Atomic: cleanup and add more operations

Cleanup:
	* Remove volatile keyword, it is of no use at all. [1]
	* Remove the enable_if stuff. It had no use either.
	  The most likely explanation why the enable_if stuff was there is that it
	  was used as something like a STATIC_ASSERT to verify that sizeof(T) is not larger
	  than sizeof(void *). This check however is not just misplaced in a place where we
	  already use a lock, it isn't needed at all, as gcc will just generate a call to
	  to the runtime if it compiles for platforms that don't support atomic instructions.
	  The runtime will then most likely use locks.

Code style fixes:
	* Prefix name of the mutex
	* Line everything up nicely, where it makes things look nice
	* Filling \ continuations with spaces is code style rule

Added operations on the atomic var:
	* Compare and swap
	* Swap

The second point of the cleanup also fixes the Android build of the next commit.

[1]: http://stackoverflow.com/q/2484980

src/threading/atomic.h

@@ -29,52 +29,72 @@ with this program; if not, write to the Free Software Foundation, Inc.,
 #define GCC_VERSION   (__GNUC__        * 100 + __GNUC_MINOR__)
 #define CLANG_VERSION (__clang_major__ * 100 + __clang_minor__)
 #if GCC_VERSION >= 407 || CLANG_VERSION >= 302
-	#define ATOMIC_LOAD(T, v)      return __atomic_load_n(&(v),       __ATOMIC_SEQ_CST)
+	#define ATOMIC_LOAD(T, v)        return __atomic_load_n    (&(v),       __ATOMIC_SEQ_CST)
 	#define ATOMIC_STORE(T, v, x)           __atomic_store     (&(v), &(x), __ATOMIC_SEQ_CST); return x
-	#define ATOMIC_ADD_EQ(T, v, x) return __atomic_add_fetch(&(v), (x), __ATOMIC_SEQ_CST)
-	#define ATOMIC_SUB_EQ(T, v, x) return __atomic_sub_fetch(&(v), (x), __ATOMIC_SEQ_CST)
-	#define ATOMIC_POST_INC(T, v)  return __atomic_fetch_add(&(v), 1, __ATOMIC_SEQ_CST)
-	#define ATOMIC_POST_DEC(T, v)  return __atomic_fetch_sub(&(v), 1, __ATOMIC_SEQ_CST)
+	#define ATOMIC_EXCHANGE(T, v, x) return __atomic_exchange_n(&(v), (x),  __ATOMIC_SEQ_CST)
+	#define ATOMIC_ADD_EQ(T, v, x)   return __atomic_add_fetch (&(v), (x),  __ATOMIC_SEQ_CST)
+	#define ATOMIC_SUB_EQ(T, v, x)   return __atomic_sub_fetch (&(v), (x),  __ATOMIC_SEQ_CST)
+	#define ATOMIC_POST_INC(T, v)    return __atomic_fetch_add (&(v), 1,    __ATOMIC_SEQ_CST)
+	#define ATOMIC_POST_DEC(T, v)    return __atomic_fetch_sub (&(v), 1,    __ATOMIC_SEQ_CST)
+	#define ATOMIC_CAS(T, v, e, d)   return __atomic_compare_exchange_n(&(v), &(e), (d), \
+		false, __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST)
 #else
 	#define ATOMIC_USE_LOCK
 	#include "threading/mutex.h"
 
 	#define ATOMIC_LOCK_OP(T, op) do { \
-			mutex.lock(); \
+			m_mutex.lock();            \
 			T _val = (op);             \
-			mutex.unlock(); \
+			m_mutex.unlock();          \
 			return _val;               \
 		} while (0)
+	#define ATOMIC_LOCK_CAS(T, v, e, d) do { \
+			m_mutex.lock();                  \
+			bool _eq = (v == e);             \
+			if (_eq)                         \
+				v = d;                       \
+			m_mutex.unlock();                \
+			return _eq;                      \
+		} while (0)
 	#define ATOMIC_LOAD(T, v)                     \
 		if (sizeof(T) <= sizeof(void*)) return v; \
 		else ATOMIC_LOCK_OP(T, v);
 	#define ATOMIC_STORE(T, v, x)                     \
 		if (sizeof(T) <= sizeof(void*)) return v = x; \
 		else ATOMIC_LOCK_OP(T, v = x);
-# if GCC_VERSION >= 401
+	#define ATOMIC_EXCHANGE(T, v, x) do { \
+			m_mutex.lock();               \
+			T _val = v;                   \
+			v = x;                        \
+			m_mutex.unlock();             \
+			return _val;                  \
+		} while (0)
+	#if GCC_VERSION >= 401
 		#define ATOMIC_ADD_EQ(T, v, x) return __sync_add_and_fetch(&(v), (x))
 		#define ATOMIC_SUB_EQ(T, v, x) return __sync_sub_and_fetch(&(v), (x))
 		#define ATOMIC_POST_INC(T, v)  return __sync_fetch_and_add(&(v), 1)
 		#define ATOMIC_POST_DEC(T, v)  return __sync_fetch_and_sub(&(v), 1)
-# else
+		#define ATOMIC_CAS(T, v, e, d) return __sync_bool_compare_and_swap(&(v), &(e), (d))
+	#else
 		#define ATOMIC_ADD_EQ(T, v, x) ATOMIC_LOCK_OP(T, v += x)
 		#define ATOMIC_SUB_EQ(T, v, x) ATOMIC_LOCK_OP(T, v -= x)
 		#define ATOMIC_POST_INC(T, v)  ATOMIC_LOCK_OP(T, v++)
 		#define ATOMIC_POST_DEC(T, v)  ATOMIC_LOCK_OP(T, v--)
-# endif
+		#define ATOMIC_CAS(T, v, e, d) ATOMIC_LOCK_CAS(T, v, e, d)
+	#endif
 #endif
 
 
 template<typename T>
-class Atomic
-{
-	// Like C++11 std::enable_if, but defaults to char since C++03 doesn't support SFINAE
-	template<bool B, typename T_ = void> struct enable_if { typedef char type; };
-	template<typename T_> struct enable_if<true, T_> { typedef T_ type; };
+class Atomic {
 public:
-	Atomic(const T &v=0) : val(v) {}
+	Atomic(const T &v = 0) : val(v) {}
 
 	operator T () { ATOMIC_LOAD(T, val); }
+
+	T exchange(T x)  { ATOMIC_EXCHANGE(T, val, x); }
+	bool compare_exchange_strong(T &expected, T desired) { ATOMIC_CAS(T, val, expected, desired); }
+
 	T operator =  (T x) { ATOMIC_STORE(T, val, x); }
 	T operator += (T x) { ATOMIC_ADD_EQ(T, val, x); }
 	T operator -= (T x) { ATOMIC_SUB_EQ(T, val, x); }
@@ -82,11 +102,10 @@ public:
 	T operator -- ()    { return *this -= 1; }
 	T operator ++ (int) { ATOMIC_POST_INC(T, val); }
 	T operator -- (int) { ATOMIC_POST_DEC(T, val); }
-
 private:
-	volatile T val;
+	T val;
 #ifdef ATOMIC_USE_LOCK
-	typename enable_if<sizeof(T) <= sizeof(void*), Mutex>::type mutex;
+	Mutex m_mutex;
 #endif
 };