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/ForceMatrixDecomposition.cpp (file contents):
Revision 1688 by gezelter, Wed Mar 14 17:56:01 2012 UTC vs.
Revision 1761 by gezelter, Fri Jun 22 20:01:37 2012 UTC

# Line 95 | Line 95 | namespace OpenMD {
95      storageLayout_ = sman_->getStorageLayout();
96      ff_ = info_->getForceField();
97      nLocal_ = snap_->getNumberOfAtoms();
98 <    
98 >  
99      nGroups_ = info_->getNLocalCutoffGroups();
100      // gather the information for atomtype IDs (atids):
101      idents = info_->getIdentArray();
# Line 109 | Line 109 | namespace OpenMD {
109      PairList* oneTwo = info_->getOneTwoInteractions();
110      PairList* oneThree = info_->getOneThreeInteractions();
111      PairList* oneFour = info_->getOneFourInteractions();
112 <
112 >    
113 >    if (needVelocities_)
114 >      snap_->cgData.setStorageLayout(DataStorage::dslPosition |
115 >                                     DataStorage::dslVelocity);
116 >    else
117 >      snap_->cgData.setStorageLayout(DataStorage::dslPosition);
118 >    
119   #ifdef IS_MPI
120  
121      MPI::Intracomm row = rowComm.getComm();
# Line 145 | Line 151 | namespace OpenMD {
151      cgRowData.resize(nGroupsInRow_);
152      cgRowData.setStorageLayout(DataStorage::dslPosition);
153      cgColData.resize(nGroupsInCol_);
154 <    cgColData.setStorageLayout(DataStorage::dslPosition);
155 <        
154 >    if (needVelocities_)
155 >      // we only need column velocities if we need them.
156 >      cgColData.setStorageLayout(DataStorage::dslPosition |
157 >                                 DataStorage::dslVelocity);
158 >    else    
159 >      cgColData.setStorageLayout(DataStorage::dslPosition);
160 >      
161      identsRow.resize(nAtomsInRow_);
162      identsCol.resize(nAtomsInCol_);
163      
# Line 165 | Line 176 | namespace OpenMD {
176      pot_row.resize(nAtomsInRow_);
177      pot_col.resize(nAtomsInCol_);
178  
179 +    expot_row.resize(nAtomsInRow_);
180 +    expot_col.resize(nAtomsInCol_);
181 +
182      AtomRowToGlobal.resize(nAtomsInRow_);
183      AtomColToGlobal.resize(nAtomsInCol_);
184      AtomPlanIntRow->gather(AtomLocalToGlobal, AtomRowToGlobal);
# Line 450 | Line 464 | namespace OpenMD {
464      }
465    }
466  
453 –
467    groupCutoffs ForceMatrixDecomposition::getGroupCutoffs(int cg1, int cg2) {
468      int i, j;  
469   #ifdef IS_MPI
# Line 474 | Line 487 | namespace OpenMD {
487    void ForceMatrixDecomposition::zeroWorkArrays() {
488      pairwisePot = 0.0;
489      embeddingPot = 0.0;
490 +    excludedPot = 0.0;
491 +    excludedSelfPot = 0.0;
492  
493   #ifdef IS_MPI
494      if (storageLayout_ & DataStorage::dslForce) {
# Line 490 | Line 505 | namespace OpenMD {
505           Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
506  
507      fill(pot_col.begin(), pot_col.end(),
508 +         Vector<RealType, N_INTERACTION_FAMILIES> (0.0));  
509 +
510 +    fill(expot_row.begin(), expot_row.end(),
511 +         Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
512 +
513 +    fill(expot_col.begin(), expot_col.end(),
514           Vector<RealType, N_INTERACTION_FAMILIES> (0.0));  
515  
516      if (storageLayout_ & DataStorage::dslParticlePot) {    
# Line 525 | Line 546 | namespace OpenMD {
546             atomColData.skippedCharge.end(), 0.0);
547      }
548  
549 +    if (storageLayout_ & DataStorage::dslFlucQForce) {      
550 +      fill(atomRowData.flucQFrc.begin(),
551 +           atomRowData.flucQFrc.end(), 0.0);
552 +      fill(atomColData.flucQFrc.begin(),
553 +           atomColData.flucQFrc.end(), 0.0);
554 +    }
555 +
556 +    if (storageLayout_ & DataStorage::dslElectricField) {    
557 +      fill(atomRowData.electricField.begin(),
558 +           atomRowData.electricField.end(), V3Zero);
559 +      fill(atomColData.electricField.begin(),
560 +           atomColData.electricField.end(), V3Zero);
561 +    }
562 +
563 +    if (storageLayout_ & DataStorage::dslFlucQForce) {    
564 +      fill(atomRowData.flucQFrc.begin(), atomRowData.flucQFrc.end(),
565 +           0.0);
566 +      fill(atomColData.flucQFrc.begin(), atomColData.flucQFrc.end(),
567 +           0.0);
568 +    }
569 +
570   #endif
571      // even in parallel, we need to zero out the local arrays:
572  
# Line 537 | Line 579 | namespace OpenMD {
579        fill(snap_->atomData.density.begin(),
580             snap_->atomData.density.end(), 0.0);
581      }
582 +
583      if (storageLayout_ & DataStorage::dslFunctional) {
