17.2 Threads

Bigloo supports fair threads (see Section Thread), a specification of cooperative threads. In this framework a thread must explicitly or implicitly yield the processor to the scheduler (see Section Scheduler). Explicit cooperation is achieved by library functions such as thread-yield! or thread-sleep!. The scheduler does not preempt a running thread to allocate the processor to another waiting thread. Fair threads have two drawbacks over preemptive threads:

• Cooperative threads are not skilled to benefit of multi processors platforms.
• Single threads programs must be adapted in order to be ran concurrently.

On the other hand, Fair threads have advantages that make them suitable for a high level programming language such as Scheme:

• Fair threads have a strong and well defined semantic. Multi threaded programs using Fair threads are deterministic thus programs that deploy Fair threads are predictable.
• Fair threads are easier to program with because they hide most the of the concurrent programming pitfalls. In particular, since Fair threads enforce a strong synchronization, there is no need to deploy techniques such as mutex, semaphore or condition variables.

This whole chapter has been written in collaboration with Frédéric Boussinot. It uses materials on Fair threads that can be found at http://www-sop.inria.fr/indes/rp/FairThreads/html/FairThreads.html.

This document was generated on March 31, 2014 using texi2html 5.0.