groupListCol_.clear();
->
+    groupListCol_.resize(nGroupsInCol_);
->
+    for (int i = 0; i < nGroupsInCol_; i++) {
->
+      int gid = cgColToGlobal[i];
->
+      for (int j = 0; j < nAtomsInCol_; j++) {
->
+        int aid = AtomColToGlobal[j];
->
+        if (globalGroupMembership[aid] == gid)
->
+          groupListCol_[i].push_back(j);
->
+      }
->
+    }
->
->
+    excludesForAtom.clear();
->
+    excludesForAtom.resize(nAtomsInRow_);
->
+    toposForAtom.clear();
->
+    toposForAtom.resize(nAtomsInRow_);
->
+    topoDist.clear();
->
+    topoDist.resize(nAtomsInRow_);
->
+    for (int i = 0; i < nAtomsInRow_; i++) {
->
+      int iglob = AtomRowToGlobal[i];
->
->
+      for (int j = 0; j < nAtomsInCol_; j++) {
->
+        int jglob = AtomColToGlobal[j];
->
->
+        if (excludes->hasPair(iglob, jglob))
->
+          excludesForAtom[i].push_back(j);
->
->
+        if (oneTwo->hasPair(iglob, jglob)) {
->
+          toposForAtom[i].push_back(j);
->
+          topoDist[i].push_back(1);
->
+        } else {
->
+          if (oneThree->hasPair(iglob, jglob)) {
->
+            toposForAtom[i].push_back(j);
->
+            topoDist[i].push_back(2);
->
+          } else {
->
+            if (oneFour->hasPair(iglob, jglob)) {
->
+              toposForAtom[i].push_back(j);
->
+              topoDist[i].push_back(3);
->
+            }
->
+          }
->
+        }
->
+      }
->
+    }
->
+#endif
-+
-+
+    // allocate memory for the parallel objects
-+
+    atypesLocal.resize(nLocal_);
-+
-+
+    for (int i = 0; i < nLocal_; i++)
-+
+      atypesLocal[i] = ff_->getAtomType(idents[i]);
-+
-+
+    groupList_.clear();
-+
+    groupList_.resize(nGroups_);
-+
+    for (int i = 0; i < nGroups_; i++) {
-+
+      int gid = cgLocalToGlobal[i];
-+
+      for (int j = 0; j < nLocal_; j++) {
-+
+        int aid = AtomLocalToGlobal[j];
-+
+        if (globalGroupMembership[aid] == gid) {
-+
+          groupList_[i].push_back(j);
-+
+        }
-+
+      }
-+
+    }
-+
-+
+    excludesForAtom.clear();
-+
+    excludesForAtom.resize(nLocal_);
-+
+    toposForAtom.clear();
-+
+    toposForAtom.resize(nLocal_);
-+
+    topoDist.clear();
-+
+    topoDist.resize(nLocal_);
-+
-+
+    for (int i = 0; i < nLocal_; i++) {
-+
+      int iglob = AtomLocalToGlobal[i];
-+
-+
+      for (int j = 0; j < nLocal_; j++) {
-+
+        int jglob = AtomLocalToGlobal[j];
-+
-+
+        if (excludes->hasPair(iglob, jglob))
-+
+          excludesForAtom[i].push_back(j);
-+
-+
+        if (oneTwo->hasPair(iglob, jglob)) {
-+
+          toposForAtom[i].push_back(j);
-+
+          topoDist[i].push_back(1);
-+
+        } else {
-+
+          if (oneThree->hasPair(iglob, jglob)) {
-+
+            toposForAtom[i].push_back(j);
-+
+            topoDist[i].push_back(2);
-+
+          } else {
-+
+            if (oneFour->hasPair(iglob, jglob)) {
-+
+              toposForAtom[i].push_back(j);
-+
+              topoDist[i].push_back(3);
-+
+            }
-+
+          }
-+
+        }
-+
+      }
-+
+    }
-+
-+
+    createGtypeCutoffMap();
-+
-+
-+
+  void ForceMatrixDecomposition::createGtypeCutoffMap() {
-+
+    RealType tol = 1e-6;
-+
+    largestRcut_ = 0.0;
-+
+    RealType rc;
-+
+    int atid;
-+
+    set<AtomType*> atypes = info_->getSimulatedAtomTypes();
-+
-+
+    map<int, RealType> atypeCutoff;
-+
-+
+    for (set<AtomType*>::iterator at = atypes.begin();
-+
+         at != atypes.end(); ++at){
-+
+      atid = (*at)->getIdent();
-+
+      if (userChoseCutoff_)
-+
+        atypeCutoff[atid] = userCutoff_;
-+
+      else
-+
+        atypeCutoff[atid] = interactionMan_->getSuggestedCutoffRadius(*at);
-+
+    }
-+
-+
+    vector<RealType> gTypeCutoffs;
-+
+    // first we do a single loop over the cutoff groups to find the
-+
+    // largest cutoff for any atypes present in this group.
