jobDistributer.h

/*  
    Copyright (c) 2017 Institute Jožef Stefan, Jamova cesta 39, SI-1000, Ljubljana, Slovenija

    Permission is hereby granted, free of charge, to any person obtaining a copy
    of this software and associated documentation files (the "Software"), to deal
    in the Software without restriction, including without limitation the rights
    to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
    copies of the Software, and to permit persons to whom the Software is
    furnished to do so, subject to the following conditions:

    The above copyright notice and this permission notice shall be included in all
    copies or substantial portions of the Software.

    THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
    IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
    FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
    AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
    LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
    OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
    SOFTWARE.

    Please cite the following works (bibtex source below):
        - DepolliAvbeljTrobec2008 for the simulator and simulation-based optimization
        - DepolliTrobecFilipic2013 for the AMS-DEMO optimizer
        - TrobecDepolliAvbelj2009 for the simulator

    @article{DepolliAvbeljTrobec2008,
        author = {Depolli, Matjaž and Avbelj, Viktor and Trobec, Roman},
        title = {Computer-Simulated Alternative Modes of {U}-Wave Genesis},
        journal = {Journal of Cardiovascular Electrophysiology},
        volume = {19},
        number = {1},
        publisher = {Blackwell Publishing Inc},
        issn = {1540-8167},
        url = {http://dx.doi.org/10.1111/j.1540-8167.2007.00978.x},
        doi = {10.1111/j.1540-8167.2007.00978.x},
        pages = {84--89},
        keywords = {U wave, ECG, action potential, repolarization, myocardium, computer simulation},
        year = {2008}
    }

    @article{DepolliTrobecFilipic2013,
        author = {Depolli, Matjaž and Trobec, Roman and Filipič, Bogdan},
        title = {Asynchronous master-slave parallelization of differential evolution for multiobjective optimization},
        journal = {Evolutionary Computation},
        volume = {21},
        number = {2},
        pages = {261-291},
        doi = {10.1162/EVCO_a_00076},
        issn = {1063-6560},
        url = {http://www.mitpressjournals.org/doi/abs/10.1162/EVCO_a_00076},
        year = {2013}
    }

    @inproceedings{TrobecDepolliAvbelj2009,
        title           = {Simulation of {ECG} repolarization phase with improved model of cell action potentials},
        author          = {Trobec, Roman and Depolli, Matja{\v{z}} and Avbelj, Viktor},
        booktitle       = {International Joint Conference on Biomedical Engineering Systems and Technologies},
        pages           = {325--332},
        year            = {2009},
        organization    = {Springer}
    }
*/

#ifndef JOB_DISTRIBUTER_H_INCLUDED
#define JOB_DISTRIBUTER_H_INCLUDED


#include "mpiWrapper.h"
#include <iostream>
#include <algorithm>
#include <stdexcept>
#include "VectorArithmetics.h"
#include "timer.h"

/// This is how Job class should look like
/**
	struct Job {
		int operator() (int) {return 0;}
	};
}
**/

template<class T>
inline size_t size(T) {
	return 1;
}

template<class T>
inline size_t size(const std::vector<T>& v) {
	return v.size();
}


template<class T>
inline void resize(T, size_t) {}

template<class T>
inline void resize(std::vector<T>& v, size_t newSize) {
	v.resize(newSize);
}

template<class T, size_t N>
inline void resize(Array<T, N>& v, size_t newSize) {
	assert(N == newSize);
}


template<class T>
inline T* begin(T& var) {
	return &var;
}

template<class T>
inline typename std::vector<T>::iterator begin(std::vector<T>& vec) {
	return vec.begin();
}

template<class T, size_t N>
inline T* begin(Array<T, N>& vec) {
	return vec.begin();
}


template<class T>
inline T* end(T& var) {
	return begin(var)+1;
}

template<class T>
inline typename std::vector<T>::iterator end(std::vector<T>& vec) {
	return vec.end();
}

template<class T, size_t N>
inline T* end(Array<T, N>& vec) {
	return vec.end();
}


template<class Chromosome, class Criteria>
class JobDistributer {
	const Mpi::Communicator 	communicator;
	
public:
	JobDistributer(const Mpi::Communicator& com) : communicator(com) {}

	Mpi::Communicator comm() {return communicator;}
	
	template<class Job>
	void executeWorker(Job& job, PrecisionTimer* pt = 0) {
		ScopeTimer t(*pt);
		
		int n = 0;
		size_t inBufferSize;
		size_t outBufferSize;
		Mpi::receive(n, 0, 1);
		
		Chromosome in;
		for (int i = 0; i < n; ++i) {
			BinaryStream inStream;
			Mpi::receive(inStream, 0, 1);
			inStream >> in;
//			std::cout << "worker " << communicator.getRank() << " received " << (i+1) << " of " << n << std::endl;
			Criteria out;
			double violation;
			{
				ExcludeScopeFromTimer et(*pt);
				violation = job(in, out);
			}
//			std::cout << "worker " << communicator.getRank() << " executed job" << std::endl;
			
