concurrency

Concurrency is the ability of a system to manage multiple tasks that make progress over overlapping time periods. It enables a program to handle several sequences of operations at once, either by interleaving their execution on a single processor or by running them in parallel on multiple cores. Concurrency focuses on structure and coordination, while parallelism emphasizes actual simultaneous execution.

A typical concurrent system involves tasks, processes, or threads that share resources or communicate with each other. Models of concurrency include shared-memory concurrency, where tasks access a common address space, and message-passing concurrency, where tasks communicate through explicit messages. The actor model is a related approach in which independent entities (actors) exchange messages and make progress independently.

Synchronization and communication are central to correctness in concurrent programs. Tools such as mutexes, locks, semaphores, monitors, and condition variables help enforce mutual exclusion and ordering. Barriers coordinate phases of work. Proper memory models ensure visibility and ordering of updates across threads. Common problems include race conditions, deadlocks, livelocks, and starvation, as well as subtle memory visibility issues.

Programs use various techniques to manage concurrency, including preemptive or cooperative scheduling, thread pools, and asynchronous I/O with futures, promises, or async/await. Design patterns include shared-memory synchronization with locks, message-passing with channels, and the actor model, as well as data parallelism and pipelining.

Many languages provide built-in support for concurrency, such as threads in Java and C++, goroutines in Go, the event loop in Node.js, or processes in Erlang. Modern hardware with multi-core CPUs enables true parallelism, though concurrency can be exploited on a single core through interleaved execution.