--- trunk/gadget/hooke.cc	2015/07/24 18:36:24	12
+++ trunk/gadget/hooke.cc	2015/09/01 19:14:33	15
@@ -156,17 +156,9 @@
   DoubleVector z(point);
 
   minf = prevbest;
-//  for (int k=0;k<point.Size(); k++)
-//	  cout << z[k] ;
-//  cout << endl;
   for (i = 0; i < point.Size(); i++) {
-
-//	  for (int k=0;k<point.Size(); k++)
-//	  	  cout << z[k] << " " ;
-//cout << endl;
     z[param[i]] = point[param[i]] + delta[param[i]];
     ftmp = EcoSystem->SimulateAndUpdate(z);
-//    cout << i <<"-z["<< param[i]<<"]:" <<z[param[i]] << " - " << ftmp << endl;
     if (ftmp < minf) {
       minf = ftmp;
     } else {
@@ -178,7 +170,6 @@
       else
         z[param[i]] = point[param[i]];
     }
-//    cout << i <<"-z["<< param[i]<<"]:" <<z[param[i]] << " - " << ftmp  <<" - " << prevbest << endl;
   }
 
   for (i = 0; i < point.Size(); i++)
@@ -193,7 +184,7 @@
  * · 2 threads per coord to parallelize the calculation of +delta/-delta
  * · parallelize the calculation of the best nearby of the coord
  */
-double OptInfoHooke::bestNearbyOMP(DoubleVector& delta, DoubleVector& point, double prevbest, IntVector& param) {
+double OptInfoHooke::bestNearbyRepro(DoubleVector& delta, DoubleVector& point, double prevbest, IntVector& param) {
   double minf;//, ftmp;
   int i, j, k;
   DoubleVector z(point);
@@ -218,12 +209,12 @@
 	  return -1;
   }
 
+  omp_set_dynamic(0);
+  omp_set_nested(1); //permit the nested parallelization
   while ( i < nvars) {
 	  if ((i + paral_tokens -1) >= nvars)
 		  paral_tokens = nvars - i;
-	  omp_set_dynamic(0);
-	  omp_set_nested(1);
-#pragma omp parallel for num_threads(paral_tokens) private(k)
+#pragma omp parallel for num_threads(paral_tokens) private(k) //parallelize the parameters (numThr/2)
 	  for (j = 0; j < paral_tokens; ++j) {
 		  storage[j].z = z;
 		  storage[j].delta = delta;
@@ -233,7 +224,7 @@
 		  v1[k] +=  delta[k];
 		  v2[k] -=  delta[k];
 
-#pragma omp parallel sections num_threads(2)
+#pragma omp parallel sections num_threads(2) //parrallelize the +/- delta simulation for each parameter
 		  {
 	#pragma omp section
 			  {
@@ -269,7 +260,7 @@
 		  }
 	  }
   }
-
+  delete[] storage;
   for (i = 0; i < nvars; ++i)
     point[i] = z[i];
   return minf;
@@ -300,7 +291,7 @@
   EcoSystem->scaleVariables();
 #ifndef NO_OPENMP
   int numThr = omp_get_max_threads ( );
-  for (i = 0; i < numThr; i++)
+  for (i = 0; i < numThr; i++) // scale the variables for the ecosystem of every thread
 	  EcoSystems[i]->scaleVariables();
 #endif
   EcoSystem->getOptScaledValues(x);
@@ -369,7 +360,7 @@
     for (i = 0; i < nvars; i++)
       trialx[i] = x[i];
 #ifndef NO_OPENMP
-    newf = this->bestNearbyOMP(delta, trialx, bestf, param);
+    newf = this->bestNearbyRepro(delta, trialx, bestf, param);
     if (newf == -1) {
     	handle.logMessage(LOGINFO, "\nStopping Hooke & Jeeves optimisation algorithm\n");
     	handle.logMessage(LOGINFO, "\nThe number of threads must be a multiple of 2\n");
@@ -464,7 +455,7 @@
         x[i] = trialx[i];
 
 #ifndef NO_OPENMP
-      newf = this->bestNearbyOMP(delta, trialx, bestf, param);
+      newf = this->bestNearbyRepro(delta, trialx, bestf, param);
       if (newf == -1) {
           	handle.logMessage(LOGINFO, "\nStopping Hooke & Jeeves optimisation algorithm\n");
           	handle.logMessage(LOGINFO, "\nThe number of threads must be a multiple of 2\n");
@@ -533,8 +524,8 @@
 }
 
 /* Functions to perform the parallelization of the algorithm of HJ with OpenMP*/
-#ifdef GADGET_OPENMP
-double OptInfoHooke::bestNearbyOMP2(DoubleVector& delta, DoubleVector& point, double prevbest, IntVector& param) {
+#ifdef SPECULATIVE
+double OptInfoHooke::bestNearbySpec(DoubleVector& delta, DoubleVector& point, double prevbest, IntVector& param) {
   double minf;
   int i, j, k, ii;
   DoubleVector z(point);
@@ -558,13 +549,13 @@
 	  return -1;
   }
 
+  omp_set_dynamic(0);
+  omp_set_nested(1); //permit the nested parallelization
   for (ii=0; ii< paral_tokens; ii++) {
 	  i = 0;
 	  while ( i < nvars) {
 		  if ((i + paral_tokens -1) >= nvars)
 			  paral_tokens = nvars - i;
-		  omp_set_dynamic(0);
-		  omp_set_nested(1);
 	#pragma omp parallel for num_threads(paral_tokens) private(k)
 		  for (j = 0; j < paral_tokens; ++j) {
 			  storage[j].z = z;
@@ -644,11 +635,9 @@
 	   IntVector trapped(nvars, 0);
 
 	   EcoSystem->scaleVariables();
-	 #ifndef NO_OPENMP
 	   int numThr = omp_get_max_threads ( );
-	   for (i = 0; i < numThr; i++)
+	   for (i = 0; i < numThr; i++) // scale the variables for the ecosystem of every thread
 	 	  EcoSystems[i]->scaleVariables();
-	 #endif
 	   EcoSystem->getOptScaledValues(x);
 	   EcoSystem->getOptLowerBounds(lowerb);
 	   EcoSystem->getOptUpperBounds(upperb);
@@ -715,7 +704,7 @@
 	     for (i = 0; i < nvars; i++)
 	       trialx[i] = x[i];
 	 #ifndef NO_OPENMP
-	     newf = this->bestNearbyOMP2(delta, trialx, bestf, param);
+	     newf = this->bestNearbySpec(delta, trialx, bestf, param);
 	     if (newf == -1) {
 	     	handle.logMessage(LOGINFO, "\nStopping Hooke & Jeeves optimisation algorithm\n");
 	     	handle.logMessage(LOGINFO, "\nThe number of threads must be a multiple of 2\n");
@@ -810,7 +799,7 @@
 	         x[i] = trialx[i];
 
 	 #ifndef NO_OPENMP
-	       newf = this->bestNearbyOMP2(delta, trialx, bestf, param);
+	       newf = this->bestNearbySpec(delta, trialx, bestf, param);
 	       if (newf == -1) {
 	           	handle.logMessage(LOGINFO, "\nStopping Hooke & Jeeves optimisation algorithm\n");
 	           	handle.logMessage(LOGINFO, "\nThe number of threads must be a multiple of 2\n");