C in a Nutshell, 2nd Edition

Chapter 14. Multithreading

C programs often perform several tasks simultaneously. For example, a program may:

• Execute procedures that accomplish intermediate tasks in parallel and so improve performance

• Process user input while carrying on time-consuming data communication or real-time operations “in the background”

Different tasks are performed simultaneously by the concurrent execution of parts of the program. Especially on modern multiprocessor systems—including multicore processors, of course—it is increasingly important for programs to take advantage of concurrency to use the system’s resources efficiently.

Until recently, C developers have had to depend on features of the operating system or appropriate libraries to implement concurrent execution. Now, however, the new C11 standard makes concurrency in C programming portable. C11 supports multithreaded execution, or multiple parallel paths of control flow within a process, and provides the same degree of concurrency as all modern operating systems. To this end, C11 defines an appropriate memory model and supports atomic operations. Support for multithreading and atomic operations are optional under the C11 standard, however. An implementation that conforms to C11 must simply define the macros __STDC_NO_THREADS__ and __STDC_NO_ATOMICS__ if it does not provide the corresponding features.

You may have already worked with the POSIX threads extension to C (called pthreads for short); that is, the library that implements ...