Building blocks for concurrent programming

The development of concurrent programs is, generally, quite difficult. Several factors can make it even more difficult: for example, it is much harder to write concurrent programs that also need to be correct and efficient (in other words, all of them). Complex programs with many mutexes, or lock-free programs, are harder still.

As was said at the conclusion of the last section, the only hope of managing this complexity is to corral it into small, well-defined sections of the code, or modules. As long as the interfaces and requirements are clear, the clients of these modules don't need to know whether the implementation is lock-free or lock-based. It does affect the performance, so the module may be too slow for a particular need until it's optimized, but we do these optimizations as needed, and they are confined to the particular module.

In this chapter, we focus on the modules that implement data structures for concurrent...