~30% performance improvement with threading enabled

include/chaiscript/chaiscript_threading.hpp

@@ -64,8 +64,7 @@ namespace chaiscript
 
       using std::recursive_mutex;
 
-
-
+#ifdef CHAISCRIPT_HAS_THREAD_LOCAL
       /// Typesafe thread specific storage. If threading is enabled, this class uses a mutex protected map. If
       /// threading is not enabled, the class always returns the same data, regardless of which thread it is called from.
       /// 
@@ -74,6 +73,52 @@ namespace chaiscript
         class Thread_Storage
         {
           public:
+
+            Thread_Storage(void *t_key)
+              : m_key(t_key)
+            {
+            }
+
+            ~Thread_Storage()
+            {
+              t().erase(m_key);
+            }
+
+            inline T *operator->() const
+            {
+              return &(t()[m_key]);
+            }
+
+            inline T &operator*() const
+            {
+              return t()[m_key];
+            }
+
+            void *m_key;
+
+          private:
+            static std::unordered_map<void*, T> &t()
+            {
+              thread_local static std::unordered_map<void *, T> my_t;
+              return my_t;
+            }
+        };
+
+#else
+
+      /// Typesafe thread specific storage. If threading is enabled, this class uses a mutex protected map. If
+      /// threading is not enabled, the class always returns the same data, regardless of which thread it is called from.
+      /// 
+      /// This version is used if the compiler does not support thread_local
+      template<typename T>
+        class Thread_Storage
+        {
+          public:
+
+            Thread_Storage(void *)
+            {
+            }
+
             inline T *operator->() const
             {
               return get_tls().get();
@@ -88,35 +133,26 @@ namespace chaiscript
           private:
             std::shared_ptr<T> get_tls() const
             {
-              
               unique_lock<mutex> lock(m_mutex);
-             
+
               auto itr = m_instances.find(std::this_thread::get_id());
 
               if (itr != m_instances.end()) { return itr->second; }
-             
+
               std::shared_ptr<T> new_instance(new T());
 
               m_instances.insert(std::make_pair(std::this_thread::get_id(), new_instance));
 
               return new_instance;
-              
-
-              /*
-              static __thread std::shared_ptr<T> *m_data = 0;
-              
-              if (!m_data) { m_data = new std::shared_ptr<T>(new T()); }
-
-              return *m_data;
-              */
             }
 
 
             mutable mutex m_mutex;
             mutable std::unordered_map<std::thread::id, std::shared_ptr<T> > m_instances;
         };
+#endif // threading enabled but no tls
 
-#else
+#else // threading disabled
       template<typename T>
       class unique_lock 
       {