Add nowait into other thread FMM (Target/Source Model and Unsafe model)

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/scalfmm/scalfmm/trunk@133 2616d619-271b-44dc-8df4-d4a8f33a7222

Src/Core/FFmmAlgorithmThreadProc.hpp

@@ -783,7 +783,7 @@ public:
                     #pragma omp parallel
                     {
                         Kernel * const myThreadkernels = kernels[omp_get_thread_num()];
-                        #pragma omp for
+                        #pragma omp for nowait
                         for(int idxCell = startIdx ; idxCell < endIdx ; ++idxCell){
                             myThreadkernels->L2L( iterArray[idxCell].getCurrentCell() , iterArray[idxCell].getCurrentChild(), idxLevel);
                         }

Src/Core/FFmmAlgorithmThreadTsm.hpp

@@ -124,7 +124,7 @@ public:
         #pragma omp parallel
         {
             Kernel * const myThreadkernels = kernels[omp_get_thread_num()];
-            #pragma omp for
+            #pragma omp for nowait
             for(int idxLeafs = 0 ; idxLeafs < numberOfLeafs ; ++idxLeafs){
                 // We need the current cell that represent the leaf
                 // and the list of particles
@@ -174,7 +174,7 @@ public:
             #pragma omp parallel
             {
                 Kernel * const myThreadkernels = kernels[omp_get_thread_num()];
-                #pragma omp for
+                #pragma omp for nowait
                 for(int idxCell = 0 ; idxCell < numberOfCells ; ++idxCell){
                     // We need the current cell and the child
                     // child is an array (of 8 child) that may be null
@@ -235,7 +235,7 @@ public:
                     Kernel * const myThreadkernels = kernels[omp_get_thread_num()];
                     const CellClass* neighbors[208];
 
-                    #pragma omp for
+                    #pragma omp for nowait
                     for(int idxCell = 0 ; idxCell < numberOfCells ; ++idxCell){
                         CellClass* const currentCell = iterArray[idxCell].getCurrentCell();
                         if(currentCell->hasTargetsChild()){
@@ -289,7 +289,7 @@ public:
                 #pragma omp parallel
                 {
                     Kernel * const myThreadkernels = kernels[omp_get_thread_num()];
-                    #pragma omp for
+                    #pragma omp for nowait
                     for(int idxCell = 0 ; idxCell < numberOfCells ; ++idxCell){
                         CellClass* potentialChild[8];
                         CellClass** const realChild = iterArray[idxCell].getCurrentChild();
@@ -339,7 +339,7 @@ public:
             // There is a maximum of 26 neighbors
             ContainerClass<ParticleClass>* neighbors[26];
 
-            #pragma omp for
+            #pragma omp for schedule(dynamic)
             for(int idxLeafs = 0 ; idxLeafs < numberOfLeafs ; ++idxLeafs){
                 myThreadkernels->L2P(iterArray[idxLeafs].getCurrentCell(), iterArray[idxLeafs].getCurrentListTargets());
                 // need the current particles and neighbors particles

Src/Core/FFmmAlgorithmThreadUs.hpp

@@ -128,7 +128,7 @@ public:
         #pragma omp parallel
         {
             Kernel * const myThreadkernels = kernels[omp_get_thread_num()];
-            #pragma omp for
+            #pragma omp for nowait
             for(int idxLeafs = 0 ; idxLeafs < numberOfLeafs ; ++idxLeafs){
                 // We need the current cell that represent the leaf
                 // and the list of particles
@@ -170,7 +170,7 @@ public:
             #pragma omp parallel
             {
                 Kernel * const myThreadkernels = kernels[omp_get_thread_num()];
-                #pragma omp for
+                #pragma omp for nowait
                 for(int idxCell = 0 ; idxCell < numberOfCells ; ++idxCell){
                     // We need the current cell and the child
                     // child is an array (of 8 child) that may be null
@@ -215,7 +215,7 @@ public:
                     Kernel * const myThreadkernels = kernels[omp_get_thread_num()];
                     CellClass* neighbors[208];
 
-                    #pragma omp for
+                    #pragma omp for nowait
                     for(int idxCell = 0 ; idxCell < numberOfCells ; ++idxCell){
                         const int counter = tree->getDistantNeighbors(neighbors, iterArray[idxCell].getCurrentGlobalCoordinate(),idxLevel);
                         if(counter) myThreadkernels->M2L(  iterArray[idxCell].getCurrentCell() , neighbors, counter, idxLevel);
@@ -253,7 +253,7 @@ public:
                 #pragma omp parallel
                 {
                     Kernel * const myThreadkernels = kernels[omp_get_thread_num()];
-                    #pragma omp for
+                    #pragma omp for nowait
                     for(int idxCell = 0 ; idxCell < numberOfCells ; ++idxCell){
                         myThreadkernels->L2L( iterArray[idxCell].getCurrentCell() , iterArray[idxCell].getCurrentChild(), idxLevel);
                     }
@@ -292,7 +292,7 @@ public:
             // There is a maximum of 26 neighbors
             ContainerClass<ParticleClass>* neighbors[26];
 
-            #pragma omp for
+            #pragma omp for schedule(dynamic)
             for(int idxLeafs = 0 ; idxLeafs < numberOfLeafs ; ++idxLeafs){
                 myThreadkernels->L2P(iterArray[idxLeafs].getCurrentCell(), iterArray[idxLeafs].getCurrentListTargets());
                 // need the current particles and neighbors particles