Page MenuHomec4science
Files F101998706

communicator.hh
No OneTemporary
Actions

File Metadata

Created
Sun, Feb 16, 00:45

communicator.hh
View Options

/**
* @file communicator.hh
*
* @author Guillaume Anciaux <guillaume.anciaux@epfl.ch>
*
* @date Fri Nov 08 12:44:53 2013
*
* @brief Mother class of LM communicators
*
* @section LICENSE
*
* Copyright INRIA and CEA
*
* The LibMultiScale is a C++ parallel framework for the multiscale
* coupling methods dedicated to material simulations. This framework
* provides an API which makes it possible to program coupled simulations
* and integration of already existing codes.
*
* This Project was initiated in a collaboration between INRIA Futurs Bordeaux
* within ScAlApplix team and CEA/DPTA Ile de France.
* The project is now continued at the Ecole Polytechnique Fédérale de Lausanne
* within the LSMS/ENAC laboratory.
*
* This software is governed by the CeCILL-C license under French law and
* abiding by the rules of distribution of free software. You can use,
* modify and/ or redistribute the software under the terms of the CeCILL-C
* license as circulated by CEA, CNRS and INRIA at the following URL
* "http://www.cecill.info".
*
* As a counterpart to the access to the source code and rights to copy,
* modify and redistribute granted by the license, users are provided only
* with a limited warranty and the software's author, the holder of the
* economic rights, and the successive licensors have only limited
* liability.
*
* In this respect, the user's attention is drawn to the risks associated
* with loading, using, modifying and/or developing or reproducing the
* software by the user in light of its specific status of free software,
* that may mean that it is complicated to manipulate, and that also
* therefore means that it is reserved for developers and experienced
* professionals having in-depth computer knowledge. Users are therefore
* encouraged to load and test the software's suitability as regards their
* requirements in conditions enabling the security of their systems and/or
* data to be ensured and, more generally, to use and operate it in the
* same conditions as regards security.
*
* The fact that you are presently reading this means that you have had
* knowledge of the CeCILL-C license and that you accept its terms.
*
*/
#ifndef __LIBMULTISCALE_COMMUNICATOR_HH__
#define __LIBMULTISCALE_COMMUNICATOR_HH__
/* -------------------------------------------------------------------------- */
#include "comm_buffer.hh"
#include <mpi.h>
#include "tensor.hh"
/* -------------------------------------------------------------------------- */
__BEGIN_LIBMULTISCALE__
/* -------------------------------------------------------------------------- */
class Geometry;
/* -------------------------------------------------------------------------- */
class Communicator{
public:
/* -------------------------------------------------------------------------- */
// Constructor/desctructor
/* -------------------------------------------------------------------------- */
virtual ~Communicator(){};
/* ------------------------------------------------------------------------ */
/* Class Members */
/* ------------------------------------------------------------------------ */
//! return the communicator used
static Communicator & getCommunicator(){
if (!communicator) LM_THROW("communicator not already created");
return *communicator;
};
//! set the communicator
template <typename T>
static void createCommunicator(){
if (communicator) LM_THROW("communicator was already created");
communicator = new T();
};
/* -------------------------------------------------------------------------- */
//Virtual functions
/* -------------------------------------------------------------------------- */
//! test if am i in the group specified by ID
virtual bool amIinGroup(CommGroup group_index)=0;
//! add a group of nb_proc processors
virtual UInt addGroup(UInt nb_procs) =0;
//! return the equivalent MPI group from one group ID
virtual MPI_Comm getMpiGroup(CommGroup group_index)=0;
//! barrier for a given group (no param is a barrier on all processors)
