Essentially, to the operating system (OS), a process consists of one or more threads, each thread processing its own state and variables. One would regard this as a hierarchical configuration, with the OS as the foundation, providing support for the running of (user) processes. Each of these processes then consists of one or more threads. Communication between processes is handled by inter-process communication (IPC), which is provided by the operating system.

In a graphical view, this looks like the following:

Each process within the OS has its own state, with each thread in a process having its own state as well as the relative to the other threads within that same process. While IPC allows processes to communicate with each other, threads can communicate with other threads within the process in a variety of ways, which we'll explore in more...

As we saw in the preceding sections, the stack together with the CPU registers define a task. As mentioned earlier, this stack consists of stack frames, each of which defines the (local) variables, parameters, data, and instructions for that particular instance of task execution. Of note is that although the stack and stack frames are primarily a software concept, it is an essential feature of any modern OS, with hardware support in many CPU instruction sets. Graphically, it can be be visualized like the following:

The SP (ESP on x86) points to the top of the stack, with another pointer (Extended Base Pointer (EBP) for x86). Each frame contains a reference to the preceding frame (caller return address), as set by the OS.

When using a debugger with one's C++ application, this is basically what one sees when requesting the backtrack--the individual frames of the...

Over the past decades, a lot of different terms related to the way tasks are processed by a computer have been coined and come into common use. Many of these are also used interchangeably, correctly or not. An example of this is multithreading in comparison with multiprocessing.

Here, the latter means running one task per processor in a system with multiple physical processors, while the former means running multiple tasks on a singular processor simultaneously, thus giving the illusion that they are all being executed simultaneously:

Another interesting distinction between multiprocessing and multitasking is that the latter uses time-slices in order to run multiple threads on a single processor core. This is different from multithreading in the sense that in a multitasking system, no tasks will ever run in a concurrent fashion on the same CPU core, though...

A number of task-scheduling algorithms exist, each focusing on a different goal. Some may seek to maximize throughput, others minimize latency, while others may seek to maximize response time. Which scheduler is the optimal choice solely depends on the application the system is being used for.

For desktop systems, the scheduler is generally kept as general-purpose as possible, usually prioritizing foreground applications over background applications in order to give the user the best possible desktop experience.

For embedded systems, especially in real-time, industrial applications would instead seek to guarantee timing. This allows processes to be executed at exactly the right time, which is crucial in, for example, driving machinery, robotics, or chemical processes where a delay of even a few milliseconds could be costly or even fatal.

The scheduler type is also dependent...