584        fill(snap_->atomData.functional.begin(),
585             snap_->atomData.functional.end(), 0.0);
586      }
587 +
588      if (storageLayout_ & DataStorage::dslFunctionalDerivative) {      
589        fill(snap_->atomData.functionalDerivative.begin(),
590             snap_->atomData.functionalDerivative.end(), 0.0);
591      }
592 +
593      if (storageLayout_ & DataStorage::dslSkippedCharge) {      
594        fill(snap_->atomData.skippedCharge.begin(),
595             snap_->atomData.skippedCharge.end(), 0.0);
596      }
597 <    
597 >
598 >    if (storageLayout_ & DataStorage::dslElectricField) {      
599 >      fill(snap_->atomData.electricField.begin(),
600 >           snap_->atomData.electricField.end(), V3Zero);
601 >    }
602    }
603  
604  
# Line 572 | Line 621 | namespace OpenMD {
621      cgPlanVectorColumn->gather(snap_->cgData.position,
622                                 cgColData.position);
623  
624 +
625 +
626 +    if (needVelocities_) {
627 +      // gather up the atomic velocities
628 +      AtomPlanVectorColumn->gather(snap_->atomData.velocity,
629 +                                   atomColData.velocity);
630 +      
631 +      cgPlanVectorColumn->gather(snap_->cgData.velocity,
632 +                                 cgColData.velocity);
633 +    }
634 +
635      
636      // if needed, gather the atomic rotation matrices
637      if (storageLayout_ & DataStorage::dslAmat) {
# Line 589 | Line 649 | namespace OpenMD {
649                                     atomColData.electroFrame);
650      }
651  
652 +    // if needed, gather the atomic fluctuating charge values
653 +    if (storageLayout_ & DataStorage::dslFlucQPosition) {
654 +      AtomPlanRealRow->gather(snap_->atomData.flucQPos,
655 +                              atomRowData.flucQPos);
656 +      AtomPlanRealColumn->gather(snap_->atomData.flucQPos,
657 +                                 atomColData.flucQPos);
658 +    }
659 +
660   #endif      
661    }
662    
# Line 611 | Line 679 | namespace OpenMD {
679        for (int i = 0; i < n; i++)
680          snap_->atomData.density[i] += rho_tmp[i];
681      }
682 +
683 +    if (storageLayout_ & DataStorage::dslElectricField) {
684 +      
685 +      AtomPlanVectorRow->scatter(atomRowData.electricField,
686 +                                 snap_->atomData.electricField);
687 +      
688 +      int n = snap_->atomData.electricField.size();
689 +      vector<Vector3d> field_tmp(n, V3Zero);
690 +      AtomPlanVectorColumn->scatter(atomColData.electricField, field_tmp);
691 +      for (int i = 0; i < n; i++)
692 +        snap_->atomData.electricField[i] += field_tmp[i];
693 +    }
694   #endif
695    }
696  
# Line 690 | Line 770 | namespace OpenMD {
770              
771      }
772      
773 +    if (storageLayout_ & DataStorage::dslFlucQForce) {
774 +
775 +      int nq = snap_->atomData.flucQFrc.size();
776 +      vector<RealType> fqfrc_tmp(nq, 0.0);
777 +
778 +      AtomPlanRealRow->scatter(atomRowData.flucQFrc, fqfrc_tmp);
779 +      for (int i = 0; i < nq; i++) {
780 +        snap_->atomData.flucQFrc[i] += fqfrc_tmp[i];
781 +        fqfrc_tmp[i] = 0.0;
782 +      }
783 +      
784 +      AtomPlanRealColumn->scatter(atomColData.flucQFrc, fqfrc_tmp);
785 +      for (int i = 0; i < nq; i++)
786 +        snap_->atomData.flucQFrc[i] += fqfrc_tmp[i];
787 +            
788 +    }
789 +
790      nLocal_ = snap_->getNumberOfAtoms();
791  
792      vector<potVec> pot_temp(nLocal_,
793                              Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
794 +    vector<potVec> expot_temp(nLocal_,
795 +                              Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
796  
797      // scatter/gather pot_row into the members of my column
798            
799      AtomPlanPotRow->scatter(pot_row, pot_temp);
800 +    AtomPlanPotRow->scatter(expot_row, expot_temp);
801  
802 <    for (int ii = 0;  ii < pot_temp.size(); ii++ )
802 >    for (int ii = 0;  ii < pot_temp.size(); ii++ )
803        pairwisePot += pot_temp[ii];
804 <    
804 >
805 >    for (int ii = 0;  ii < expot_temp.size(); ii++ )
806 >      excludedPot += expot_temp[ii];
807 >        
808 >    if (storageLayout_ & DataStorage::dslParticlePot) {
809 >      // This is the pairwise contribution to the particle pot.  The
810 >      // embedding contribution is added in each of the low level
811 >      // non-bonded routines.  In single processor, this is done in
812 >      // unpackInteractionData, not in collectData.
