lkml.org 
[lkml]   [2013]   [Dec]   [10]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH v3 09/12] sched/numa: fix task scan rate adjustment
Hi Wanpeng,

On Sun, Dec 08, 2013 at 02:14:50PM +0800, Wanpeng Li wrote:
> commit 04bb2f947 (sched/numa: Adjust scan rate in task_numa_placement) calculate
> period_slot which should be used as base value of scan rate increase if remote
> access dominate. However, current codes forget to use it, this patch fix it.
>
> Signed-off-by: Wanpeng Li <liwanp@linux.vnet.ibm.com>
> ---
> kernel/sched/fair.c | 2 +-
> 1 files changed, 1 insertions(+), 1 deletions(-)
>
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index 7073c76..b077f1b3 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -1358,7 +1358,7 @@ static void update_task_scan_period(struct task_struct *p,
> */
> period_slot = DIV_ROUND_UP(diff, NUMA_PERIOD_SLOTS);
> ratio = DIV_ROUND_UP(private * NUMA_PERIOD_SLOTS, (private + shared));
> - diff = (diff * ratio) / NUMA_PERIOD_SLOTS;
> + diff = (period_slot * ratio) / NUMA_PERIOD_SLOTS;
> }
>
> p->numa_scan_period = clamp(p->numa_scan_period + diff,

It seems to me that the original code is correct, because the mathematical
meaning of this hunk is clear:

diff = (diff calculated by local-remote ratio) * (private-shared ratio)

If you use period_slot here, diff always becomes less then 1/10 finally by
the second ratio multiplication (because we divide by NUMA_PERIOD_SLOTS twice),
and I don't see the justification.

And if my idea is correct, we don't have to recalculate period_slot when
we multiply private-shared ratio. So we can remove that line.

Thanks,
Naoya Horiguchi


\
 
 \ /
  Last update: 2013-12-10 08:01    [W:0.219 / U:0.912 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site