private void DoCrossoverForBestIndividuals()
{
int length = firstFront.Count;
int numberOfConfigurations = allConfigurations.Count;
while (nsgaPopulation.Count < numberOfConfigurations)
{
int index_parent1, index_parent2;
int count = 10;
do
{
index_parent1 = random.Next(0, bestIndividuals.Count);
index_parent2 = random.Next(0, bestIndividuals.Count);
count--;
} while (index_parent1 == index_parent2 && count != 0);
if (bestIndividuals.Count > 1)
{
if (count == 0)
{
index_parent1 = bestIndividuals.Count - 1;
index_parent2 = bestIndividuals.Count - 2;
}
Configuration parent1 = bestIndividuals[index_parent1];
Configuration parent2 = bestIndividuals[index_parent2];
List <Configuration> resultedConfigurations = Crossover(parent1, parent2);
allConfigurations.AddRange(resultedConfigurations);
var newSimulatedConfigurations = sim.Run(resultedConfigurations, selectedTraces);
foreach (var cfg in resultedConfigurations)
{
if (nsgaPopulation.Count < numberOfConfigurations)
{
double ipc_average = 0, power_average = 0, energy_average = 0;
int ct = 0;
foreach (var item in newSimulatedConfigurations)
{
if (item.cfg == cfg)
{
ipc_average += item.ipc;
power_average += item.power;
energy_average += item.energy;
ct++;
}
}
ipc_average /= ct;
power_average /= ct;
energy_average /= ct;
GeneticIndividual tempIndividual = new GeneticIndividual(cfg, ipc_average, power_average, energy_average);
nsgaPopulation.Add(tempIndividual);
}
}
}
}
}
public bool InitRandomPopulation(List <string> selectedTraces)
{
ConfigurationGenerator cg = new ConfigurationGenerator();
for (int i = 0; i < populationSize; i++)
{
population.Add(cg.RandomConfig(individualID.ToString()));
}
allConfigurations.AddRange(population);
var results = sim.Run(population, selectedTraces);
for (int i = 0; i < population.Count; i++)
{
population[i] = new GeneticIndividual(population[i], 0.0, 0.0, 0.0);
for (int j = 0; j < selectedTraces.Count; j++)
{
((GeneticIndividual)population[i]).Ipc += 1.0 / results[i + population.Count * j].ipc;
((GeneticIndividual)population[i]).Power += results[i + population.Count * j].power;
((GeneticIndividual)population[i]).Energy += results[i + population.Count * j].energy;
}
if (selectedTraces.Count > 1)
{
((GeneticIndividual)population[i]).Ipc /= selectedTraces.Count;
((GeneticIndividual)population[i]).Power /= selectedTraces.Count;
((GeneticIndividual)population[i]).Energy /= selectedTraces.Count;
}
nsgaPopulation.Add(new GeneticIndividual(population[i], ((GeneticIndividual)population[i]).Ipc, ((GeneticIndividual)population[i]).Power, ((GeneticIndividual)population[i]).Energy));
}
return(true);
}
public double Calculatedensity(GeneticIndividual individual, List <Configuration> pop)
{
foreach (var other in pop)
{
((GeneticIndividual)other).distance = EuclideanDistance(((GeneticIndividual)individual), (GeneticIndividual)other);
}
var sorted = pop.OrderBy(c => (((GeneticIndividual)c).distance)).ToArray();
int k = (int)Math.Sqrt(pop.Count);
return(1.0 / (((GeneticIndividual)sorted[k]).distance + 2.0));
}
public int Dominates(GeneticIndividual ind1, GeneticIndividual ind2)
{
if (ind1.Energy < ind2.Energy && ind1.Ipc < ind2.Ipc)
{
return(1);
}
if (ind1.Energy >= ind2.Energy && ind1.Ipc >= ind2.Ipc)
{
return(-1);
}
return(0);
}
private void DoMutationAndSimulation()
{
foreach (var bestIndividual in bestIndividuals)
{ //preiau ultimul individ de pe primul front ( 0 )
Configuration configuration = new Configuration();
if (bestIndividual.FrontNumber == 0)
{
try
{
foreach (var individual in population)
{
if ((bestIndividual.Name.Substring(bestIndividual.Name.LastIndexOf('_') + 1)).Equals(individual.Name))
{
configuration = individual;
break;
}
}
Configuration mutatedConfiguration = Mutate((Configuration)configuration.Clone());
mutatedConfiguration.Name = id.ToString();
id++;
allConfigurations.Add(mutatedConfiguration);
//rulare simulator doar pt mutatedConfiguration
var newSimulatedConfigurations = sim.Run(new List <Configuration> {
mutatedConfiguration
}, selectedTraces);
double ipc_average = 0, power_average = 0, energy_average = 0;
foreach (var item in newSimulatedConfigurations)
{
ipc_average += item.ipc;
power_average += item.power;
energy_average += item.energy;
}
ipc_average /= newSimulatedConfigurations.Count;
power_average /= newSimulatedConfigurations.Count;
energy_average /= newSimulatedConfigurations.Count;
GeneticIndividual tempIndividual = new GeneticIndividual(mutatedConfiguration, ipc_average, power_average, energy_average);
nsgaPopulation.Add(tempIndividual);
// Thread.Sleep(20);
}
catch (Exception e)
{
Console.WriteLine(e.Message);
}
}
}
}
public void CalculateObjectives()
{
results = sim.Run(population, selectedTraces);
for (int i = 0; i < population.Count; i++)
{
population[i] = new GeneticIndividual(population[i], 0.0, 0.0, 0.0);
for (int j = 0; j < selectedTraces.Count; j++)
{
((GeneticIndividual)population[i]).Ipc += 1.0 / results[i + population.Count * j].ipc;
((GeneticIndividual)population[i]).Power += results[i + population.Count * j].power;
((GeneticIndividual)population[i]).Energy += results[i + population.Count * j].energy;
}
if (selectedTraces.Count > 1)
{
((GeneticIndividual)population[i]).Ipc /= selectedTraces.Count;
((GeneticIndividual)population[i]).Power /= selectedTraces.Count;
((GeneticIndividual)population[i]).Energy /= selectedTraces.Count;
}
}
}
private double EuclideanDistance(GeneticIndividual individual, GeneticIndividual other)
{
return(Math.Sqrt(Math.Pow(individual.Ipc - other.Ipc, 2) + Math.Pow(individual.Energy - other.Energy, 2)));
}
