Medium
Work-stealing deque with owner and thieves on opposite ends
In a work-stealing task scheduler, each worker thread owns a deque of tasks. The owner exclusively pushes and pops from one end ("bottom"); other worker threads that have run out of local work ("thieves") steal from the opposite end ("top"). Why is splitting the deque this way, rather than having every thread operate on a single shared end, the key design decision for low contention?
25 of 51 in Concurrency