Fiber (computer science)

In computer science, a fiber is a particularly lightweight thread of execution.

Like threads, fibers share address space. However, fibers use cooperative multitasking while threads use preemptive multitasking. Threads often depend on the kernel's thread scheduler to preempt a busy thread and resume another thread; fibers yield themselves to run another fiber while executing.

Fibers and coroutines

Fibers describe essentially the same concept as coroutines. The distinction, if there is any, is that coroutines are a language-level construct, a form of control flow, while fibers are a systems-level construct, viewed as threads that happen to not run in parallel. It is contentious which of the two concepts has priority: fibers may be viewed as an implementation of coroutines,[1] or as a substrate on which to implement coroutines.[2]

Advantages and disadvantages

Because fibers multitask cooperatively, thread safety is less of an issue than with preemptively scheduled threads, and synchronization constructs including spinlocks and atomic operations are unnecessary when writing fibered code, as they are implicitly synchronized. However, many libraries yield a fiber implicitly as a method of conducting non-blocking I/O; as such, some caution and documentation reading is advised. A disadvantage is that fibers cannot utilize multiprocessor machines without also using preemptive threads; however, an M:N threading model with no more preemptive threads than CPU cores can be more efficient than either pure fibers or pure preemptive threading.

In some server programs fibers are used to soft block themselves to allow their single-threaded parent programs to continue working. In this design, fibers are used mostly for I/O access which does not need CPU processing. This allows the main program to continue with what it is doing. Fibers yield control to the single-threaded main program, and when the I/O operation is completed fibers continue where they left off.

Operating system support

Less support from the operating system is needed for fibers than for threads. They can be implemented in modern Unix systems using the library functions getcontext, setcontext and swapcontext in ucontext.h, as in GNU Portable Threads, or in assembler as boost.fiber.

On Microsoft Windows, fibers are created using the ConvertThreadToFiber and CreateFiber calls; a fiber that is currently suspended may be resumed in any thread. Fiber-local storage, analogous to thread-local storage, may be used to create unique copies of variables.[3]

Symbian OS used a similar concept to fibers in its Active Scheduler. An active object contained one fiber to be executed by the Active Scheduler when one of several outstanding asynchronous calls completed. Several Active objects could be waiting to execute (based on priority) and each one had to restrict its own execution time.

Fiber implementation examples

Fibers can be implemented without operating system support, although some operating systems or libraries provide explicit support for them.

See also

References

  1. A Fiber Class
  2. Shankar, Ajai (September 2003). "Implementing Coroutines for .NET by Wrapping the Unmanaged Fiber API". MSDN Magazine. Microsoft. Archived from the original on 19 August 2003. Retrieved 24 December 2006. This article shows how coroutines can be implemented for the .NET Framework by using the Fiber API and Managed Extensions for C++, and how they can be easily used with other .NET-compliant languages
  3. Fibers, MSDN Library
  4. CreateFiber, MSDN
This article is issued from Wikipedia. The text is licensed under Creative Commons - Attribution - Sharealike. Additional terms may apply for the media files.