improved

src/files/Supervisor.h

@@ -5,6 +5,7 @@
 #include <vector>
 #include <queue>
 #include <math.h>
+#include <thread>
 #include "File.h"   
 #include "Task.h"
 #include "Array.h"
@@ -107,15 +108,30 @@ public:
 
         map<int, queue<T*>, std::greater<int>> sortedByKernelNeeds;
         size_t activeTasks = 0;
-        
+        int maxNeededKernels = 0;
+
         for (auto& task : this->getElements()) {
             if (task->getState() == WorkspaceReady) {
                 activeTasks++;
                 sortedByKernelNeeds[task->getKernels()].push(task);
+                if (maxNeededKernels < task->getKernels())
+                    maxNeededKernels = task->getKernels();
             }
         }
 
-        printf("total tasks count = %ld, active task count %ld, maxkernels %d\n", this->getLength(), activeTasks, maxKernels);
+        const int maxThreads = std::thread::hardware_concurrency();
+        printf("total tasks count = %ld, active task count %ld, maxKernels %d, maxNeededKernels %d, maxThreads %d\n", 
+                this->getLength(), activeTasks, maxKernels, maxNeededKernels, maxThreads);
+        
+        if (maxKernels > maxThreads) {
+            printf("truncated maxKernels to maxThreads: %d -> %d\n", maxKernels, maxThreads);
+            maxKernels = maxThreads;
+        }
+        
+        if (maxNeededKernels > maxKernels) {
+            printf("increased maxKernels to maxNeededKernels: %d -> %d\n", maxKernels, maxNeededKernels);
+            maxKernels = maxNeededKernels;
+        }
 
         int busyKernels = 0;
         set<T*> activeTaskSet;
@@ -129,7 +145,6 @@ public:
         vector<T*> toDel;
         
         size_t done = 0;
-        //size_t step = activeTasks * 0.01; // step == 1%
         size_t step = ceil(activeTasks * 0.01); // step == 1%
         const double total = activeTasks;