virtual void synchronize(CommGroup group_index = group_all)=0;
//! return number of processors present in group specified
virtual UInt getNBprocsOnGroup(CommGroup i)=0;
//! return my real rank from a group rank and group ID
virtual UInt realRank(UInt i, CommGroup group)=0;
//! return my group rank from a real rank and group ID
virtual UInt groupRank(UInt i, CommGroup group)=0;
virtual bool isInGroup(UInt i , CommGroup group)=0;
virtual UInt getNBGroups()=0;
//! wait for unsynchronisated comunications
virtual void waitForPendingComs()=0;
//! global comunication over a group to send geometries structures
virtual void sendLocalGeometriesToGroup(Geometry & geom,CommGroup group)=0;
//! global comunication over a group to receive geometries structures
virtual void receiveLocalGeometriesFromGroup(Geometry ** geom,CommGroup group)=0;
//! global comunication Over a group to send boolean tables
virtual void sendCommunicationTable(std::vector<UInt> & com_with,
CommGroup destgroup)=0;
//! global comunication over a group to receive boolean tables
virtual void receiveCommunicationTable(std::vector<UInt> & com_with,
CommGroup fromgroup)=0;
//! receive array from a processor on by its ID in a given group
virtual void receiveUInts(CommBuffer<UInt> & d,UInt nb,UInt from,CommGroup group,
const std::string & comment)=0;
//! send array to a processor on by its ID in a given group
virtual void sendUInts(CommBuffer<UInt> & i,UInt nb,UInt dest,
CommGroup group,const std::string & comment)=0;
//! receive array from a processor on by its ID in a given group
virtual void receiveReals(CommBuffer<Real> & d,UInt nb,UInt from,CommGroup group,
const std::string & comment)=0;
//! send array to a processor on by its ID in a given group
virtual void sendReals(CommBuffer<Real> & i,UInt nb,UInt dest,
CommGroup group,const std::string & comment)=0;
//! reduction operation over a given group
virtual void reduceUInt(CommBuffer<UInt> & contrib,UInt nb,CommGroup group,
const std::string & comment, Operator op)=0;
//! reduction operation over a given group
virtual void reduceReal(CommBuffer<Real> & contrib,UInt nb,CommGroup group,
const std::string & comment, Operator op)=0;
//! reduction operation over a given group
virtual void allReduceUInt(CommBuffer<UInt> & contrib,UInt nb,CommGroup group,
const std::string & comment, Operator op)=0;
//! reduction operation over a given group
virtual void allReduceReal(CommBuffer<Real> & contrib,UInt nb,CommGroup group,
const std::string & comment, Operator op)=0;
//! printself function
virtual void printself(std::ostream & stream){};
/* -------------------------------------------------------------------------- */
// Generic functions
/* -------------------------------------------------------------------------- */
//receive functions
//! receive an nb-data array from a processor specified by its ID in a given group
template <typename T>
inline void receive(std::vector<T> & d,UInt nb,UInt from,CommGroup group,const std::string & comment);
//! receive array from a processor specified by its ID in a given group
template <typename T>
inline void receive(std::vector<T> & d,UInt from,CommGroup group,const std::string & comment);
//! receive an nb-data array from a processor specified by its ID in a given group
template <typename T>
inline void receive(CommBuffer<T> & d,UInt nb, UInt from,CommGroup group,const std::string & comment);
//! receive array from a processor specified by its ID in a given group
template <typename T>
inline void receive(CommBuffer<T> & d,UInt from,CommGroup group,const std::string & comment);
//! receive a C-style array from a processor specified by its ID in a given group
template <typename T>
inline void receive(T * d,UInt nb,UInt from,CommGroup group,const std::string & comment);
//sending functions
//! send an nb-data array to a processor specified by its ID in a given group
template <typename T>
inline void send(std::vector<T> & d,UInt nb,UInt to,CommGroup group,const std::string & comment);
//! send array to a processor specified by its ID in a given group
template <typename T>
