Multithreading

In computer science, single-threading is similar to traditional sequential processing, executing a single command at any given time. On the other hand, multithreading implements more than one thread to exist and execute in a single process, simultaneously. By allowing multiple threads to access shared resources/contexts and be executed independently, this programming technique can help applications to gain speed in the execution of independent tasks.

Multithreading can primarily be achieved in two ways. In single-processor systems, multithreading is typically implemented via time slicing, a technique that allows the CPU to switch between different software running on different threads. In time slicing, the CPU switches its execution so quickly and so often that users usually perceive that the software is running in parallel (for example, when you open two different software at the same time on a single-processor computer):

As opposed to single-processor systems, systems with multiple processors or cores can easily implement multithreading, by executing each thread in a separate process or core, simultaneously. Additionally, time slicing is an option, as these multiprocess or multicore systems can have only one processor/core to switch between tasks—although this is generally not a good practice.

Multithreaded applications have a number of advantages, as compared to traditional sequential applications; some of them are listed as follows:

• Faster execution time: One of the main advantages of concurrency through multithreading is the speedup that is achieved. Separate threads in the same program can be executed concurrently or in parallel, if they are sufficiently independent of one another.

• Responsiveness: A single-threaded program can only process one piece of input at a time; therefore, if the main execution thread blocks on a long-running task (that is, a piece of input that requires heavy computation and processing), the whole program will not be able to continue with other input, and hence, it will appear to be frozen. By using separate threads to perform computation and remain running to take in different user input simultaneously, a multithreaded program can provide better responsiveness.
• Efficiency in resource consumption: As we mentioned previously, multiple threads within the same process can share and access the same resources. Consequently, multithreaded programs can serve and process many client requests for data concurrently, using significantly fewer resources than would be needed when using single-threaded or multiprocess programs. This also leads to quicker communication between threads.

That being said, multithreaded programs also have their disadvantages, as follows:

• Crashes: Even though a process can contain multiple threads, a single illegal operation within one thread can negatively affect the processing of all of the other threads in the process, and can crash the entire program as a result.
• Synchronization: Even though sharing the same resources can be an advantage over traditional sequential programming or multiprocessing programs, careful consideration is also needed for the shared resources. Usually, threads must be coordinated in a deliberate and systematic manner, so that shared data is computed and manipulated correctly. Unintuitive problems that can be caused by careless thread coordination include deadlocks, livelocks, and race conditions, all of which will be discussed in future chapters.