src/brains/SimCreator.cpp

/*
 * Copyright (c) 2005 The University of Notre Dame. All Rights Reserved.
 *
 * The University of Notre Dame grants you ("Licensee") a
 * non-exclusive, royalty free, license to use, modify and
 * redistribute this software in source and binary code form, provided
 * that the following conditions are met:
 *
 * 1. Acknowledgement of the program authors must be made in any
 *    publication of scientific results based in part on use of the
 *    program.  An acceptable form of acknowledgement is citation of
 *    the article in which the program was described (Matthew
 *    A. Meineke, Charles F. Vardeman II, Teng Lin, Christopher
 *    J. Fennell and J. Daniel Gezelter, "OOPSE: An Object-Oriented
 *    Parallel Simulation Engine for Molecular Dynamics,"
 *    J. Comput. Chem. 26, pp. 252-271 (2005))
 *
 * 2. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 *
 * 3. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the
 *    distribution.
 *
 * This software is provided "AS IS," without a warranty of any
 * kind. All express or implied conditions, representations and
 * warranties, including any implied warranty of merchantability,
 * fitness for a particular purpose or non-infringement, are hereby
 * excluded.  The University of Notre Dame and its licensors shall not
 * be liable for any damages suffered by licensee as a result of
 * using, modifying or distributing the software or its
 * derivatives. In no event will the University of Notre Dame or its
 * licensors be liable for any lost revenue, profit or data, or for
 * direct, indirect, special, consequential, incidental or punitive
 * damages, however caused and regardless of the theory of liability,
 * arising out of the use of or inability to use software, even if the
 * University of Notre Dame has been advised of the possibility of
 * such damages.
 */
 
/**
 * @file SimCreator.cpp
 * @author tlin
 * @date 11/03/2004
 * @time 13:51am
 * @version 1.0
 */
#include <exception>
#include <iostream>
#include <sstream>
#include <string>

#include "brains/MoleculeCreator.hpp"
#include "brains/SimCreator.hpp"
#include "brains/SimSnapshotManager.hpp"
#include "io/DumpReader.hpp"
#include "UseTheForce/ForceFieldFactory.hpp"
#include "utils/simError.h"
#include "utils/StringUtils.hpp"
#include "math/SeqRandNumGen.hpp"
#include "mdParser/MDLexer.hpp"
#include "mdParser/MDParser.hpp"
#include "mdParser/MDTreeParser.hpp"
#include "mdParser/SimplePreprocessor.hpp"
#include "antlr/ANTLRException.hpp"
#include "antlr/TokenStreamRecognitionException.hpp"
#include "antlr/TokenStreamIOException.hpp"
#include "antlr/TokenStreamException.hpp"
#include "antlr/RecognitionException.hpp"
#include "antlr/CharStreamException.hpp"

#include "antlr/MismatchedCharException.hpp"
#include "antlr/MismatchedTokenException.hpp"
#include "antlr/NoViableAltForCharException.hpp"
#include "antlr/NoViableAltException.hpp"

#ifdef IS_MPI
#include "math/ParallelRandNumGen.hpp"
#endif

namespace oopse {
  
Globals* SimCreator::parseFile(const std::string mdFileName){
        Globals* simParams = NULL;
        try {

            // Create a preprocessor that preprocesses md file into an ostringstream
            std::stringstream ppStream;
#ifdef IS_MPI            
            int streamSize;
            const int masterNode = 0;
            int commStatus;
            if (worldRank == masterNode) {
#endif 
                
                SimplePreprocessor preprocessor;
                preprocessor.preprocess(mdFileName, ppStream);
                
#ifdef IS_MPI            
                //brocasting the stream size
                streamSize = ppStream.str().size() +1;
                commStatus = MPI_Bcast(&streamSize, 1, MPI_LONG, masterNode, MPI_COMM_WORLD);                   

                commStatus = MPI_Bcast(static_cast<void*>(const_cast<char*>(ppStream.str().c_str())), streamSize, MPI_CHAR, masterNode, MPI_COMM_WORLD); 
            