-+
+#ifdef IS_MPI
-+
+    vector<RealType> groupCutoffRow(nGroupsInRow_, 0.0);
-+
+    groupRowToGtype.resize(nGroupsInRow_);
-+
+    for (int cg1 = 0; cg1 < nGroupsInRow_; cg1++) {
-+
+      vector<int> atomListRow = getAtomsInGroupRow(cg1);
-+
+      for (vector<int>::iterator ia = atomListRow.begin();
-+
+           ia != atomListRow.end(); ++ia) {
-+
+        int atom1 = (*ia);
-+
+        atid = identsRow[atom1];
-+
+        if (atypeCutoff[atid] > groupCutoffRow[cg1]) {
-+
+          groupCutoffRow[cg1] = atypeCutoff[atid];
-+
+        }
-+
+      }
-+
+      bool gTypeFound = false;
-+
+      for (int gt = 0; gt < gTypeCutoffs.size(); gt++) {
-+
+        if (abs(groupCutoffRow[cg1] - gTypeCutoffs[gt]) < tol) {
-+
+          groupRowToGtype[cg1] = gt;
-+
+          gTypeFound = true;
-+
+        }
-+
+      }
-+
+      if (!gTypeFound) {
-+
+        gTypeCutoffs.push_back( groupCutoffRow[cg1] );
-+
+        groupRowToGtype[cg1] = gTypeCutoffs.size() - 1;
-+
+      }
-+
-+
+    }
-+
+    vector<RealType> groupCutoffCol(nGroupsInCol_, 0.0);
-+
+    groupColToGtype.resize(nGroupsInCol_);
-+
+    for (int cg2 = 0; cg2 < nGroupsInCol_; cg2++) {
-+
+      vector<int> atomListCol = getAtomsInGroupColumn(cg2);
-+
+      for (vector<int>::iterator jb = atomListCol.begin();
-+
+           jb != atomListCol.end(); ++jb) {
-+
+        int atom2 = (*jb);
-+
+        atid = identsCol[atom2];
-+
+        if (atypeCutoff[atid] > groupCutoffCol[cg2]) {
-+
+          groupCutoffCol[cg2] = atypeCutoff[atid];
-+
+        }
-+
+      }
-+
+      bool gTypeFound = false;
-+
+      for (int gt = 0; gt < gTypeCutoffs.size(); gt++) {
-+
+        if (abs(groupCutoffCol[cg2] - gTypeCutoffs[gt]) < tol) {
-+
+          groupColToGtype[cg2] = gt;
-+
+          gTypeFound = true;
-+
+        }
-+
+      }
-+
+      if (!gTypeFound) {
-+
+        gTypeCutoffs.push_back( groupCutoffCol[cg2] );
-+
+        groupColToGtype[cg2] = gTypeCutoffs.size() - 1;
-+
+      }
-+
+    }
-+
+#else
-<
+  void ForceDecomposition::distributeData()  {
->
+    vector<RealType> groupCutoff(nGroups_, 0.0);
->
+    groupToGtype.resize(nGroups_);
->
+    for (int cg1 = 0; cg1 < nGroups_; cg1++) {
->
+      groupCutoff[cg1] = 0.0;
->
+      vector<int> atomList = getAtomsInGroupRow(cg1);
->
+      for (vector<int>::iterator ia = atomList.begin();
->
+           ia != atomList.end(); ++ia) {
->
+        int atom1 = (*ia);
->
+        atid = idents[atom1];
->
+        if (atypeCutoff[atid] > groupCutoff[cg1])
->
+          groupCutoff[cg1] = atypeCutoff[atid];
->
+      }
->
->
+      bool gTypeFound = false;
->
+      for (int gt = 0; gt < gTypeCutoffs.size(); gt++) {
->
+        if (abs(groupCutoff[cg1] - gTypeCutoffs[gt]) < tol) {
->
+          groupToGtype[cg1] = gt;
->
+          gTypeFound = true;
->
+        }
->
+      }
->
+      if (!gTypeFound) {
->
+        gTypeCutoffs.push_back( groupCutoff[cg1] );
->
+        groupToGtype[cg1] = gTypeCutoffs.size() - 1;
->
+      }
->
+    }
->
+#endif
->
->
+    // Now we find the maximum group cutoff value present in the simulation
->
->
+    RealType groupMax = *max_element(gTypeCutoffs.begin(),
->
+                                     gTypeCutoffs.end());
->
+#ifdef IS_MPI
-<
+    Snapshot* snap = sman_->getCurrentSnapshot();
->
+    MPI::COMM_WORLD.Allreduce(&groupMax, &groupMax, 1, MPI::REALTYPE,
->
+                              MPI::MAX);
->
+#endif
-+
+    RealType tradRcut = groupMax;
-+
-+
+    for (int i = 0; i < gTypeCutoffs.size();  i++) {
-+
+      for (int j = 0; j < gTypeCutoffs.size();  j++) {
-+
+        RealType thisRcut;
-+
+        switch(cutoffPolicy_) {
-+
+        case TRADITIONAL:
-+
+          thisRcut = tradRcut;
-+
+          break;
-+
+        case MIX:
-+
+          thisRcut = 0.5 * (gTypeCutoffs[i] + gTypeCutoffs[j]);
-+
+          break;
-+
+        case MAX:
-+
+          thisRcut = max(gTypeCutoffs[i], gTypeCutoffs[j]);
-+
+          break;
-+
+        default:
-+
+          sprintf(painCave.errMsg,
-+
+                  "ForceMatrixDecomposition::createGtypeCutoffMap "
-+
+                  "hit an unknown cutoff policy!\n");
-+
+          painCave.severity = OPENMD_ERROR;
-+
+          painCave.isFatal = 1;
-+
+          simError();
-+
+          break;
-+
+        }
-+
-+
+        pair<int,int> key = make_pair(i,j);
-+
+        gTypeCutoffMap[key].first = thisRcut;
-+
+        if (thisRcut > largestRcut_) largestRcut_ = thisRcut;
-+
+        gTypeCutoffMap[key].second = thisRcut*thisRcut;
-+
+        gTypeCutoffMap[key].third = pow(thisRcut + skinThickness_, 2);
-+
+        // sanity check
-+
-+
+        if (userChoseCutoff_) {
-+
+          if (abs(gTypeCutoffMap[key].first - userCutoff_) > 0.0001) {
-+
+            sprintf(painCave.errMsg,
-+
+                    "ForceMatrixDecomposition::createGtypeCutoffMap "
-+
+                    "user-specified rCut (%lf) does not match computed group Cutoff\n", userCutoff_);
-+
+            painCave.severity = OPENMD_ERROR;
-+
+            painCave.isFatal = 1;
-+
+            simError();
-+
+          }
-+
+        }
-+
+      }
-+
+    }
-+
+  }
-+
-+
-+
+  groupCutoffs ForceMatrixDecomposition::getGroupCutoffs(int cg1, int cg2) {
-+
+    int i, j;
-+
+#ifdef IS_MPI
-+
+    i = groupRowToGtype[cg1];
-+
+    j = groupColToGtype[cg2];
-+
+#else
-+
+    i = groupToGtype[cg1];
-+
+    j = groupToGtype[cg2];
-+
+#endif
-+
+    return gTypeCutoffMap[make_pair(i,j)];
-+
+  }
-+
-+
+  int ForceMatrixDecomposition::getTopologicalDistance(int atom1, int atom2) {
-+
+    for (int j = 0; j < toposForAtom[atom1].size(); j++) {
-+
+      if (toposForAtom[atom1][j] == atom2)
-+
+        return topoDist[atom1][j];
-+
+    }
-+
+    return 0;
-+
+  }
-+
-+
