Learning about concurrency and order
As the reader was reminded earlier in this chapter, any program that accesses any shared data without access synchronization (mutexes or atomic accesses, usually) has undefined behavior that is usually called a data race. This seems simple enough, at least in theory. But our motivational example was too simple: it had just one variable shared between threads. There is more to concurrency than locking shared variables, as we are about to see.
The need for order
Now consider this example known as the producer-consumer queue. Let us say that we have two threads. The first thread, the producer, prepares some data by constructing objects. The second thread, the consumer, processes the data (does work on each object). For simplicity, let us say that we have a large memory buffer that is initially uninitialized and the producer thread constructs new objects in the buffer as if they were array elements:
size_t N; /...