                
            } else {
                //get stream size
                commStatus = MPI_Bcast(&streamSize, 1, MPI_LONG, masterNode, MPI_COMM_WORLD);   
                
                  char* buf = new char[streamSize];
                  assert(buf);
                
                  //receive file content
                  commStatus = MPI_Bcast(buf, streamSize, MPI_CHAR, masterNode, MPI_COMM_WORLD); 
                
                  ppStream.str(buf);
                  delete buf;

            }
#endif            
            // Create a scanner that reads from the input stream
            MDLexer lexer(ppStream);
            lexer.setFilename(mdFileName);
            lexer.initDeferredLineCount();
    
            // Create a parser that reads from the scanner
            MDParser parser(lexer);
            parser.setFilename(mdFileName);

            // Create an observer that synchorizes file name change
            FilenameObserver observer;
            observer.setLexer(&lexer);
            observer.setParser(&parser);
            lexer.setObserver(&observer);
    
            antlr::ASTFactory factory;
            parser.initializeASTFactory(factory);
            parser.setASTFactory(&factory);
            parser.mdfile();

            // Create a tree parser that reads information into Globals
            MDTreeParser treeParser;
            treeParser.initializeASTFactory(factory);
            treeParser.setASTFactory(&factory);
             simParams = treeParser.walkTree(parser.getAST());

        }

      
      catch(antlr::MismatchedCharException& e) {
          sprintf(painCave.errMsg, 
                  "parser exception: %s %s:%d:%d\n",
                  e.getMessage().c_str(),e.getFilename().c_str(), e.getLine(), e.getColumn());
          painCave.isFatal = 1;
          simError();           
      }
      catch(antlr::MismatchedTokenException &e) {
          sprintf(painCave.errMsg, 
                  "parser exception: %s %s:%d:%d\n",
                  e.getMessage().c_str(),e.getFilename().c_str(), e.getLine(), e.getColumn());
          painCave.isFatal = 1;
          simError();   
      }
      catch(antlr::NoViableAltForCharException &e) {
          sprintf(painCave.errMsg, 
                  "parser exception: %s %s:%d:%d\n",
                  e.getMessage().c_str(),e.getFilename().c_str(), e.getLine(), e.getColumn());
          painCave.isFatal = 1;
          simError();   
      }
      catch(antlr::NoViableAltException &e) {
          sprintf(painCave.errMsg, 
                  "parser exception: %s %s:%d:%d\n",
                  e.getMessage().c_str(),e.getFilename().c_str(), e.getLine(), e.getColumn());
          painCave.isFatal = 1;
          simError();   
      }
      
        catch(antlr::TokenStreamRecognitionException& e) {
          sprintf(painCave.errMsg, 
                  "parser exception: %s %s:%d:%d\n",
                  e.getMessage().c_str(),e.getFilename().c_str(), e.getLine(), e.getColumn());
          painCave.isFatal = 1;
          simError();   
        }
        
        catch(antlr::TokenStreamIOException& e) {
          sprintf(painCave.errMsg, 
                  "parser exception: %s\n",
                  e.getMessage().c_str());
          painCave.isFatal = 1;
          simError();
        }
        
        catch(antlr::TokenStreamException& e) {
          sprintf(painCave.errMsg, 
                  "parser exception: %s\n",
                  e.getMessage().c_str());
          painCave.isFatal = 1;
          simError();
        }        
       catch (antlr::RecognitionException& e) {
          sprintf(painCave.errMsg, 
                  "parser exception: %s %s:%d:%d\n",
                  e.getMessage().c_str(),e.getFilename().c_str(), e.getLine(), e.getColumn());
          painCave.isFatal = 1;
          simError();          
       }
       catch (antlr::CharStreamException& e) {
          sprintf(painCave.errMsg, 
                  "parser exception: %s\n",
                  e.getMessage().c_str());
          painCave.isFatal = 1;
          simError();        
       }
       catch (OOPSEException& e) {
          sprintf(painCave.errMsg, 
                  "%s\n",
                  e.getMessage().c_str());
          painCave.isFatal = 1;
          simError();
       }
       catch (std::exception& e) {
          sprintf(painCave.errMsg, 
                  "parser exception: %s\n",
                  e.what());
          painCave.isFatal = 1;
          simError();
       }

        return simParams;
  }
  
  SimInfo*  SimCreator::createSim(const std::string & mdFileName, 
                                  bool loadInitCoords) {

