|
|
Multiprocessing
In FairThreads, users have control on the way threads are scheduled. Fair threads which are
linked to a scheduler are scheduled in a cooperative way by it.
When a fair thread unlinks from a scheduler, it becomes an autonomous
native thread which can be run in real parallelism, on a distinct
processor.
An important point is that FairThreads provides users with programming primitives allowing threads to dynamically link to
schedulers and to dynamically unlink from them.
Precise Semantics
Linked threads have a precise and clear semantics (which can be
formally given). The point is that systems exclusively made of threads
linked to one unique scheduler are completely deterministic.
Simplicity
FairThreads offers a very simple framework for concurrent and parallel
programming. Simple cooperative systems can be coded in a simple way,
without the need of locks to protect data. Instants give automatic
synchronizations that can also simplify programming in certain
situations.
Compatibility with Pthreads
FairThreads is fully compatible with the standard Pthreads library. Indeed,
unlinked fair threads are actually just pthreads. In this respect,
FairThreads is an extension of Pthreads, which allows users to define
cooperative contexts with a clear and simple semantics, in which
threads execute at the same pace and events are instantaneously
broadcast.
Automata
Auxiliary tasks can be implemented using automata instead of standard
fair threads. Implementation of an automaton is lightweight and does
not require a dedicated native thread. Automata are useful for
short-lived small tasks or when a large number of
tasks is needed. Automata are an alternative to standard techniques
such as "chunks" or "chores", sometimes used in thread-based
programming.
Implementation
A first implementation of FairThreads is available (under the
Gnu Lesser General Public License
as a library called fthread [6]
which must be used with the standard Pthreads library.
|