Chapter 11: Concurrency

Abstract

CPU manufacturers have recently struggled to improve CPU frequencies at the same rate of progress as was possible in the 1990s. CPU performance improvements over time have been maintained by clever CPU design and multiple processors being included on a single chip. This means programmers today must embrace concurrent programming or multithreaded programming if they wish their programs to perform quickly on modern computer chips. Concurrent programming can be a challenge for programmers to get right. Many pitfalls await concurrent programs, including data that gets out of sync and therefore is wrong as well as deadlocks once your tasks require the use of locks to manage access. The recipes in this chapter introduce you to some practical applications of the STL features supplied by C++ to help you write concurrent programs. 11-1. Using Threads to Execute Concurrent Tasks Problem You're writing a program that is performing poorly, and you'd like to speed up execution by using multiple processors in a system. Solution C++ provides the thread type, which can be used to create a native operating system thread. Program threads can be run on more than a single processor and therefore allow you to write programs that can use multiple CPUs and CPU cores. How It Works Detecting the Number of Logical CPU Cores The C++ thread library provides a feature set that lets programs use all the cores and CPUs available in a given computer system. The first important function supplied by the C++ threading capabilities that you should be aware of allows you to query the number of execution units the computer contains. Listing 11-1 shows the C++ thread::hardware_concurrency method.