813 >      for (int ii = 0; ii < N_INTERACTION_FAMILIES; ii++) {
814 >        for (int i = 0; i < nLocal_; i++) {
815 >          // factor of two is because the total potential terms are divided
816 >          // by 2 in parallel due to row/ column scatter      
817 >          snap_->atomData.particlePot[i] += 2.0 * pot_temp[i](ii);
818 >        }
819 >      }
820 >    }
821 >
822      fill(pot_temp.begin(), pot_temp.end(),
823           Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
824 +    fill(expot_temp.begin(), expot_temp.end(),
825 +         Vector<RealType, N_INTERACTION_FAMILIES> (0.0));
826        
827      AtomPlanPotColumn->scatter(pot_col, pot_temp);    
828 +    AtomPlanPotColumn->scatter(expot_col, expot_temp);    
829      
830      for (int ii = 0;  ii < pot_temp.size(); ii++ )
831        pairwisePot += pot_temp[ii];    
832 +
833 +    for (int ii = 0;  ii < expot_temp.size(); ii++ )
834 +      excludedPot += expot_temp[ii];    
835 +
836 +    if (storageLayout_ & DataStorage::dslParticlePot) {
837 +      // This is the pairwise contribution to the particle pot.  The
838 +      // embedding contribution is added in each of the low level
839 +      // non-bonded routines.  In single processor, this is done in
840 +      // unpackInteractionData, not in collectData.
841 +      for (int ii = 0; ii < N_INTERACTION_FAMILIES; ii++) {
842 +        for (int i = 0; i < nLocal_; i++) {
843 +          // factor of two is because the total potential terms are divided
844 +          // by 2 in parallel due to row/ column scatter      
845 +          snap_->atomData.particlePot[i] += 2.0 * pot_temp[i](ii);
846 +        }
847 +      }
848 +    }
849      
850 +    if (storageLayout_ & DataStorage::dslParticlePot) {
851 +      int npp = snap_->atomData.particlePot.size();
852 +      vector<RealType> ppot_temp(npp, 0.0);
853 +
854 +      // This is the direct or embedding contribution to the particle
855 +      // pot.
856 +      
857 +      AtomPlanRealRow->scatter(atomRowData.particlePot, ppot_temp);
858 +      for (int i = 0; i < npp; i++) {
859 +        snap_->atomData.particlePot[i] += ppot_temp[i];
860 +      }
861 +
862 +      fill(ppot_temp.begin(), ppot_temp.end(), 0.0);
863 +      
864 +      AtomPlanRealColumn->scatter(atomColData.particlePot, ppot_temp);
865 +      for (int i = 0; i < npp; i++) {
866 +        snap_->atomData.particlePot[i] += ppot_temp[i];
867 +      }
868 +    }
869 +
870      for (int ii = 0; ii < N_INTERACTION_FAMILIES; ii++) {
871        RealType ploc1 = pairwisePot[ii];
872        RealType ploc2 = 0.0;
# Line 718 | Line 875 | namespace OpenMD {
875      }
876  
877      for (int ii = 0; ii < N_INTERACTION_FAMILIES; ii++) {
878 <      RealType ploc1 = embeddingPot[ii];
878 >      RealType ploc1 = excludedPot[ii];
879        RealType ploc2 = 0.0;
880        MPI::COMM_WORLD.Allreduce(&ploc1, &ploc2, 1, MPI::REALTYPE, MPI::SUM);
881 <      embeddingPot[ii] = ploc2;
881 >      excludedPot[ii] = ploc2;
882      }
883  
884 +    // Here be dragons.
