WorkerThreadPool: Avoid most runtime allocations

Just a little optimization.

**NOTE:**
With `RID_Owner` we could replace each pair of `PagedAllocator` and
`HashMap`-of-ids-to-pointers. However, that would force us to expose
`RID` as the task/group id, instead of `int`, which would break the
API. Too bad. Let's wait until Godot 5.0.

core/object/worker_thread_pool.cpp

@@ -613,13 +613,14 @@ void WorkerThreadPool::finish() {
 		return;
 	}
 
-	task_mutex.lock();
-	SelfList<Task> *E = low_priority_task_queue.first();
-	while (E) {
-		print_error("Task waiting was never re-claimed: " + E->self()->description);
-		E = E->next();
+	{
+		MutexLock lock(task_mutex);
+		SelfList<Task> *E = low_priority_task_queue.first();
+		while (E) {
+			print_error("Task waiting was never re-claimed: " + E->self()->description);
+			E = E->next();
+		}
 	}
-	task_mutex.unlock();
 
 	{
 		MutexLock lock(task_mutex);
@@ -632,6 +633,13 @@ void WorkerThreadPool::finish() {
 		data.thread.wait_to_finish();
 	}
 
+	{
+		MutexLock lock(task_mutex);
+		for (KeyValue<TaskID, Task *> &E : tasks) {
+			task_allocator.free(E.value);
+		}
+	}
+
 	threads.clear();
 }
 

core/object/worker_thread_pool.h

@@ -95,8 +95,11 @@ class WorkerThreadPool : public Object {
 				task_elem(this) {}
 	};
 
-	PagedAllocator<Task> task_allocator;
-	PagedAllocator<Group> group_allocator;
+	static const uint32_t TASKS_PAGE_SIZE = 1024;
+	static const uint32_t GROUPS_PAGE_SIZE = 256;
+
+	PagedAllocator<Task, false, TASKS_PAGE_SIZE> task_allocator;
+	PagedAllocator<Group, false, GROUPS_PAGE_SIZE> group_allocator;
 
 	SelfList<Task>::List low_priority_task_queue;
 	SelfList<Task>::List task_queue;
@@ -117,8 +120,20 @@ class WorkerThreadPool : public Object {
 	bool exit_threads = false;
 
 	HashMap<Thread::ID, int> thread_ids;
-	HashMap<TaskID, Task *> tasks;
-	HashMap<GroupID, Group *> groups;
+	HashMap<
+			TaskID,
+			Task *,
+			HashMapHasherDefault,
+			HashMapComparatorDefault<TaskID>,
+			PagedAllocator<HashMapElement<TaskID, Task *>, false, TASKS_PAGE_SIZE>>
+			tasks;
+	HashMap<
+			GroupID,
+			Group *,
+			HashMapHasherDefault,
+			HashMapComparatorDefault<GroupID>,
+			PagedAllocator<HashMapElement<GroupID, Group *>, false, GROUPS_PAGE_SIZE>>
+			groups;
 
 	uint32_t max_low_priority_threads = 0;
 	uint32_t low_priority_threads_used = 0;

core/templates/paged_allocator.h

@@ -40,7 +40,7 @@
 #include <type_traits>
 #include <typeinfo>
 
-template <class T, bool thread_safe = false>
+template <class T, bool thread_safe = false, uint32_t DEFAULT_PAGE_SIZE = 4096>
 class PagedAllocator {
 	T **page_pool = nullptr;
 	T ***available_pool = nullptr;
@@ -53,10 +53,6 @@ class PagedAllocator {
 	SpinLock spin_lock;
 
 public:
-	enum {
-		DEFAULT_PAGE_SIZE = 4096
-	};
-
 	template <class... Args>
 	T *alloc(Args &&...p_args) {
 		if (thread_safe) {