-
Howard Chu authored
Remove poolq_hash, it wasn't distributing work evenly to the queues. Just walk through all queues and use the one with smallest active+pending count. Since pool_retract also relied on the hash, a different means of locating the thread to retract was needed. Add pool_submit2 which returns the threadpool task structure, and record which poolq this task lives on.
e12ca8b6