paradiseo/ParadisEO-GPU/src/eval/moCudakernelEval.h

204 lines
8.1 KiB
C++

/*
<moCudakernelEval.h>
Copyright (C) DOLPHIN Project-Team, INRIA Lille - Nord Europe, 2006-2010
Karima Boufaras, Thé Van LUONG
This software is governed by the CeCILL 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
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 license and that you accept its terms.
ParadisEO WebSite : http://paradiseo.gforge.inria.fr
Contact: paradiseo-help@lists.gforge.inria.fr
*/
#ifndef __moCudakernelEval_H
#define __moCudakernelEval_H
///////////////////////////////////////////////////////////////////////////////////////////////////////////////
/**
* The kernel function called from the host and executed in device to compute all neighbors fitness at one time
* @param _eval how to evaluate each neighbor
* @param _solution representation of solution( vector of int,float....)
* @param _allFitness Array of Fitness to save all neighbors fitness
* @param _fitness the current solution fitness
* @param _neighborhoodsize the size of the neighborhood
*/
template<class EOT, class Fitness, class Neighbor, class IncrementEval>
__global__ void kernelEval(IncrementEval _eval, EOT _solution, Fitness* _allFitness,
Fitness _fitness, unsigned _neighborhoodsize) {
// The thread identifier within a grid block's
int id = blockIdx.x * blockDim.x + threadIdx.x;
// array to save index to be changed
unsigned int index[1];
// In this representation each id identify one and only one neighbor in neighborhood
if (id < _neighborhoodsize) {
//Change the id'th element of solution
index[0]=id;
//Compute fitness for id'th neighbor
_allFitness[id] = _eval(_solution, _fitness,index);
}
}
///////////////////////////////////////////////////////////////////////////////////////////////////////////////
/**
* The kernel function called from the host and executed in device to compute all flip neighbors fitness at one time
* @param _eval how to evaluate each neighbor
* @param _solution representation of solution to flip
* @param _allFitness Array of Fitness type to save all neighbors fitness
* @param _fitness the current solution fitness
* @param _neighborhoodsize the size of the neighborhood
* @param _mapping the neighborhood mapping
* @param _Kflip the number of bit to flip
*/
template<class EOT, class Fitness, class Neighbor, class IncrementEval>
__global__ void kernelKflip(IncrementEval _eval, EOT _solution, Fitness* _allFitness,
Fitness _fitness, unsigned _neighborhoodsize, unsigned * _mapping,unsigned _Kflip) {
// The thread identifier within a grid block's
int id = blockIdx.x * blockDim.x + threadIdx.x;
//save temporary fitness
unsigned tmp_fitness;
//counter of number of flip to do
unsigned i;
// array to save index to be changed
unsigned index[1];
// In this representation each id identify one and only one neighbor in neighborhood
if (id < _neighborhoodsize) {
//Init fitness with fitness of solution
tmp_fitness=_fitness;
//Evaluate neighbor after Kflip
for(i=0;i<=_Kflip;i++){
//The designed index to flip
index[0]=_mapping[id + i * _neighborhoodsize];
//Evaluate the neighbor
tmp_fitness= _eval(_solution, tmp_fitness, index);
}
//The final fitness of the Id'th neighbor
_allFitness[id]=tmp_fitness;
}
}
///////////////////////////////////////////////////////////////////////////////////////////////////////////////
/**
* The kernel function called from the host and executed in device to compute all swap neighbors fitness at one time
* @param _eval how to evaluate each neighbor
* @param _solution representation ofsolution to swap
* @param _sol_tmp to save temporary a solution element to swap
* @param _allFitness Array of Fitness type to save all neighbors fitness
* @param _fitness the current solution fitness
* @param _neighborhoodsize the size of the neighborhood
* @param _mapping the neighborhood mapping
* @param _Kswap the number of swap to do
* @param _size the solution size
*/
template<class EOT,class Fitness, class Neighbor, class IncrementEval>
__global__ void kernelKswap(IncrementEval _eval,EOT _solution ,EOT _sol_tmp, Fitness* _allFitness,
Fitness _fitness, unsigned _neighborhoodsize, unsigned * _mapping,unsigned _Kswap,unsigned _size) {
// The thread identifier within a grid block's
int id = blockIdx.x * blockDim.x + threadIdx.x;
//save temporary fitness
int tmp_fitness;
//counter of number of swap to do
unsigned i;
// array to save index to be changed, solution size & thread id
unsigned index[4];
// In this representation each id identify one and only one neighbor in neighborhood
if (id < _neighborhoodsize) {
//the first index to swap
index[0]=_mapping[id];
//the second index to swap
index[1]=_mapping[id +_neighborhoodsize];
//the solution size
index[2]=_size;
//the thread id
index[3]=id;
//Init the temporary fitness with the initial solution fitness
tmp_fitness=_fitness;
//Evaluate neighbor after K-swap
for(i=2;i<=_Kswap+1;i++){
//Evaluate neighbor with index case
tmp_fitness=_eval(_solution, tmp_fitness, index);
//Permut the solution
_sol_tmp[id]=_solution[index[0]+id*index[2]];
_solution[index[0]+id*index[2]]=_solution[index[1]+id*index[2]];
_solution[index[1]+id*index[2]]=_sol_tmp[id];
//Init the next swap to do
index[0]=index[1];
index[1]=_mapping[id +i*_neighborhoodsize];
}
//save the final fitness of the id'th neighbor
_allFitness[id]=tmp_fitness;
}
}
///////////////////////////////////////////////////////////////////////////////////////////////////////////////
/**
* The kernel function called from the host and executed in device to compute all permutation neighbors fitness at one time
* @param _eval how to evaluate each neighbor
* @param _solution representation of solution
* @param _allFitness Array of Fitness type to save all neighbors fitness
* @param _fitness the current solution fitness
* @param _neighborhoodsize the size of the neighborhood
* @param _mapping the neighborhood mapping
* @param _size the solution size
*/
template<class EOT, class Fitness, class Neighbor, class IncrementEval>
__global__ void kernelPermutation(IncrementEval _eval, EOT _solution, Fitness* _allFitness,
Fitness _fitness, unsigned _neighborhoodsize, unsigned * _mapping,unsigned _size) {
// The thread identifier within a grid block's
int id = blockIdx.x * blockDim.x + threadIdx.x;
// array to save index to be changed, solution size
unsigned index[4];
// In this representation each id identify one and only one neighbor in neighborhood
if (id < _neighborhoodsize) {
//The first index of permutation
index[0]=_mapping[id];
//The second index of permutation
index[1]=_mapping[id +_neighborhoodsize];
//The solution size
index[2]=_size;
//Puch 0 in the 3 index
index[3]=0;
_allFitness[id]=_eval(_solution,_fitness,index);
}
}
#endif