lkml.org 
[lkml]   [2014]   [Jul]   [18]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 1/2] workqueue: remove unneeded test before wake up next worker
On Wed, Jul 16, 2014 at 06:09:58PM +0800, Lai Jiangshan wrote:
> In this code:
> if ((worker->flags & WORKER_UNBOUND) && need_more_worker(pool))
> wake_up_worker(pool);
>
> the first test is unneeded. Even the first test is removed, it doesn't affect
> the wake-up logic when WORKER_UNBOUND. And it will not introduce any useless
> wake-up when !WORKER_UNBOUND since the nr_running >= 1 except only one case.
> It will introduce useless/redundant wake-up when cpu_intensive, but this
> case is rare and next patch will also remove this redundant wake-up.
>
> Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
> ---
> kernel/workqueue.c | 7 ++-----
> 1 files changed, 2 insertions(+), 5 deletions(-)
>
> diff --git a/kernel/workqueue.c b/kernel/workqueue.c
> index f8d54c1..6d11b9a 100644
> --- a/kernel/workqueue.c
> +++ b/kernel/workqueue.c
> @@ -2047,11 +2047,8 @@ __acquires(&pool->lock)
> if (unlikely(cpu_intensive))
> worker_set_flags(worker, WORKER_CPU_INTENSIVE, true);
>
> - /*
> - * Unbound pool isn't concurrency managed and work items should be
> - * executed ASAP. Wake up another worker if necessary.
> - */
> - if ((worker->flags & WORKER_UNBOUND) && need_more_worker(pool))
> + /* Wake up another worker if necessary. */
> + if (need_more_worker(pool))
> wake_up_worker(pool);

What does this buy us? Sure, it may achieve about the same operation
but it's a lot more confusing. need_more_worker() is specifically for
concurrency management. Applying it to unmanaged workers could lead
to okay behavior but conflating the two to save one test on worker
flags doesn't seem like a good trade-off to me.

Thanks.

--
tejun


\
 
 \ /
  Last update: 2014-07-19 00:41    [W:0.137 / U:0.104 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site