inline void send(std::vector<T> & d,UInt to,CommGroup group,const std::string & comment);
//! send an nb-data array to a processor specified by its ID in a given group
template <typename T>
inline void send(CommBuffer<T> & d,UInt nb,UInt to,CommGroup group,const std::string & comment);
//! send array to a processor specified by its ID in a given group
template <typename T>
inline void send(CommBuffer<T> & d,UInt to,CommGroup group,const std::string & comment);
//! send a C-style array to a processor specified by its ID in a given group
template <typename T>
inline void send(T * d,UInt nb, UInt to,CommGroup group,const std::string & comment);
//reducing functions
//! reduce by sum from a commbuffer
template <typename T>
inline void reduce(CommBuffer<T> & d,UInt nb,CommGroup group,const std::string & comment, Operator op);
//! reduce from a vector
template <typename T>
inline void reduce(std::vector<T> & d,UInt nb,CommGroup group,const std::string & comment, Operator op);
//! reduce a C-style array (for compatibility : involve additional copy)
template <typename T>
inline void reduce(T *d,UInt nb,CommGroup group,const std::string & comment, Operator op);
//! reduce by sum from a commbuffer
template <typename T>
inline void allReduce(CommBuffer<T> & d,UInt nb,CommGroup group,const std::string & comment, Operator op);
//! reduce from a vector
template <typename T>
inline void allReduce(std::vector<T> & d,UInt nb,CommGroup group,const std::string & comment, Operator op);
//! reduce a C-style array (for compatibility : involve additional copy)
template <typename T>
inline void allReduce(T *d,UInt nb,CommGroup group,const std::string & comment, Operator op);
/* ------------------------------------------------------------------------ */
/* Class Members */
/* ------------------------------------------------------------------------ */
static Communicator * communicator;
};
/* -------------------------------------------------------------------------- */
template <typename T>
inline void Communicator::receive(std::vector<T> & d,UInt from,CommGroup group,
const std::string & comment){
CommBuffer<T> buf(d);
receive(buf,from,group,comment);
}
/* -------------------------------------------------------------------------- */
template <typename T>
inline void Communicator::receive(std::vector<T> & d,UInt nb,UInt from,CommGroup group,
const std::string & comment){
CommBuffer<T> buf(d);
receive(buf,nb,from,group,comment);
}
/* -------------------------------------------------------------------------- */
template <typename T>
inline void Communicator::send(std::vector<T> & d,UInt nb,UInt from,CommGroup group,
const std::string & comment){
CommBuffer<T> buf(d);
send(buf,nb,from,group,comment);
}
/* -------------------------------------------------------------------------- */
template <typename T>
inline void Communicator::send(std::vector<T> & d,UInt from,CommGroup group,
const std::string & comment){
send(d,d.size(),from,group,comment);
}
/* -------------------------------------------------------------------------- */
template <typename T>
inline void Communicator::receive(CommBuffer<T> & d,UInt from,CommGroup group,
const std::string & comment){
receive(d,UINT_MAX,from,group,comment);
}
/* -------------------------------------------------------------------------- */
template <>
inline void Communicator::receive(CommBuffer<Real> & d,UInt nb,UInt from,CommGroup group,
const std::string & comment){
receiveReals(d,nb,from,group,comment);
}
/* -------------------------------------------------------------------------- */
template <>
inline void Communicator::send(CommBuffer<Real> & d,UInt nb,UInt to,CommGroup group,
const std::string & comment){
sendReals(d,nb,to,group,comment);
}
/* -------------------------------------------------------------------------- */
template <>
inline void Communicator::receive(CommBuffer<UInt> & d,UInt nb,UInt from,CommGroup group,
const std::string & comment){
receiveUInts(d,nb,from,group,comment);
}
/* -------------------------------------------------------------------------- */
template <>
inline void Communicator::send(CommBuffer<UInt> & d,UInt nb,UInt to,CommGroup group,