885 +    MPI::Intracomm col = colComm.getComm();
886 +
887 +    col.Allreduce(MPI::IN_PLACE,
888 +                  &snap_->frameData.conductiveHeatFlux[0], 3,
889 +                  MPI::REALTYPE, MPI::SUM);
890 +
891 +
892   #endif
893  
894    }
895  
896 +  /**
897 +   * Collects information obtained during the post-pair (and embedding
898 +   * functional) loops onto local data structures.
899 +   */
900 +  void ForceMatrixDecomposition::collectSelfData() {
901 +    snap_ = sman_->getCurrentSnapshot();
902 +    storageLayout_ = sman_->getStorageLayout();
903 +
904 + #ifdef IS_MPI
905 +    for (int ii = 0; ii < N_INTERACTION_FAMILIES; ii++) {
906 +      RealType ploc1 = embeddingPot[ii];
907 +      RealType ploc2 = 0.0;
908 +      MPI::COMM_WORLD.Allreduce(&ploc1, &ploc2, 1, MPI::REALTYPE, MPI::SUM);
909 +      embeddingPot[ii] = ploc2;
910 +    }    
911 +    for (int ii = 0; ii < N_INTERACTION_FAMILIES; ii++) {
912 +      RealType ploc1 = excludedSelfPot[ii];
913 +      RealType ploc2 = 0.0;
914 +      MPI::COMM_WORLD.Allreduce(&ploc1, &ploc2, 1, MPI::REALTYPE, MPI::SUM);
915 +      excludedSelfPot[ii] = ploc2;
916 +    }    
917 + #endif
918 +    
919 +  }
920 +
921 +
922 +
923    int ForceMatrixDecomposition::getNAtomsInRow() {  
924   #ifdef IS_MPI
925      return nAtomsInRow_;
# Line 766 | Line 958 | namespace OpenMD {
958      
959      snap_->wrapVector(d);
960      return d;    
961 +  }
962 +
963 +  Vector3d ForceMatrixDecomposition::getGroupVelocityColumn(int cg2){
964 + #ifdef IS_MPI
965 +    return cgColData.velocity[cg2];
966 + #else
967 +    return snap_->cgData.velocity[cg2];
968 + #endif
969 +  }
970 +
971 +  Vector3d ForceMatrixDecomposition::getAtomVelocityColumn(int atom2){
972 + #ifdef IS_MPI
973 +    return atomColData.velocity[atom2];
974 + #else
975 +    return snap_->atomData.velocity[atom2];
976 + #endif
977    }
978  
979  
# Line 834 | Line 1042 | namespace OpenMD {
1042     * We need to exclude some overcounted interactions that result from
1043     * the parallel decomposition.
1044     */
1045 <  bool ForceMatrixDecomposition::skipAtomPair(int atom1, int atom2) {
1046 <    int unique_id_1, unique_id_2;
1045 >  bool ForceMatrixDecomposition::skipAtomPair(int atom1, int atom2, int cg1, int cg2) {
1046 >    int unique_id_1, unique_id_2, group1, group2;
1047          
1048   #ifdef IS_MPI
1049      // in MPI, we have to look up the unique IDs for each atom
1050      unique_id_1 = AtomRowToGlobal[atom1];
1051      unique_id_2 = AtomColToGlobal[atom2];
1052 +    group1 = cgRowToGlobal[cg1];
1053 +    group2 = cgColToGlobal[cg2];
1054   #else
1055      unique_id_1 = AtomLocalToGlobal[atom1];
1056      unique_id_2 = AtomLocalToGlobal[atom2];
1057 +    group1 = cgLocalToGlobal[cg1];
1058 +    group2 = cgLocalToGlobal[cg2];
1059   #endif  
1060  
1061      if (unique_id_1 == unique_id_2) return true;
# Line 855 | Line 1067 | namespace OpenMD {
1067      } else {
1068        if ((unique_id_1 + unique_id_2) % 2 == 1) return true;
1069      }
1070 + #endif    
1071 +
1072 + #ifndef IS_MPI
1073 +    if (group1 == group2) {
1074 +      if (unique_id_1 < unique_id_2) return true;
1075 +    }
1076   #endif
1077      
1078      return false;
# Line 950 | Line 1168 | namespace OpenMD {
1168        idat.skippedCharge2 = &(atomColData.skippedCharge[atom2]);
1169      }
1170  
1171 +    if (storageLayout_ & DataStorage::dslFlucQPosition) {              