+  void ForceMatrixDecomposition::zeroWorkArrays() {
-+
+    pairwisePot = 0.0;
-+
+    embeddingPot = 0.0;
-+
-+
+#ifdef IS_MPI
-+
+    if (storageLayout_ & DataStorage::dslForce) {
-+
+      fill(atomRowData.force.begin(), atomRowData.force.end(), V3Zero);
-+
+      fill(atomColData.force.begin(), atomColData.force.end(), V3Zero);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslTorque) {
-+
+      fill(atomRowData.torque.begin(), atomRowData.torque.end(), V3Zero);
-+
+      fill(atomColData.torque.begin(), atomColData.torque.end(), V3Zero);
-+
+    }
-+
-+
+    fill(pot_row.begin(), pot_row.end(),
-+
+         Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
-+
-+
+    fill(pot_col.begin(), pot_col.end(),
-+
+         Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
-+
-+
+    if (storageLayout_ & DataStorage::dslParticlePot) {
-+
+      fill(atomRowData.particlePot.begin(), atomRowData.particlePot.end(),
-+
+.0);
-+
+      fill(atomColData.particlePot.begin(), atomColData.particlePot.end(),
-+
+.0);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslDensity) {
-+
+      fill(atomRowData.density.begin(), atomRowData.density.end(), 0.0);
-+
+      fill(atomColData.density.begin(), atomColData.density.end(), 0.0);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslFunctional) {
-+
+      fill(atomRowData.functional.begin(), atomRowData.functional.end(),
-+
+.0);
-+
+      fill(atomColData.functional.begin(), atomColData.functional.end(),
-+
+.0);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
-+
+      fill(atomRowData.functionalDerivative.begin(),
-+
+           atomRowData.functionalDerivative.end(), 0.0);
-+
+      fill(atomColData.functionalDerivative.begin(),
-+
+           atomColData.functionalDerivative.end(), 0.0);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslSkippedCharge) {
-+
+      fill(atomRowData.skippedCharge.begin(),
-+
+           atomRowData.skippedCharge.end(), 0.0);
-+
+      fill(atomColData.skippedCharge.begin(),
-+
+           atomColData.skippedCharge.end(), 0.0);
-+
+    }
-+
-+
+#endif
-+
+    // even in parallel, we need to zero out the local arrays:
-+
-+
+    if (storageLayout_ & DataStorage::dslParticlePot) {
-+
+      fill(snap_->atomData.particlePot.begin(),
-+
+           snap_->atomData.particlePot.end(), 0.0);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslDensity) {
-+
+      fill(snap_->atomData.density.begin(),
-+
+           snap_->atomData.density.end(), 0.0);
-+
+    }
-+
+    if (storageLayout_ & DataStorage::dslFunctional) {
-+
+      fill(snap_->atomData.functional.begin(),
-+
+           snap_->atomData.functional.end(), 0.0);
-+
+    }
-+
+    if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
-+
+      fill(snap_->atomData.functionalDerivative.begin(),
-+
+           snap_->atomData.functionalDerivative.end(), 0.0);
-+
+    }
-+
+    if (storageLayout_ & DataStorage::dslSkippedCharge) {
-+
+      fill(snap_->atomData.skippedCharge.begin(),
-+
+           snap_->atomData.skippedCharge.end(), 0.0);
-+
+    }
-+
-+
+  }
-+
-+
-+
+  void ForceMatrixDecomposition::distributeData()  {
-+
+    snap_ = sman_->getCurrentSnapshot();
-+
+    storageLayout_ = sman_->getStorageLayout();
-+
+#ifdef IS_MPI
-+
+    // gather up the atomic positions
-<
+    AtomCommVectorI->gather(snap->atomData.position,
-<
+                            snap->atomIData.position);
-<
+    AtomCommVectorJ->gather(snap->atomData.position,
-<
+                            snap->atomJData.position);
->
+    AtomPlanVectorRow->gather(snap_->atomData.position,
->
+                              atomRowData.position);
->
+    AtomPlanVectorColumn->gather(snap_->atomData.position,
->
+                                 atomColData.position);
+    // gather up the cutoff group positions
-<
+    cgCommVectorI->gather(snap->cgData.position,
-<
+                          snap->cgIData.position);
-<
+    cgCommVectorJ->gather(snap->cgData.position,
-<
+                          snap->cgJData.position);
->
->
+    cgPlanVectorRow->gather(snap_->cgData.position,
->
+                            cgRowData.position);
->
->
+    cgPlanVectorColumn->gather(snap_->cgData.position,
->
+                               cgColData.position);
->
+    // if needed, gather the atomic rotation matrices
-<
+    if (snap->atomData.getStorageLayout() & DataStorage::dslAmat) {
-<
+      AtomCommMatrixI->gather(snap->atomData.aMat,
-<
+                              snap->atomIData.aMat);
-<
+      AtomCommMatrixJ->gather(snap->atomData.aMat,
-<
+                              snap->atomJData.aMat);
->
+    if (storageLayout_ & DataStorage::dslAmat) {
->
+      AtomPlanMatrixRow->gather(snap_->atomData.aMat,
->
+                                atomRowData.aMat);
->
+      AtomPlanMatrixColumn->gather(snap_->atomData.aMat,
->
+                                   atomColData.aMat);
+    // if needed, gather the atomic eletrostatic frames
-<
+    if (snap->atomData.getStorageLayout() & DataStorage::dslElectroFrame) {
-<
+      AtomCommMatrixI->gather(snap->atomData.electroFrame,
-<
+                              snap->atomIData.electroFrame);
-<
+      AtomCommMatrixJ->gather(snap->atomData.electroFrame,
-<
+                              snap->atomJData.electroFrame);
->
+    if (storageLayout_ & DataStorage::dslElectroFrame) {
->
+      AtomPlanMatrixRow->gather(snap_->atomData.electroFrame,
->
+                                atomRowData.electroFrame);
->
+      AtomPlanMatrixColumn->gather(snap_->atomData.electroFrame,
->
+                                   atomColData.electroFrame);
-+
+#endif
-<
+  void ForceDecomposition::collectIntermediateData() {
->
+  /* collects information obtained during the pre-pair loop onto local
->
+   * data structures.
