Intel Threading Building Blocks

Task Scheduling Overview

The library provides a task scheduler, which is the engine that drives the algorithm templates. You may also call it directly. This is worth considering if your application meets the criteria described earlier that make the default task scheduler inefficient.

Tasks are logical units of computation. The scheduler maps these onto physical threads. The mapping is non-preemptive. Each thread has an execute() method. Once a thread starts running execute(), the task is bound to that thread until execute() returns. During that time, the thread services other tasks only when it waits on child tasks, at which time it may run the child tasks or—if there are no pending child tasks—service tasks created by other threads.

The task scheduler is intended for parallelizing computationally intensive work. Since task objects are not scheduled preemptively, they should not make calls that might block for long periods because, meanwhile, the blocked thread (and its associated processor) are precluded from servicing other tasks.

Warning

There is no guarantee that potentially parallel tasks actually execute in parallel, because the scheduler adjusts actual parallelism to fit available worker threads. For example, when given a single worker thread, the scheduler obviously cannot create parallelism. Furthermore, it is unsafe to use tasks in a producer/consumer relationship if the consumer needs to do some initialization or other work before the producer task completes, because there ...