1172 +      idat.flucQ1 = &(atomRowData.flucQPos[atom1]);
1173 +      idat.flucQ2 = &(atomColData.flucQPos[atom2]);
1174 +    }
1175 +
1176   #else
1177      
955 –
956 –    // cerr << "atoms = " << atom1 << " " << atom2 << "\n";
957 –    // cerr << "pos1 = " << snap_->atomData.position[atom1] << "\n";
958 –    // cerr << "pos2 = " << snap_->atomData.position[atom2] << "\n";
959 –
1178      idat.atypes = make_pair( atypesLocal[atom1], atypesLocal[atom2]);
961 –    //idat.atypes = make_pair( ff_->getAtomType(idents[atom1]),
962 –    //                         ff_->getAtomType(idents[atom2]) );
1179  
1180      if (storageLayout_ & DataStorage::dslAmat) {
1181        idat.A1 = &(snap_->atomData.aMat[atom1]);
# Line 999 | Line 1215 | namespace OpenMD {
1215      if (storageLayout_ & DataStorage::dslSkippedCharge) {
1216        idat.skippedCharge1 = &(snap_->atomData.skippedCharge[atom1]);
1217        idat.skippedCharge2 = &(snap_->atomData.skippedCharge[atom2]);
1218 +    }
1219 +
1220 +    if (storageLayout_ & DataStorage::dslFlucQPosition) {              
1221 +      idat.flucQ1 = &(snap_->atomData.flucQPos[atom1]);
1222 +      idat.flucQ2 = &(snap_->atomData.flucQPos[atom2]);
1223      }
1224 +
1225   #endif
1226    }
1227  
# Line 1008 | Line 1230 | namespace OpenMD {
1230   #ifdef IS_MPI
1231      pot_row[atom1] += RealType(0.5) *  *(idat.pot);
1232      pot_col[atom2] += RealType(0.5) *  *(idat.pot);
1233 +    expot_row[atom1] += RealType(0.5) *  *(idat.excludedPot);
1234 +    expot_col[atom2] += RealType(0.5) *  *(idat.excludedPot);
1235  
1236      atomRowData.force[atom1] += *(idat.f1);
1237      atomColData.force[atom2] -= *(idat.f1);
1238 +
1239 +    if (storageLayout_ & DataStorage::dslFlucQForce) {              
1240 +      atomRowData.flucQFrc[atom1] -= *(idat.dVdFQ1);
1241 +      atomColData.flucQFrc[atom2] -= *(idat.dVdFQ2);
1242 +    }
1243 +
1244 +    if (storageLayout_ & DataStorage::dslElectricField) {              
1245 +      atomRowData.electricField[atom1] += *(idat.eField1);
1246 +      atomColData.electricField[atom2] += *(idat.eField2);
1247 +    }
1248 +
1249   #else
1250      pairwisePot += *(idat.pot);
1251 +    excludedPot += *(idat.excludedPot);
1252  
1253      snap_->atomData.force[atom1] += *(idat.f1);
1254      snap_->atomData.force[atom2] -= *(idat.f1);
1255 +
1256 +    if (idat.doParticlePot) {
1257 +      // This is the pairwise contribution to the particle pot.  The
1258 +      // embedding contribution is added in each of the low level
1259 +      // non-bonded routines.  In parallel, this calculation is done
1260 +      // in collectData, not in unpackInteractionData.
1261 +      snap_->atomData.particlePot[atom1] += *(idat.vpair) * *(idat.sw);
1262 +      snap_->atomData.particlePot[atom2] += *(idat.vpair) * *(idat.sw);
1263 +    }
1264 +    
1265 +    if (storageLayout_ & DataStorage::dslFlucQForce) {              
1266 +      snap_->atomData.flucQFrc[atom1] -= *(idat.dVdFQ1);
1267 +      snap_->atomData.flucQFrc[atom2] -= *(idat.dVdFQ2);
1268 +    }
1269 +
1270 +    if (storageLayout_ & DataStorage::dslElectricField) {              
1271 +      snap_->atomData.electricField[atom1] += *(idat.eField1);
1272 +      snap_->atomData.electricField[atom2] += *(idat.eField2);
1273 +    }
1274 +
1275   #endif
1276      
1277    }

Diff Legend

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