->
+   */
->
+  void ForceMatrixDecomposition::collectIntermediateData() {
->
+    snap_ = sman_->getCurrentSnapshot();
->
+    storageLayout_ = sman_->getStorageLayout();
+#ifdef IS_MPI
-–
+    Snapshot* snap = sman_->getCurrentSnapshot();
-<
+    if (snap->atomData.getStorageLayout() & DataStorage::dslDensity) {
-<
+      AtomCommRealI->scatter(snap->atomIData.density,
-<
+                             snap->atomData.density);
-<
+      std::vector<RealType> rho_tmp;
-<
+      int n = snap->getNumberOfAtoms();
-<
+      rho_tmp.reserve( n );
-<
+      AtomCommRealJ->scatter(snap->atomJData.density, rho_tmp);
->
+    if (storageLayout_ & DataStorage::dslDensity) {
->
->
+      AtomPlanRealRow->scatter(atomRowData.density,
->
+                               snap_->atomData.density);
->
->
+      int n = snap_->atomData.density.size();
->
+      vector<RealType> rho_tmp(n, 0.0);
->
+      AtomPlanRealColumn->scatter(atomColData.density, rho_tmp);
+      for (int i = 0; i < n; i++)
-<
+        snap->atomData.density[i] += rho_tmp[i];
->
+        snap_->atomData.density[i] += rho_tmp[i];
+#endif
-<
-<
+  void ForceDecomposition::distributeIntermediateData() {
->
->
+  /*
->
+   * redistributes information obtained during the pre-pair loop out to
->
+   * row and column-indexed data structures
->
+   */
->
+  void ForceMatrixDecomposition::distributeIntermediateData() {
->
+    snap_ = sman_->getCurrentSnapshot();
->
+    storageLayout_ = sman_->getStorageLayout();
+#ifdef IS_MPI
-<
+    Snapshot* snap = sman_->getCurrentSnapshot();
-<
+    if (snap->atomData.getStorageLayout() & DataStorage::dslFunctional) {
-<
+      AtomCommRealI->gather(snap->atomData.functional,
-<
+                            snap->atomIData.functional);
-<
+      AtomCommRealJ->gather(snap->atomData.functional,
-<
+                            snap->atomJData.functional);
->
+    if (storageLayout_ & DataStorage::dslFunctional) {
->
+      AtomPlanRealRow->gather(snap_->atomData.functional,
->
+                              atomRowData.functional);
->
+      AtomPlanRealColumn->gather(snap_->atomData.functional,
->
+                                 atomColData.functional);
-<
+    if (snap->atomData.getStorageLayout() & DataStorage::dslFunctionalDerivative) {
-<
+      AtomCommRealI->gather(snap->atomData.functionalDerivative,
-<
+                            snap->atomIData.functionalDerivative);
-<
+      AtomCommRealJ->gather(snap->atomData.functionalDerivative,
-<
+                            snap->atomJData.functionalDerivative);
->
+    if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
->
+      AtomPlanRealRow->gather(snap_->atomData.functionalDerivative,
->
+                              atomRowData.functionalDerivative);
->
+      AtomPlanRealColumn->gather(snap_->atomData.functionalDerivative,
->
+                                 atomColData.functionalDerivative);
+#endif
-<
+  void ForceDecomposition::collectData() {
-<
+#ifdef IS_MPI
-<
+    Snapshot* snap = sman_->getCurrentSnapshot();
-<
+    int n = snap->getNumberOfAtoms();
-<
-<
+    std::vector<Vector3d> frc_tmp;
-<
+    frc_tmp.reserve( n );
->
+  void ForceMatrixDecomposition::collectData() {
->
+    snap_ = sman_->getCurrentSnapshot();
->
+    storageLayout_ = sman_->getStorageLayout();
->
+#ifdef IS_MPI
->
+    int n = snap_->atomData.force.size();
->
+    vector<Vector3d> frc_tmp(n, V3Zero);
-<
+    AtomCommVectorI->scatter(snap->atomIData.force, frc_tmp);
-<
+    for (int i = 0; i < n; i++)
-<
+      snap->atomData.force[i] += frc_tmp[i];
->
+    AtomPlanVectorRow->scatter(atomRowData.force, frc_tmp);
->
+    for (int i = 0; i < n; i++) {
->
+      snap_->atomData.force[i] += frc_tmp[i];
->
+      frc_tmp[i] = 0.0;
->
+    }
-<
+    AtomCommVectorJ->scatter(snap->atomJData.force, frc_tmp);
-<
+    for (int i = 0; i < n; i++)
-<
+      snap->atomData.force[i] += frc_tmp[i];
->
+    AtomPlanVectorColumn->scatter(atomColData.force, frc_tmp);
->
+    for (int i = 0; i < n; i++) {
->
+      snap_->atomData.force[i] += frc_tmp[i];
->
+    }
->
->
+    if (storageLayout_ & DataStorage::dslTorque) {
->
->
+      int nt = snap_->atomData.torque.size();
->
+      vector<Vector3d> trq_tmp(nt, V3Zero);
->
->
+      AtomPlanVectorRow->scatter(atomRowData.torque, trq_tmp);
->
+      for (int i = 0; i < nt; i++) {
->
+        snap_->atomData.torque[i] += trq_tmp[i];
->
+        trq_tmp[i] = 0.0;
->
+      }
->
->
+      AtomPlanVectorColumn->scatter(atomColData.torque, trq_tmp);
->
+      for (int i = 0; i < nt; i++)
->
+        snap_->atomData.torque[i] += trq_tmp[i];
->
+    }
->
->
+    if (storageLayout_ & DataStorage::dslSkippedCharge) {
