Synchronization patterns

Synchronization patterns have one overarching purpose: to ensure correct operations on data shared by multiple threads. These patterns are critically important for the absolute majority of concurrent programs. The only programs that do not have any need for synchronization are the ones that execute several entirely independent tasks that do not involve any common data (except for, possibly, reading shared and immutable inputs) and produce separate results. For every other program, there is a need to manage some shared state, which exposes us to the danger of the dreaded data races. Formally, the C++ standard says that concurrent access to the same object (same memory location) without the appropriate synchronization that guarantees exclusive access for each thread results in undefined behavior. To be precise, the behavior is undefined if at least one thread can modify the shared data: if the data is never changed by any thread, then there is no possibility...