    //parse meta-data file
    Globals* simParams = parseFile(mdFileName);
    
    //create the force field
    ForceField * ff = ForceFieldFactory::getInstance()
      ->createForceField(simParams->getForceField());
    
    if (ff == NULL) {
      sprintf(painCave.errMsg, 
              "ForceField Factory can not create %s force field\n",
              simParams->getForceField().c_str());
      painCave.isFatal = 1;
      simError();
    }
    
    if (simParams->haveForceFieldFileName()) {
      ff->setForceFieldFileName(simParams->getForceFieldFileName());
    }
    
    std::string forcefieldFileName;
    forcefieldFileName = ff->getForceFieldFileName();
    
    if (simParams->haveForceFieldVariant()) {
      //If the force field has variant, the variant force field name will be
      //Base.variant.frc. For exampel EAM.u6.frc
      
      std::string variant = simParams->getForceFieldVariant();
      
      std::string::size_type pos = forcefieldFileName.rfind(".frc");
      variant = "." + variant;
      if (pos != std::string::npos) {
        forcefieldFileName.insert(pos, variant);
      } else {
        //If the default force field file name does not containt .frc suffix, just append the .variant
        forcefieldFileName.append(variant);
      }
    } 
    
    ff->parse(forcefieldFileName);
    ff->setFortranForceOptions();
    //create SimInfo
    SimInfo * info = new SimInfo(ff, simParams);
     
    //gather parameters (SimCreator only retrieves part of the parameters)
    gatherParameters(info, mdFileName);
    
    //divide the molecules and determine the global index of molecules
#ifdef IS_MPI
    divideMolecules(info);
#endif 
    
    //create the molecules
    createMolecules(info);
    
    
    //allocate memory for DataStorage(circular reference, need to break it)
    info->setSnapshotManager(new SimSnapshotManager(info));
    
    //set the global index of atoms, rigidbodies and cutoffgroups (only need to be set once, the
    //global index will never change again). Local indices of atoms and rigidbodies are already set by 
    //MoleculeCreator class which actually delegates the responsibility to LocalIndexManager. 
    setGlobalIndex(info);
    
    //Alought addExculdePairs is called inside SimInfo's addMolecule method, at that point
    //atoms don't have the global index yet  (their global index are all initialized to -1).
    //Therefore we have to call addExcludePairs explicitly here. A way to work around is that
    //we can determine the beginning global indices of atoms before they get created.
    SimInfo::MoleculeIterator mi;
    Molecule* mol;
    for (mol= info->beginMolecule(mi); mol != NULL; mol = info->nextMolecule(mi)) {
      info->addExcludePairs(mol);
    }
    
    if (loadInitCoords)
      loadCoordinates(info);    
    
    return info;
  }
  
  void SimCreator::gatherParameters(SimInfo *info, const std::string& mdfile) {
    
    //figure out the output file names
    std::string prefix;
    
#ifdef IS_MPI
    
    if (worldRank == 0) {
#endif // is_mpi
      Globals * simParams = info->getSimParams();
      if (simParams->haveFinalConfig()) {
        prefix = getPrefix(simParams->getFinalConfig());
      } else {
        prefix = getPrefix(mdfile);
      }
      
      info->setFinalConfigFileName(prefix + ".eor");
      info->setDumpFileName(prefix + ".dump");
      info->setStatFileName(prefix + ".stat");
      info->setRestFileName(prefix + ".zang");
      