->
->
+      int ns = snap_->atomData.skippedCharge.size();
->
+      vector<RealType> skch_tmp(ns, 0.0);
->
->
+      AtomPlanRealRow->scatter(atomRowData.skippedCharge, skch_tmp);
->
+      for (int i = 0; i < ns; i++) {
->
+        snap_->atomData.skippedCharge[i] += skch_tmp[i];
->
+        skch_tmp[i] = 0.0;
->
+      }
->
->
+      AtomPlanRealColumn->scatter(atomColData.skippedCharge, skch_tmp);
->
+      for (int i = 0; i < ns; i++)
->
+        snap_->atomData.skippedCharge[i] += skch_tmp[i];
->
+    }
-+
+    nLocal_ = snap_->getNumberOfAtoms();
-+
-+
+    vector<potVec> pot_temp(nLocal_,
-+
+                            Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
-+
-+
+    // scatter/gather pot_row into the members of my column
-+
-+
+    AtomPlanPotRow->scatter(pot_row, pot_temp);
-+
-+
+    for (int ii = 0;  ii < pot_temp.size(); ii++ )
-+
+      pairwisePot += pot_temp[ii];
-<
+    if (snap->atomData.getStorageLayout() & DataStorage::dslTorque) {
-<
+      std::vector<Vector3d> trq_tmp;
-<
+      trq_tmp.reserve( n );
->
+    fill(pot_temp.begin(), pot_temp.end(),
->
+         Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
-<
+      AtomCommVectorI->scatter(snap->atomIData.torque, trq_tmp);
-<
+      for (int i = 0; i < n; i++)
-<
+        snap->atomData.torque[i] += trq_tmp[i];
-<
-<
+      AtomCommVectorJ->scatter(snap->atomJData.torque, trq_tmp);
-<
+      for (int i = 0; i < n; i++)
-<
+        snap->atomData.torque[i] += trq_tmp[i];
->
+    AtomPlanPotColumn->scatter(pot_col, pot_temp);
->
->
+    for (int ii = 0;  ii < pot_temp.size(); ii++ )
->
+      pairwisePot += pot_temp[ii];
->
->
+    for (int ii = 0; ii < N_INTERACTION_FAMILIES; ii++) {
->
+      RealType ploc1 = pairwisePot[ii];
->
+      RealType ploc2 = 0.0;
->
+      MPI::COMM_WORLD.Allreduce(&ploc1, &ploc2, 1, MPI::REALTYPE, MPI::SUM);
->
+      pairwisePot[ii] = ploc2;
-+
-+
+#endif
-+
-+
+  }
-+
-+
+  int ForceMatrixDecomposition::getNAtomsInRow() {
-+
+#ifdef IS_MPI
-+
+    return nAtomsInRow_;
-+
+#else
-+
+    return nLocal_;
-+
+#endif
-+
+  }
-+
-+
+  /**
-+
+   * returns the list of atoms belonging to this group.
-+
+   */
-+
+  vector<int> ForceMatrixDecomposition::getAtomsInGroupRow(int cg1){
-+
+#ifdef IS_MPI
-+
+    return groupListRow_[cg1];
-+
+#else
-+
+    return groupList_[cg1];
-+
+#endif
-+
+  }
-+
-+
+  vector<int> ForceMatrixDecomposition::getAtomsInGroupColumn(int cg2){
-+
+#ifdef IS_MPI
-+
+    return groupListCol_[cg2];
-+
+#else
-+
+    return groupList_[cg2];
-+
+#endif
-+
+  }
-+
-+
+  Vector3d ForceMatrixDecomposition::getIntergroupVector(int cg1, int cg2){
-+
+    Vector3d d;
-<
+    // Still need pot!
->
+#ifdef IS_MPI
->
+    d = cgColData.position[cg2] - cgRowData.position[cg1];
->
+#else
->
+    d = snap_->cgData.position[cg2] - snap_->cgData.position[cg1];
->
+#endif
-+
+    snap_->wrapVector(d);
-+
+    return d;
-+
+  }
-+
+  Vector3d ForceMatrixDecomposition::getAtomToGroupVectorRow(int atom1, int cg1){
-+
-+
+    Vector3d d;
-+
-+
+#ifdef IS_MPI
-+
+    d = cgRowData.position[cg1] - atomRowData.position[atom1];
-+
+#else
-+
+    d = snap_->cgData.position[cg1] - snap_->atomData.position[atom1];
+#endif
-+
-+
+    snap_->wrapVector(d);
-+
+    return d;
-+
+  Vector3d ForceMatrixDecomposition::getAtomToGroupVectorColumn(int atom2, int cg2){
-+
+    Vector3d d;
-+
-+
+#ifdef IS_MPI
-+
+    d = cgColData.position[cg2] - atomColData.position[atom2];
-+
+#else
-+
+    d = snap_->cgData.position[cg2] - snap_->atomData.position[atom2];
-+
+#endif
-+
-+
+    snap_->wrapVector(d);
-+
+    return d;
-+
+  }
-+
-+
+  RealType ForceMatrixDecomposition::getMassFactorRow(int atom1) {
-+
+#ifdef IS_MPI
-+
+    return massFactorsRow[atom1];
-+
+#else
-+
+    return massFactors[atom1];
-+
+#endif
-+
+  }
-+
-+
+  RealType ForceMatrixDecomposition::getMassFactorColumn(int atom2) {
-+
+#ifdef IS_MPI
-+
+    return massFactorsCol[atom2];
-+
+#else
-+
+    return massFactors[atom2];
-+
+#endif
-+
-+
+  }
-+
-+
+  Vector3d ForceMatrixDecomposition::getInteratomicVector(int atom1, int atom2){
-+
+    Vector3d d;
-+
-+
+#ifdef IS_MPI
-+
+    d = atomColData.position[atom2] - atomRowData.position[atom1];
-+
+#else
-+
+    d = snap_->atomData.position[atom2] - snap_->atomData.position[atom1];
-+
+#endif
-+
-+
+    snap_->wrapVector(d);
-+
+    return d;
-+
+  }
-+
-+
+  vector<int> ForceMatrixDecomposition::getExcludesForAtom(int atom1) {
-+
+    return excludesForAtom[atom1];
-+
+  }
-+
-+
+  /**
-+
+   * We need to exclude some overcounted interactions that result from
-+
+   * the parallel decomposition.
