* Use atomic operations to determine if the size of the queue exceeds the limit.

* Optimizing reflection performance