#ifdef IS_MPI
      
    }
    
#endif
    
  }
  
#ifdef IS_MPI
  void SimCreator::divideMolecules(SimInfo *info) {
    double numerator;
    double denominator;
    double precast;
    double x;
    double y;
    double a;
    int old_atoms;
    int add_atoms;
    int new_atoms;
    int nTarget;
    int done;
    int i;
    int j;
    int loops;
    int which_proc;
    int nProcessors;
    std::vector<int> atomsPerProc;
    int nGlobalMols = info->getNGlobalMolecules();
    std::vector<int> molToProcMap(nGlobalMols, -1); // default to an error condition:
    
    MPI_Comm_size(MPI_COMM_WORLD, &nProcessors);
    
    if (nProcessors > nGlobalMols) {
      sprintf(painCave.errMsg,
              "nProcessors (%d) > nMol (%d)\n"
              "\tThe number of processors is larger than\n"
              "\tthe number of molecules.  This will not result in a \n"
              "\tusable division of atoms for force decomposition.\n"
              "\tEither try a smaller number of processors, or run the\n"
              "\tsingle-processor version of OOPSE.\n", nProcessors, nGlobalMols);
      
      painCave.isFatal = 1;
      simError();
    }
    
    int seedValue;
    Globals * simParams = info->getSimParams();
    SeqRandNumGen* myRandom; //divide labor does not need Parallel random number generator
    if (simParams->haveSeed()) {
      seedValue = simParams->getSeed();
      myRandom = new SeqRandNumGen(seedValue);
    }else {
      myRandom = new SeqRandNumGen();
    }   
    
    
    a = 3.0 * nGlobalMols / info->getNGlobalAtoms();
    
    //initialize atomsPerProc
    atomsPerProc.insert(atomsPerProc.end(), nProcessors, 0);
    
    if (worldRank == 0) {
      numerator = info->getNGlobalAtoms();
      denominator = nProcessors;
      precast = numerator / denominator;
      nTarget = (int)(precast + 0.5);
      
      for(i = 0; i < nGlobalMols; i++) {
        done = 0;
        loops = 0;
        
        while (!done) {
          loops++;
          
          // Pick a processor at random
          
          which_proc = (int) (myRandom->rand() * nProcessors);
          
          //get the molecule stamp first
          int stampId = info->getMoleculeStampId(i);
          MoleculeStamp * moleculeStamp = info->getMoleculeStamp(stampId);
          
          // How many atoms does this processor have so far?
          old_atoms = atomsPerProc[which_proc];
          add_atoms = moleculeStamp->getNAtoms();
          new_atoms = old_atoms + add_atoms;
          
          // If we've been through this loop too many times, we need
          // to just give up and assign the molecule to this processor
          // and be done with it. 
          
          if (loops > 100) {
            sprintf(painCave.errMsg,
                    "I've tried 100 times to assign molecule %d to a "
                    " processor, but can't find a good spot.\n"
                    "I'm assigning it at random to processor %d.\n",
                    i, which_proc);
            
            painCave.isFatal = 0;
            simError();
            
            molToProcMap[i] = which_proc;
            atomsPerProc[which_proc] += add_atoms;
            
            done = 1;
            continue;
          }
          
          // If we can add this molecule to this processor without sending
          // it above nTarget, then go ahead and do it:
          
          if (new_atoms <= nTarget) {
            molToProcMap[i] = which_proc;
            atomsPerProc[which_proc] += add_atoms;
            
            done = 1;
            continue;
          }
          
          // The only situation left is when new_atoms > nTarget.  We
          // want to accept this with some probability that dies off the
          // farther we are from nTarget
          
          // roughly:  x = new_atoms - nTarget
          //           Pacc(x) = exp(- a * x)
          // where a = penalty / (average atoms per molecule)
          
          x = (double)(new_atoms - nTarget);
          y = myRandom->rand();
          
          if (y < exp(- a * x)) {
            molToProcMap[i] = which_proc;
            atomsPerProc[which_proc] += add_atoms;
            
            done = 1;
            continue;
          } else {
            continue;
          }
        }
      }
      
      delete myRandom;
      
      // Spray out this nonsense to all other processors:
      
      MPI_Bcast(&molToProcMap[0], nGlobalMols, MPI_INT, 0, MPI_COMM_WORLD);
    } else {
      