-+
+   */
-+
+  bool ForceMatrixDecomposition::skipAtomPair(int atom1, int atom2) {
-+
+    int unique_id_1, unique_id_2;
-+
-+
+#ifdef IS_MPI
-+
+    // in MPI, we have to look up the unique IDs for each atom
-+
+    unique_id_1 = AtomRowToGlobal[atom1];
-+
+    unique_id_2 = AtomColToGlobal[atom2];
-+
-+
+    // this situation should only arise in MPI simulations
-+
+    if (unique_id_1 == unique_id_2) return true;
-+
-+
+    // this prevents us from doing the pair on multiple processors
-+
+    if (unique_id_1 < unique_id_2) {
-+
+      if ((unique_id_1 + unique_id_2) % 2 == 0) return true;
-+
+    } else {
-+
+      if ((unique_id_1 + unique_id_2) % 2 == 1) return true;
-+
+    }
-+
+#endif
-+
+    return false;
-+
+  }
-+
-+
+  /**
-+
+   * We need to handle the interactions for atoms who are involved in
-+
+   * the same rigid body as well as some short range interactions
-+
+   * (bonds, bends, torsions) differently from other interactions.
-+
+   * We'll still visit the pairwise routines, but with a flag that
-+
+   * tells those routines to exclude the pair from direct long range
-+
+   * interactions.  Some indirect interactions (notably reaction
-+
+   * field) must still be handled for these pairs.
-+
+   */
-+
+  bool ForceMatrixDecomposition::excludeAtomPair(int atom1, int atom2) {
-+
+    int unique_id_2;
-+
+#ifdef IS_MPI
-+
+    // in MPI, we have to look up the unique IDs for the row atom.
-+
+    unique_id_2 = AtomColToGlobal[atom2];
-+
+#else
-+
+    // in the normal loop, the atom numbers are unique
-+
+    unique_id_2 = atom2;
-+
+#endif
-+
-+
+    for (vector<int>::iterator i = excludesForAtom[atom1].begin();
-+
+         i != excludesForAtom[atom1].end(); ++i) {
-+
+      if ( (*i) == unique_id_2 ) return true;
-+
+    }
-+
-+
+    return false;
-+
+  }
-+
-+
-+
+  void ForceMatrixDecomposition::addForceToAtomRow(int atom1, Vector3d fg){
-+
+#ifdef IS_MPI
-+
+    atomRowData.force[atom1] += fg;
-+
+#else
-+
+    snap_->atomData.force[atom1] += fg;
-+
+#endif
-+
+  }
-+
-+
+  void ForceMatrixDecomposition::addForceToAtomColumn(int atom2, Vector3d fg){
-+
+#ifdef IS_MPI
-+
+    atomColData.force[atom2] += fg;
-+
+#else
-+
+    snap_->atomData.force[atom2] += fg;
-+
+#endif
-+
+  }
-+
-+
+    // filling interaction blocks with pointers
-+
+  void ForceMatrixDecomposition::fillInteractionData(InteractionData &idat,
-+
+                                                     int atom1, int atom2) {
-+
-+
+    idat.excluded = excludeAtomPair(atom1, atom2);
-+
-+
+#ifdef IS_MPI
-+
+    idat.atypes = make_pair( atypesRow[atom1], atypesCol[atom2]);
-+
+    //idat.atypes = make_pair( ff_->getAtomType(identsRow[atom1]),
-+
+    //                         ff_->getAtomType(identsCol[atom2]) );
-+
-+
+    if (storageLayout_ & DataStorage::dslAmat) {
-+
+      idat.A1 = &(atomRowData.aMat[atom1]);
-+
+      idat.A2 = &(atomColData.aMat[atom2]);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslElectroFrame) {
-+
+      idat.eFrame1 = &(atomRowData.electroFrame[atom1]);
-+
+      idat.eFrame2 = &(atomColData.electroFrame[atom2]);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslTorque) {
-+
+      idat.t1 = &(atomRowData.torque[atom1]);
-+
+      idat.t2 = &(atomColData.torque[atom2]);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslDensity) {
-+
+      idat.rho1 = &(atomRowData.density[atom1]);
-+
+      idat.rho2 = &(atomColData.density[atom2]);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslFunctional) {
-+
+      idat.frho1 = &(atomRowData.functional[atom1]);
-+
+      idat.frho2 = &(atomColData.functional[atom2]);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
-+
+      idat.dfrho1 = &(atomRowData.functionalDerivative[atom1]);
-+
+      idat.dfrho2 = &(atomColData.functionalDerivative[atom2]);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslParticlePot) {
-+
+      idat.particlePot1 = &(atomRowData.particlePot[atom1]);
-+
+      idat.particlePot2 = &(atomColData.particlePot[atom2]);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslSkippedCharge) {
-+
+      idat.skippedCharge1 = &(atomRowData.skippedCharge[atom1]);
-+
+      idat.skippedCharge2 = &(atomColData.skippedCharge[atom2]);
-+
+    }
-+
-+
+#else
-+
-+
+    idat.atypes = make_pair( atypesLocal[atom1], atypesLocal[atom2]);
-+
+    //idat.atypes = make_pair( ff_->getAtomType(idents[atom1]),
-+
+    //                         ff_->getAtomType(idents[atom2]) );
-+
-+
+    if (storageLayout_ & DataStorage::dslAmat) {
-+
+      idat.A1 = &(snap_->atomData.aMat[atom1]);
-+
+      idat.A2 = &(snap_->atomData.aMat[atom2]);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslElectroFrame) {
-+
+      idat.eFrame1 = &(snap_->atomData.electroFrame[atom1]);
-+
+      idat.eFrame2 = &(snap_->atomData.electroFrame[atom2]);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslTorque) {
-+
+      idat.t1 = &(snap_->atomData.torque[atom1]);
-+
+      idat.t2 = &(snap_->atomData.torque[atom2]);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslDensity) {
-+
+      idat.rho1 = &(snap_->atomData.density[atom1]);
-+
+      idat.rho2 = &(snap_->atomData.density[atom2]);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslFunctional) {
