The current algorithm requires threads to commit/cancel waiting in order
they called Prewait. Spinning caused by that serialization can consume
lots of CPU time on some workloads. Restructure the algorithm to not
require that serialization and remove spin waits from Commit/CancelWait.
Note: this reduces max number of threads from 2^16 to 2^14 to leave
more space for ABA counter (which is now 22 bits).
Implementation details are explained in comments.
* Adds a hint to ThreadPool allowing us to turn off spin waiting. Currently each reader and record yielder op in a graph creates a threadpool with a thread that spins for 1000 iterations through the work stealing loop before yielding. This is wasteful for such ops that process I/O.
* This also changes the number of iterations through the steal loop to be inversely proportional to the number of threads. Since the time of each iteration is proportional to the number of threads, this yields roughly a constant spin time.
* Implement a separate worker loop for the num_threads == 1 case since there is no point in going through the expensive steal loop. Moreover, since Steal() calls PopBack() on the victim queues it might reverse the order in which ops are executed, compared to the order in which they are scheduled, which is usually counter-productive for the types of I/O workloads the single thread pools tend to be used for.
* Store num_threads in a member variable for simplicity and to avoid a data race between the thread creation loop and worker threads calling threads_.size().
* Use a pseudo-random permutation of queue indices during random stealing. This ensures that all the queues are considered.
* Directly pop from a non-empty queue when we are waiting for work,
instead of first noticing that there is a non-empty queue and
then doing another round of random stealing to re-discover the non-empty
queue.
* Steal only 1 task from a remote queue instead of half of tasks.
