change fitness to throughput

Appli-iso3dfd/ACO.py

@@ -68,7 +68,7 @@ def ACO(Me, NbP, comm, alpha, rho, Q, nb_ants, tau_0, n_iter, n1=256, n2=256, n3
 		comm.Gather(costs, all_costs, root=0)
 
 		if Me == 0:
-			tau, best_p, best_cost = compute_tau.compute_tau(tau, all_paths, all_costs, Q, rho, n1, n2, n3)
+			tau, best_p, best_cost = compute_tau.compute_tau(tau, all_paths, all_costs, Q, rho, n1, n2, n3, fancy_strategy, sub_threshold, sup_threshold)
 			if best_cost < cost_opti:
 				cost_opti = best_cost
 				path_opti = best_p
@@ -90,9 +90,9 @@ if __name__ == "__main__":
 	# Initialization of hyperparameters
 	alpha = 1
 	rho = 0.1
-	Q = 10
+	Q = 0.1
 	nb_ants = 2
-	tau_0 = Q/nb_ants
+	tau_0 = Q*100
 	n_iter = 1
 
 	# Parameters for compilation and execution of iso3dfd
@@ -113,8 +113,9 @@ if __name__ == "__main__":
 		path_opti, cost_opti = ACO(Me, NbP, comm, alpha, rho, Q, nb_ants, tau_0, n_iter,
 									n1=n1, n2=n2, n3=n3, fancy_strategy=fancy_strategy,
 									nb_threads=nb_threads, reps=reps)
-		print(path_opti)
-		print(cost_opti)
+		print(f"Le chemin optimal est {path_opti}.")
+		print(f"Le throughput associé est alors {-1*cost_opti}.")
+
 	else:
 		ACO(Me, NbP, comm, alpha, rho, Q, nb_ants, tau_0, n_iter,
 			n1=n1, n2=n2, n3=n3, fancy_strategy=fancy_strategy,

Appli-iso3dfd/compute_tau.py

@@ -32,16 +32,16 @@ def compute_tau(tau, all_paths, all_costs, Q, rho, n1, n2, n3, fancy_strategy='A
     if fancy_strategy == "AS" or fancy_strategy == 'ElitistAS':
       #Classic Ant System : each path is rewarded according to its length
       #Elitist Ant System : each path is rewarded according to its length
-      tau[0,p[0]] += Q/cost
-      tau[p[0],n1//16+p[1]] += Q/cost
-      tau[n1//16+p[1],n1//16+n2+p[2]] += Q/cost
+      tau[0,p[0]] += Q*(-1)*cost
+      tau[p[0],n1//16+p[1]] += Q*(-1)*cost
+      tau[n1//16+p[1],n1//16+n2+p[2]] += Q*(-1)*cost
     
   if fancy_strategy == 'ElitistAS' or fancy_strategy == "MMAS":
     #Elitist Ant System : The best ant is rewarded a second time
     #Max-Min Ant System : Only the winner ant is rewarded but the pheromon are limited within a minimal and a maximal threshold
-    tau[0,best_p[0]] += Q/best_cost
-    tau[best_p[0],n1//16+best_p[1]] += Q/best_cost
-    tau[n1//16+best_p[1],n1//16+n2+best_p[2]] += Q/best_cost
+    tau[0,best_p[0]] += Q*(-1)*best_cost
+    tau[best_p[0],n1//16+best_p[1]] += Q*(-1)*best_cost
+    tau[n1//16+best_p[1],n1//16+n2+best_p[2]] += Q*(-1)*best_cost
 
   if fancy_strategy == "MMAS":
     #verification of the threshold constraint

Appli-iso3dfd/launcher_SUBP.py

@@ -41,12 +41,12 @@ def deploySUBP(n1, n2, n3, nb_threads, reps, cbx, cby, cbz):
     times, throughputs, gflops, runs = tools.commandLineExtract(res)
     # Fitness calculation
     fitness = 0.
-    for time in times:
-      fitness += time
+    for throughput in throughputs:
+      fitness += throughput
     if res:
-      fitness /= len(times)
+      fitness /= len(throughputs)
 
-    return fitness
+    return -1*fitness
 
 
 #-----------------------------------------------------------------