-+
+      idat.frho1 = &(snap_->atomData.functional[atom1]);
-+
+      idat.frho2 = &(snap_->atomData.functional[atom2]);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslFunctionalDerivative) {
-+
+      idat.dfrho1 = &(snap_->atomData.functionalDerivative[atom1]);
-+
+      idat.dfrho2 = &(snap_->atomData.functionalDerivative[atom2]);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslParticlePot) {
-+
+      idat.particlePot1 = &(snap_->atomData.particlePot[atom1]);
-+
+      idat.particlePot2 = &(snap_->atomData.particlePot[atom2]);
-+
+    }
-+
-+
+    if (storageLayout_ & DataStorage::dslSkippedCharge) {
-+
+      idat.skippedCharge1 = &(snap_->atomData.skippedCharge[atom1]);
-+
+      idat.skippedCharge2 = &(snap_->atomData.skippedCharge[atom2]);
-+
+    }
-+
+#endif
-+
+  }
-+
-+
-+
+  void ForceMatrixDecomposition::unpackInteractionData(InteractionData &idat, int atom1, int atom2) {
-+
+#ifdef IS_MPI
-+
+    pot_row[atom1] += 0.5 *  *(idat.pot);
-+
+    pot_col[atom2] += 0.5 *  *(idat.pot);
-+
-+
+    atomRowData.force[atom1] += *(idat.f1);
-+
+    atomColData.force[atom2] -= *(idat.f1);
-+
+#else
-+
+    pairwisePot += *(idat.pot);
-+
-+
+    snap_->atomData.force[atom1] += *(idat.f1);
-+
+    snap_->atomData.force[atom2] -= *(idat.f1);
-+
+#endif
-+
-+
+  }
-+
-+
+  /*
-+
+   * buildNeighborList
-+
+   *
-+
+   * first element of pair is row-indexed CutoffGroup
-+
+   * second element of pair is column-indexed CutoffGroup
-+
+   */
-+
+  vector<pair<int, int> > ForceMatrixDecomposition::buildNeighborList() {
-+
-+
+    vector<pair<int, int> > neighborList;
-+
+    groupCutoffs cuts;
-+
+    bool doAllPairs = false;
-+
-+
+#ifdef IS_MPI
-+
+    cellListRow_.clear();
-+
+    cellListCol_.clear();
-+
+#else
-+
+    cellList_.clear();
-+
+#endif
-+
-+
+    RealType rList_ = (largestRcut_ + skinThickness_);
-+
+    RealType rl2 = rList_ * rList_;
-+
+    Snapshot* snap_ = sman_->getCurrentSnapshot();
-+
+    Mat3x3d Hmat = snap_->getHmat();
-+
+    Vector3d Hx = Hmat.getColumn(0);
-+
+    Vector3d Hy = Hmat.getColumn(1);
-+
+    Vector3d Hz = Hmat.getColumn(2);
-+
-+
+    nCells_.x() = (int) ( Hx.length() )/ rList_;
-+
+    nCells_.y() = (int) ( Hy.length() )/ rList_;
-+
+    nCells_.z() = (int) ( Hz.length() )/ rList_;
-+
-+
+    // handle small boxes where the cell offsets can end up repeating cells
-+
-+
+    if (nCells_.x() < 3) doAllPairs = true;
-+
+    if (nCells_.y() < 3) doAllPairs = true;
-+
+    if (nCells_.z() < 3) doAllPairs = true;
-+
-+
+    Mat3x3d invHmat = snap_->getInvHmat();
-+
+    Vector3d rs, scaled, dr;
-+
+    Vector3i whichCell;
-+
+    int cellIndex;
-+
+    int nCtot = nCells_.x() * nCells_.y() * nCells_.z();
-+
-+
+#ifdef IS_MPI
-+
+    cellListRow_.resize(nCtot);
-+
+    cellListCol_.resize(nCtot);
-+
+#else
-+
+    cellList_.resize(nCtot);
-+
+#endif
-+
-+
+    if (!doAllPairs) {
-+
+#ifdef IS_MPI
-+
-+
+      for (int i = 0; i < nGroupsInRow_; i++) {
-+
+        rs = cgRowData.position[i];
-+
-+
+        // scaled positions relative to the box vectors
-+
+        scaled = invHmat * rs;
-+
-+
+        // wrap the vector back into the unit box by subtracting integer box
-+
+        // numbers
-+
+        for (int j = 0; j < 3; j++) {
-+
+          scaled[j] -= roundMe(scaled[j]);
-+
+          scaled[j] += 0.5;
-+
+        }
-+
-+
+        // find xyz-indices of cell that cutoffGroup is in.
-+
+        whichCell.x() = nCells_.x() * scaled.x();
-+
+        whichCell.y() = nCells_.y() * scaled.y();
-+
+        whichCell.z() = nCells_.z() * scaled.z();
-+
-+
+        // find single index of this cell:
-+
+        cellIndex = Vlinear(whichCell, nCells_);
-+
-+
+        // add this cutoff group to the list of groups in this cell;
-+
+        cellListRow_[cellIndex].push_back(i);
-+
+      }
-+
+      for (int i = 0; i < nGroupsInCol_; i++) {
-+
+        rs = cgColData.position[i];
-+
-+
+        // scaled positions relative to the box vectors
-+
+        scaled = invHmat * rs;
-+
-+
+        // wrap the vector back into the unit box by subtracting integer box
-+
+        // numbers
-+
+        for (int j = 0; j < 3; j++) {
-+
+          scaled[j] -= roundMe(scaled[j]);
-+
+          scaled[j] += 0.5;
-+
+        }
-+
-+
+        // find xyz-indices of cell that cutoffGroup is in.
-+
+        whichCell.x() = nCells_.x() * scaled.x();
-+
+        whichCell.y() = nCells_.y() * scaled.y();
-+
+        whichCell.z() = nCells_.z() * scaled.z();
-+
-+
+        // find single index of this cell:
-+
+        cellIndex = Vlinear(whichCell, nCells_);
-+
-+
+        // add this cutoff group to the list of groups in this cell;
-+
+        cellListCol_[cellIndex].push_back(i);
-+
+      }
-+
-+
+#else
-+
+      for (int i = 0; i < nGroups_; i++) {
-+
+        rs = snap_->cgData.position[i];
-+
-+
+        // scaled positions relative to the box vectors
-+
+        scaled = invHmat * rs;
-+
-+
+        // wrap the vector back into the unit box by subtracting integer box
-+
+        // numbers
-+
+        for (int j = 0; j < 3; j++) {
-+
+          scaled[j] -= roundMe(scaled[j]);
-+
+          scaled[j] += 0.5;
-+
+        }
-+
-+
+        // find xyz-indices of cell that cutoffGroup is in.