			BinaryStream outStream;
			outStream << in << out << violation;
			Mpi::send(outStream, 0, 1);
//			std::cout << "worker " << communicator.getRank() << " sent     " << (i+1) << " of " << n << " (" << outBufferSize << ")" << std::endl;
		}
	}
	
	template<class InputCont, class Job>
	void executeBoss(InputCont& in, Job& job, PrecisionTimer* pt = 0) {
		ScopeTimer t(*pt);
		
		int nCPU = communicator.getSize();
		int nThis = communicator.getRank();
		if (nThis != 0)
			throw std::runtime_error("distributer node can only have rank 0!");
		
		if (nCPU == 1) {
			// no parallelism
			ExcludeScopeFromTimer et(*pt) ;
			for (size_t i = 0; i < in.size(); ++i)
				in[i].violation = job(in[i].chromosome, in[i].criteria);
		} else {
			// count number of individuals that need evaluation (violation must be -1)
			size_t numEvals = 0;
			for (size_t i = 0; i < in.size(); ++i)
				if (in[i].violation == -1) ++numEvals;
			
			size_t sendSize = size(in[0].chromosome);
			size_t receiveSize = size(in[0].chromosome) + size(in[0].criteria) + 1;
			
			// send orders (num of jobs to execute, size of chromosome and size of 
			// returned buffer - the whole individual [chromosome vector, criteria 
			// vector and violation code])
			for (int n = 1; n < nCPU; ++n) {
//				std::cout << "boss sent order " << (in.size() + nCPU - 1 - n) / (nCPU - 1) << " to " << n << std::endl;
				Mpi::send((int)((numEvals + nCPU - 1 - n) / (nCPU - 1)), n, 1);
				Mpi::send(sendSize, n, 1);
				Mpi::send(receiveSize, n, 1);
			}
			
			std::vector<double> receiveBuffer(receiveSize);
			
			// send data
			// indices hold index of each evaluated individual for each CPU
			std::vector<size_t> indices;
			indices.reserve(nCPU - 1);
			for (size_t i = 0; i < in.size();) {
				indices.clear();
				for (int n = 1; (n < nCPU) && (i < in.size());) {
					if (in[i].violation == -1) {
						indices.push_back(i);
						++n;
					}
					++i;
				}
				
				for (int n = 1; (n < nCPU) && (n <= indices.size()); ++n) {
					BinaryStream inStream;
					inStream << in[indices[n-1]].chromosome;
					Mpi::send(inStream, n, 1);
//					std::cout << "boss sent     " << (i+n) << std::endl;
				}
				
				for (int n = 1; (n < nCPU) && (n <= indices.size()); ++n) {
					BinaryStream stream;
//					std::cout << "boss expecting to receive results from " << (i+n) << " (" << receiveSize << ") \n";
					Mpi::receive(stream, n, 1);
					stream >> in[indices[n-1]].chromosome >> in[indices[n-1]].criteria 
						>> in[indices[n-1]].violation;
//					std::cout << "boss received " << (i+n) << std::endl;
				}
			}
//			std::cout << "boss: sending finished" << std::endl;
		}
	}
	
	template<class InputCont, class Job>
	void executeBossAndWorker(InputCont& in, Job& job, PrecisionTimer* pt = 0) {
		ScopeTimer t(*pt);
		
		int nCPU = communicator.getSize();
		int nThis = communicator.getRank();
		if (nThis != 0)
			throw std::runtime_error("distributer node can only have rank 0!");
		
		if (nCPU == 1) {
			// no parallelism
			ExcludeScopeFromTimer et(*pt);
			for (size_t i = 0; i < in.size(); ++i)
				in[i].violation = job(in[i].chromosome, in[i].criteria);
		} else {
			// count number of individuals that need evaluation (violation must be -1)
			size_t numEvals = 0;
			for (size_t i = 0; i < in.size(); ++i)
				if (in[i].violation == -1) ++numEvals;
			
			// packet sizes
			size_t sendSize = in[0].chromosome.size();
			size_t receiveSize = in[0].chromosome.size() + size(in[0].criteria) + 1;
			
			// send orders (num of jobs to execute, size of chromosome and size of 
			// returned buffer - the whole individual [chromosome vector, criteria 
			// vector and violation code])
			for (int n = 1; n < nCPU; ++n) {
//				std::cout << "boss sent order " << (in.size() + nCPU - 1 - n) / (nCPU - 1) << " to " << n << std::endl;
				Mpi::send((int)((numEvals + nCPU - 1 - n) / nCPU), n, 1); //++++
				Mpi::send(sendSize, n, 1);
				Mpi::send(receiveSize, n, 1);
			}
			
			std::vector<double> receiveBuffer(receiveSize);
			
			// send data
			// indices hold index of each evaluated individual for each CPU
			std::vector<size_t> indices;
			indices.reserve(nCPU);
			for (size_t i = 0; i < in.size();) {
				// select only individuals with no violation
				indices.clear();
				for (int n = 0; (n < nCPU) && (i < in.size());) {
					if (in[i].violation == -1) {
						indices.push_back(i);
						++n;
					}
					++i;
				}
				
				// send individuals to evaluate
				for (int n = 1; (n < nCPU) && (n < indices.size()); ++n) {
					BinaryStream inStream;
					inStream << in[indices[n]].chromosome;
					Mpi::send(inStream, n, 1);
//					std::cout << "boss sent     " << (i+n) << std::endl;
				} 
				
				// execute one evaluation locally
				{
					ExcludeScopeFromTimer et(*pt);
					in[indices[0]].violation = job(in[indices[0]].chromosome, in[indices[0]].criteria);
				}
				
				// gather results
				for (int n = 1; (n < nCPU) && (n < indices.size()); ++n) {
					BinaryStream stream;
//					std::cout << "boss expecting to receive results from " << (i+n) << " (" << receiveSize << ") \n";
					Mpi::receive(stream, n, 1);
					stream >> in[indices[n-1]].chromosome >> in[indices[n-1]].criteria 
						>> in[indices[n-1]].violation;
				}
			}
//			std::cout << "boss: sending finished" << std::endl;
		}
	}
};


#endif // JOB_DISTRIBUTER_H_INCLUDED