Fix: don't wait after completion of a work queue job batch
authorJérémie Galarneau <jeremie.galarneau@efficios.com>
Fri, 7 Dec 2018 22:06:38 +0000 (17:06 -0500)
committerMathieu Desnoyers <mathieu.desnoyers@efficios.com>
Sun, 9 Dec 2018 11:47:12 +0000 (06:47 -0500)
As indicated in the previous patch of this series, waiting on
completion of a job batch from the work queue artificially increases
the latency of the work queue.

The previous patch removed the wait that is performed when the
work queue is observed to be empty and was observed as the cause of a
performance problem.

It is likely that waiting when the queue is observed to be non-empty
is similarly unintended. Note that I have not observed such a problem
myself.

If a workqueue user even need the explicit delay for batching (e.g. if
a call-rcu implementation would ever use the workqueue worker thread),
it can add it within the worker_before_wait_fct callback received as
argument from workqueue creation.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
src/workqueue.c

index 0b1a9ea..39d0e07 100644 (file)
@@ -244,8 +244,6 @@ static void *workqueue_thread(void *arg)
                                 * urcu_work list.
                                 */
                                cmm_smp_mb();
-                       } else {
-                               (void) poll(NULL, 0, 10);
                        }
                } else {
                        (void) poll(NULL, 0, 10);
This page took 0.057243 seconds and 4 git commands to generate.