-+
+        whichCell.x() = nCells_.x() * scaled.x();
-+
+        whichCell.y() = nCells_.y() * scaled.y();
-+
+        whichCell.z() = nCells_.z() * scaled.z();
-+
-+
+        // find single index of this cell:
-+
+        cellIndex = Vlinear(whichCell, nCells_);
-+
-+
+        // add this cutoff group to the list of groups in this cell;
-+
+        cellList_[cellIndex].push_back(i);
-+
+      }
-+
-+
+#endif
-+
-+
+      for (int m1z = 0; m1z < nCells_.z(); m1z++) {
-+
+        for (int m1y = 0; m1y < nCells_.y(); m1y++) {
-+
+          for (int m1x = 0; m1x < nCells_.x(); m1x++) {
-+
+            Vector3i m1v(m1x, m1y, m1z);
-+
+            int m1 = Vlinear(m1v, nCells_);
-+
-+
+            for (vector<Vector3i>::iterator os = cellOffsets_.begin();
-+
+                 os != cellOffsets_.end(); ++os) {
-+
-+
+              Vector3i m2v = m1v + (*os);
-+
-+
-+
+              if (m2v.x() >= nCells_.x()) {
-+
+                m2v.x() = 0;
-+
+              } else if (m2v.x() < 0) {
-+
+                m2v.x() = nCells_.x() - 1;
-+
+              }
-+
-+
+              if (m2v.y() >= nCells_.y()) {
-+
+                m2v.y() = 0;
-+
+              } else if (m2v.y() < 0) {
-+
+                m2v.y() = nCells_.y() - 1;
-+
+              }
-+
-+
+              if (m2v.z() >= nCells_.z()) {
-+
+                m2v.z() = 0;
-+
+              } else if (m2v.z() < 0) {
-+
+                m2v.z() = nCells_.z() - 1;
-+
+              }
-+
-+
+              int m2 = Vlinear (m2v, nCells_);
-+
-+
+#ifdef IS_MPI
-+
+              for (vector<int>::iterator j1 = cellListRow_[m1].begin();
-+
+                   j1 != cellListRow_[m1].end(); ++j1) {
-+
+                for (vector<int>::iterator j2 = cellListCol_[m2].begin();
-+
+                     j2 != cellListCol_[m2].end(); ++j2) {
-+
-+
+                  // In parallel, we need to visit *all* pairs of row
-+
+                  // & column indicies and will divide labor in the
-+
+                  // force evaluation later.
-+
+                  dr = cgColData.position[(*j2)] - cgRowData.position[(*j1)];
-+
+                  snap_->wrapVector(dr);
-+
+                  cuts = getGroupCutoffs( (*j1), (*j2) );
-+
+                  if (dr.lengthSquare() < cuts.third) {
-+
+                    neighborList.push_back(make_pair((*j1), (*j2)));
-+
+                  }
-+
+                }
-+
+              }
-+
+#else
-+
-+
+              for (vector<int>::iterator j1 = cellList_[m1].begin();
-+
+                   j1 != cellList_[m1].end(); ++j1) {
-+
+                for (vector<int>::iterator j2 = cellList_[m2].begin();
-+
+                     j2 != cellList_[m2].end(); ++j2) {
-+
-+
+                  // Always do this if we're in different cells or if
-+
+                  // we're in the same cell and the global index of the
-+
+                  // j2 cutoff group is less than the j1 cutoff group
-+
-+
+                  if (m2 != m1 || (*j2) < (*j1)) {
-+
+                    dr = snap_->cgData.position[(*j2)] - snap_->cgData.position[(*j1)];
-+
+                    snap_->wrapVector(dr);
-+
+                    cuts = getGroupCutoffs( (*j1), (*j2) );
-+
+                    if (dr.lengthSquare() < cuts.third) {
-+
+                      neighborList.push_back(make_pair((*j1), (*j2)));
-+
+                    }
-+
+                  }
-+
+                }
-+
+              }
-+
+#endif
-+
+            }
-+
+          }
-+
+        }
-+
+      }
-+
+    } else {
-+
+      // branch to do all cutoff group pairs
-+
+#ifdef IS_MPI
-+
+      for (int j1 = 0; j1 < nGroupsInRow_; j1++) {
-+
+        for (int j2 = 0; j2 < nGroupsInCol_; j2++) {
-+
+          dr = cgColData.position[j2] - cgRowData.position[j1];
-+
+          snap_->wrapVector(dr);
-+
+          cuts = getGroupCutoffs( j1, j2 );
-+
+          if (dr.lengthSquare() < cuts.third) {
-+
+            neighborList.push_back(make_pair(j1, j2));
-+
+          }
-+
+        }
-+
+      }
-+
+#else
-+
+      for (int j1 = 0; j1 < nGroups_ - 1; j1++) {
-+
+        for (int j2 = j1 + 1; j2 < nGroups_; j2++) {
-+
+          dr = snap_->cgData.position[j2] - snap_->cgData.position[j1];
-+
+          snap_->wrapVector(dr);
-+
+          cuts = getGroupCutoffs( j1, j2 );
-+
+          if (dr.lengthSquare() < cuts.third) {
-+
+            neighborList.push_back(make_pair(j1, j2));
-+
+          }
-+
+        }
-+
+      }
-+
+#endif
-+
+    }
-+
-+
+    // save the local cutoff group positions for the check that is
-+
+    // done on each loop:
-+
+    saved_CG_positions_.clear();
-+
+    for (int i = 0; i < nGroups_; i++)
-+
+      saved_CG_positions_.push_back(snap_->cgData.position[i]);
-+
-+
+    return neighborList;
-+
+  }
+} //end namespace OpenMD

Diff Legend

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

Comparing: branches/development/src/parallel/ForceDecomposition.cpp (file contents), Revision 1540 by gezelter, Mon Jan 17 21:34:36 2011 UTC vs. branches/development/src/parallel/ForceMatrixDecomposition.cpp (file contents), Revision 1612 by gezelter, Fri Aug 12 19:59:56 2011 UTC

Diff Legend

Comparing:
branches/development/src/parallel/ForceDecomposition.cpp (file contents), Revision 1540 by gezelter, Mon Jan 17 21:34:36 2011 UTC vs.
branches/development/src/parallel/ForceMatrixDecomposition.cpp (file contents), Revision 1612 by gezelter, Fri Aug 12 19:59:56 2011 UTC