      // Listen to your marching orders from processor 0:
      
      MPI_Bcast(&molToProcMap[0], nGlobalMols, MPI_INT, 0, MPI_COMM_WORLD);
    }
    
    info->setMolToProcMap(molToProcMap);
    sprintf(checkPointMsg,
            "Successfully divided the molecules among the processors.\n");
    MPIcheckPoint();
  }
  
#endif
  
  void SimCreator::createMolecules(SimInfo *info) {
    MoleculeCreator molCreator;
    int stampId;
    
    for(int i = 0; i < info->getNGlobalMolecules(); i++) {
      
#ifdef IS_MPI
      
      if (info->getMolToProc(i) == worldRank) {
#endif
        
        stampId = info->getMoleculeStampId(i);
        Molecule * mol = molCreator.createMolecule(info->getForceField(), info->getMoleculeStamp(stampId),
                                                   stampId, i, info->getLocalIndexManager());
        
        info->addMolecule(mol);
        
#ifdef IS_MPI
        
      }
      
#endif
      
    } //end for(int i=0)   
  }
    
  void SimCreator::setGlobalIndex(SimInfo *info) {
    SimInfo::MoleculeIterator mi;
    Molecule::AtomIterator ai;
    Molecule::RigidBodyIterator ri;
    Molecule::CutoffGroupIterator ci;
    Molecule * mol;
    Atom * atom;
    RigidBody * rb;
    CutoffGroup * cg;
    int beginAtomIndex;
    int beginRigidBodyIndex;
    int beginCutoffGroupIndex;
    int nGlobalAtoms = info->getNGlobalAtoms();
    
#ifndef IS_MPI
    
    beginAtomIndex = 0;
    beginRigidBodyIndex = 0;
    beginCutoffGroupIndex = 0;
    
#else
    
    int nproc;
    int myNode;
    
    myNode = worldRank;
    MPI_Comm_size(MPI_COMM_WORLD, &nproc);
    
    std::vector < int > tmpAtomsInProc(nproc, 0);
    std::vector < int > tmpRigidBodiesInProc(nproc, 0);
    std::vector < int > tmpCutoffGroupsInProc(nproc, 0);
    std::vector < int > NumAtomsInProc(nproc, 0);
    std::vector < int > NumRigidBodiesInProc(nproc, 0);
    std::vector < int > NumCutoffGroupsInProc(nproc, 0);
    
    tmpAtomsInProc[myNode] = info->getNAtoms();
    tmpRigidBodiesInProc[myNode] = info->getNRigidBodies();
    tmpCutoffGroupsInProc[myNode] = info->getNCutoffGroups();
    
    //do MPI_ALLREDUCE to exchange the total number of atoms, rigidbodies and cutoff groups
    MPI_Allreduce(&tmpAtomsInProc[0], &NumAtomsInProc[0], nproc, MPI_INT,
                  MPI_SUM, MPI_COMM_WORLD);
    MPI_Allreduce(&tmpRigidBodiesInProc[0], &NumRigidBodiesInProc[0], nproc,
                  MPI_INT, MPI_SUM, MPI_COMM_WORLD);
    MPI_Allreduce(&tmpCutoffGroupsInProc[0], &NumCutoffGroupsInProc[0], nproc,
                  MPI_INT, MPI_SUM, MPI_COMM_WORLD);
    
    beginAtomIndex = 0;
    beginRigidBodyIndex = 0;
    beginCutoffGroupIndex = 0;
    
    for(int i = 0; i < myNode; i++) {
      beginAtomIndex += NumAtomsInProc[i];
      beginRigidBodyIndex += NumRigidBodiesInProc[i];
      beginCutoffGroupIndex += NumCutoffGroupsInProc[i];
    }
    
#endif
    
    //rigidbody's index begins right after atom's
    beginRigidBodyIndex += info->getNGlobalAtoms();
    
    for(mol = info->beginMolecule(mi); mol != NULL;
        mol = info->nextMolecule(mi)) {
      
