Cycles / CPU Rendering:

* "Auto Detect" now again uses the umber of cores, instead number of cores + 1.

This was added before we had Tile rendering and benchmarks on several systems showed that there is no gain with this now. There might be some slight difference (0.5% or so) slower/faster depending on the scene, but this is negligible.

intern/cycles/device/device_cpu.cpp

@@ -307,8 +307,7 @@ public:
 
 	void task_add(DeviceTask& task)
 	{
-		/* split task into smaller ones, more than number of threads for uneven
-		 * workloads where some parts of the image render slower than others */
+		/* split task into smaller ones */
 		list<DeviceTask> tasks;
 		task.split(tasks, TaskScheduler::num_threads());
 

intern/cycles/util/util_task.cpp

@@ -186,12 +186,12 @@ void TaskScheduler::init(int num_threads)
 		do_exit = false;
 
 		if(num_threads == 0) {
-			/* automatic number of threads will be main thread + num cores */
+			/* automatic number of threads */
 			num_threads = system_cpu_thread_count();
 		}
 		else {
-			/* main thread will also work, for fixed threads we count it too */
-			num_threads -= 1;
+			/* manual number of threads */
+			num_threads;
 		}
 
 		/* launch threads that will be waiting for work */

intern/cycles/util/util_task.h

@@ -94,8 +94,8 @@ public:
 	static void init(int num_threads = 0);
 	static void exit();
 
-	/* number of threads that can work on tasks, main thread counts too */
-	static int num_threads() { return threads.size() + 1; }
+	/* number of threads that can work on task */
+	static int num_threads() { return threads.size(); }
 
 	/* test if any session is using the scheduler */
 	static bool active() { return users != 0; }