Lets say you need a singleton in a multi-threaded application. The naive approach acquires a lock every time you access the instance. With thousands of threads hitting getInstance() per second, that lock becomes a bottleneck. You're paying synchronization costs on every call, even though the object only needs to be created once.

Double-checked locking solves this with an elegant optimization. Check if the instance exists without locking (fast path). Only acquire the lock if the instance is null, then check again inside the lock before creating. After the first creation, all subsequent calls take the fast path, avoiding synchronization entirely.

In this chapter, we'll explore why double-checked locking is trickier than it looks, understand the memory visibility issues that break naive implementations, and learn how to implement it correctly across languages.

What is Double-Checked Locking?