ForkJoinPool & Work Stealing

ForkJoinPool is a specialized thread pool for divide-and-conquer parallelism where tasks recursively fork into subtasks, process in parallel with work-stealing for load balancing, and join results back together.

ForkJoinPool.commonPool() is the default executor for both CompletableFuture async methods and parallelStream; CF uses it as a regular pool while parallelStream uses true fork/join splitting, and both sharing the pool creates isolation risks.

ForkJoinPool uses work-stealing where each worker pops tasks from the front of its own deque (LIFO for cache locality) and idle workers steal from the back of busy workers' deques (FIFO for largest subtasks), but blocking IO on the shared common pool starves all parallel streams.

Fork/join splits a big CPU-bound job into subtasks with fork, compute and join, using a sequential threshold so tiny pieces are solved directly and splitting overhead does not swamp the work.

Arrays.parallelSort, parallelPrefix and parallelSetAll offload large array work to the common ForkJoinPool, but they only beat the sequential versions when the array is large and the machine has multiple cores.