Is the current worker involved in job theft?

Question

In ForkJoinPool ForkJoinTask , does the current workflow work when a job is stolen?

I realized that the fork connection pool could work while stealing from blocked or waiting threads. The current worker seems an obvious candidate. As soon as the worker calls .join() in another task, this task is essentially blocked.

On the other hand, I see many articles that imply different conclusions. For example, there is a general consensus that the current workflow should work before waiting for forked tasks.

There are several articles that discuss the use of ForkJoinTask.getSurplusQueuedTaskCount as a method of balancing queued work if the current worker is doing some of the work. If the current worker also steals, then this does not seem necessary.

Naturally, I would like to maximize thread operations and maximize all workflows. Understanding that the current thread is also stealing work (for example, when calling .join ) will help clarify.

Answer 1

Responsibility for thread management rests with ForkJoinPool. Client code should serve its tasks, and not mix stream processing. Note that tasks and threads are two different things; tasks are units of work performed, and threads are executed.

ForkJoinTask.compute () should convert fork () into smaller subtasks if the task is large enough to use the running parts of the task in parallel and simply process the task if the task is small enough to be run on a single thread. If the job is more than expected, it can fork () do some of the work and do the rest.

If ForkJoinTask.compute () translates into smaller subtasks, it can call join () before returning. ForkJoinPool will then either free the thread to work on other tasks, or create a temporary thread to work on other tasks to make full use of the available parallelism.

I consider it reasonable to assume that the corresponding number of worker threads remains busy as long as there are unfinished tasks, unless you explicitly block the thread in the compute () method.

The Sun tutorial provides more details on how to use these classes:

https://docs.oracle.com/javase/tutorial/essential/concurrency/forkjoin.html