A thread pool is a managed set of worker threads that executes submitted tasks. A fixed-size pool imposes a maximum number of workers, making resource availability part of a parallel program’s behavior. section 7.3.3
The central hazard appears when a task running in the pool submits another task and blocks on its Future. With one worker, the outer task occupies the only thread while waiting for the inner task, which cannot start. section 7.3.3
Trace a suspected deadlock by asking:
- How many workers can the ExecutorService run simultaneously? section 7.3.3
- Does a running task submit another task before it completes? section 7.3.3
- Does the running task wait on the submitted task’s Future? section 7.3.3
- Are all workers blocked while another logical task waits to run? section 7.9
Nested fork with one worker
With Executors.newFixedThreadPool(1), running a computation such as fork(a) can deadlock when fork submits a Callable that evaluates a(es).get. The sole worker runs the outer Callable, waits for the inner Callable, and leaves no worker available for that inner computation. section 7.3.3
val a = lazyUnit(42 + 1) val es = Executors.newFixedThreadPool(1) println(Par.equal(es)(a, fork(a)))
Run the equality check with a one-worker pool to observe the interaction between fork, ExecutorService capacity, and blocking evaluation. section 7.3.3