      //local index(index in DataStorge) of atom is important
      for(atom = mol->beginAtom(ai); atom != NULL; atom = mol->nextAtom(ai)) {
        atom->setGlobalIndex(beginAtomIndex++);
      }
      
      for(rb = mol->beginRigidBody(ri); rb != NULL;
          rb = mol->nextRigidBody(ri)) {
        rb->setGlobalIndex(beginRigidBodyIndex++);
      }
      
      //local index of cutoff group is trivial, it only depends on the order of travesing
      for(cg = mol->beginCutoffGroup(ci); cg != NULL;
          cg = mol->nextCutoffGroup(ci)) {
        cg->setGlobalIndex(beginCutoffGroupIndex++);
      }
    }
    
    //fill globalGroupMembership
    std::vector<int> globalGroupMembership(info->getNGlobalAtoms(), 0);
    for(mol = info->beginMolecule(mi); mol != NULL; mol = info->nextMolecule(mi)) {        
      for (cg = mol->beginCutoffGroup(ci); cg != NULL; cg = mol->nextCutoffGroup(ci)) {
        
        for(atom = cg->beginAtom(ai); atom != NULL; atom = cg->nextAtom(ai)) {
          globalGroupMembership[atom->getGlobalIndex()] = cg->getGlobalIndex();
        }
        
      }       
    }
    
#ifdef IS_MPI    
    // Since the globalGroupMembership has been zero filled and we've only
    // poked values into the atoms we know, we can do an Allreduce
    // to get the full globalGroupMembership array (We think).
    // This would be prettier if we could use MPI_IN_PLACE like the MPI-2
    // docs said we could.
    std::vector<int> tmpGroupMembership(nGlobalAtoms, 0);
    MPI_Allreduce(&globalGroupMembership[0], &tmpGroupMembership[0], nGlobalAtoms,
                  MPI_INT, MPI_SUM, MPI_COMM_WORLD);
    info->setGlobalGroupMembership(tmpGroupMembership);
#else
    info->setGlobalGroupMembership(globalGroupMembership);
#endif
    
    //fill molMembership
    std::vector<int> globalMolMembership(info->getNGlobalAtoms(), 0);
    
    for(mol = info->beginMolecule(mi); mol != NULL; mol = info->nextMolecule(mi)) {
      
      for(atom = mol->beginAtom(ai); atom != NULL; atom = mol->nextAtom(ai)) {
        globalMolMembership[atom->getGlobalIndex()] = mol->getGlobalIndex();
      }
    }
    
#ifdef IS_MPI
    std::vector<int> tmpMolMembership(nGlobalAtoms, 0);
    
    MPI_Allreduce(&globalMolMembership[0], &tmpMolMembership[0], nGlobalAtoms,
                  MPI_INT, MPI_SUM, MPI_COMM_WORLD);
    
    info->setGlobalMolMembership(tmpMolMembership);
#else
    info->setGlobalMolMembership(globalMolMembership);
#endif
    
  }
  
  void SimCreator::loadCoordinates(SimInfo* info) {
    Globals* simParams;
    simParams = info->getSimParams();
    
    if (!simParams->haveInitialConfig()) {
      sprintf(painCave.errMsg,
              "Cannot intialize a simulation without an initial configuration file.\n");
      painCave.isFatal = 1;;
      simError();
    }
    
    DumpReader reader(info, simParams->getInitialConfig());
    int nframes = reader.getNFrames();
    
    if (nframes > 0) {
      reader.readFrame(nframes - 1);
    } else {
      //invalid initial coordinate file
      sprintf(painCave.errMsg, 
              "Initial configuration file %s should at least contain one frame\n",
              simParams->getInitialConfig().c_str());
      painCave.isFatal = 1;
      simError();
    }
    
    //copy the current snapshot to previous snapshot
    info->getSnapshotManager()->advance();
  }
  
} //end namespace oopse


Revision:	845
Committed:	Wed Jan 11 19:01:20 2006 UTC (19 years, 3 months ago) by tim
File size:	23292 byte(s)
Log Message:	instead of printing to std::cout, throwing an exception when error is found.