const std::string & comment){
sendUInts(d,nb,to,group,comment);
}
/* -------------------------------------------------------------------------- */
template <typename T>
inline void Communicator::send(CommBuffer<T> & d,UInt from,CommGroup group,
const std::string & comment){
send(d,d.size(),from,group,comment);
}
/* -------------------------------------------------------------------------- */
template <>
inline void Communicator::reduce(CommBuffer<Real> & d,UInt nb,CommGroup group,
const std::string & comment, Operator op){
reduceReal(d,nb,group,comment,op);
}
/* -------------------------------------------------------------------------- */
template <>
inline void Communicator::reduce(CommBuffer<UInt> & d,UInt nb,CommGroup group,
const std::string & comment, Operator op){
reduceUInt(d,nb,group,comment,op);
}
/* -------------------------------------------------------------------------- */
template <typename T>
inline void Communicator::reduce(std::vector<T> & d,UInt nb,CommGroup group,
const std::string & comment, Operator op){
CommBuffer<T> buf(d);
reduce(buf,nb,group,comment,op);
}
/* -------------------------------------------------------------------------- */
template <typename T>
inline void Communicator::reduce(T *d,UInt nb,CommGroup group,
const std::string & comment, Operator op){
CommBuffer<T> buf;
buf.resize(nb);
std::copy(d,d+nb,buf.buffer());
reduce(buf,nb,group,comment,op);
std::copy(buf.buffer(),buf.buffer()+nb,d);
}
/* -------------------------------------------------------------------------- */
template <>
inline void Communicator::allReduce(CommBuffer<Real> & d,UInt nb,CommGroup group,
const std::string & comment, Operator op){
allReduceReal(d,nb,group,comment,op);
}
/* -------------------------------------------------------------------------- */
template <>
inline void Communicator::allReduce(CommBuffer<UInt> & d,UInt nb,CommGroup group,
const std::string & comment, Operator op){
allReduceUInt(d,nb,group,comment,op);
}
/* -------------------------------------------------------------------------- */
template <typename T>
inline void Communicator::allReduce(std::vector<T> & d,UInt nb,CommGroup group,
const std::string & comment, Operator op){
CommBuffer<T> buf(d);
allReduce(buf,nb,group,comment,op);
}
/* -------------------------------------------------------------------------- */
template <typename T>
inline void Communicator::allReduce(T *d,UInt nb,CommGroup group,
const std::string & comment, Operator op){
CommBuffer<T> buf;
buf.resize(nb);
std::copy(d,d+nb,buf.buffer());
allReduce(buf,nb,group,comment,op);
std::copy(buf.buffer(),buf.buffer()+nb,d);
}
/* -------------------------------------------------------------------------- */
template <>
inline void Communicator::allReduce(Tensor * d,UInt nb, CommGroup group,
const std::string & comment, Operator op){
allReduce(d->getVec(),9,group,comment,op);
}
/* -------------------------------------------------------------------------- */
template <typename T>
inline void Communicator::send(T * d,UInt nb, UInt to,CommGroup group,
const std::string & comment){
CommBuffer<T> buf;
buf.resize(nb);
std::copy(d,d+nb,buf.buffer());
send(buf,nb,to,group,comment);
}
/* -------------------------------------------------------------------------- */
template <typename T>
inline void Communicator::receive(T * d,UInt nb,UInt from,CommGroup group,
const std::string & comment){
CommBuffer<T> buf;
buf.resize(nb);
receive(buf,nb,from,group,comment);
std::copy(buf.buffer(),buf.buffer()+nb,d);
}
/* -------------------------------------------------------------------------- */
inline std::ostream& operator << (std::ostream& os,Communicator &c){
c.printself(os);
return os;
}
/* -------------------------------------------------------------------------- */
__END_LIBMULTISCALE__
//#include "communicator_world.hh"
//__BEGIN_LIBMULTISCALE__
//static WorldCommunicator WORLD_COMMUNICATOR;
///* -------------------------------------------------------------------------- */
//__END_LIBMULTISCALE__
#endif /* __LIBMULTISCALE_COMMUNICATOR_HH__ */

Event Timeline

c4science · Help