tagged pointer for lockfree algorithm(64bit assumed)

tagged_pointer.cpp

template<class T>
class TaggedPointer
{
public:
    explicit TaggedPointer(T* ptr = nullptr)
        :ptr_(ptr),
        counter_(0)
    {}

    inline T* load_ptr() noexcept
    {
        return ptr_.load(memory_order_acquire);
    }

    inline uint64_t load_counter() noexcept
    {
        return counter_.load(memory_order_acquire);
    }

    inline bool compare_exchange(T* expected_ptr, uint64_t expected_counter, T* desired_ptr, uint64_t desired_counter) noexcept
    {
        bool result;
        asm volatile (
                "lock cmpxchg16b %0;"
                "setz %3;"
                :"+m"(*this), "+a"(expected_ptr), "+d"(expected_counter), "=q"(result)
                :"b"(desired_ptr), "c"(desired_counter)
                :"cc", "memory"
                );
        return result;
    }
private:
    // atomic in case of compiler keep in registers?
    atomic<T*> ptr_;
    atomic<uint64_t> counter_;
} __attribute__ (( __aligned__(16) ));