From: Nick Piggin <nickpiggin@yahoo.com.au>
To: Marcelo Tosatti <marcelo.tosatti@cyclades.com>
Cc: Nikita Danilov <nikita@clusterfs.com>,
Linux Kernel Mailing List <Linux-Kernel@vger.kernel.org>,
Andrew Morton <AKPM@Osdl.ORG>,
Linux MM Mailing List <linux-mm@kvack.org>
Subject: Re: [PATCH]: 1/4 batch mark_page_accessed()
Date: Sat, 27 Nov 2004 11:37:17 +1100 [thread overview]
Message-ID: <41A7CC3D.9030405@yahoo.com.au> (raw)
In-Reply-To: <20041126185833.GA7740@logos.cnet>
Marcelo Tosatti wrote:
> On Sun, Nov 21, 2004 at 06:44:04PM +0300, Nikita Danilov wrote:
>
>>Batch mark_page_accessed() (a la lru_cache_add() and lru_cache_add_active()):
>>page to be marked accessed is placed into per-cpu pagevec
>>(page_accessed_pvec). When pagevec is filled up, all pages are processed in a
>>batch.
>>
>>This is supposed to decrease contention on zone->lru_lock.
>
>
> Here are the STP 8way results:
>
> 8way:
>
...
> kernbench
>
> Decreases performance significantly (on -j4 more notably), probably due to
> the additional atomic operations as noted by Andrew:
>
> kernel: nikita-b2 kernel: patch-2.6.10-rc2
> Host: stp8-002 Host: stp8-003
>
...
>
> Average Half Load -j 4 Run: Average Half Load -j 4 Run:
> Elapsed Time 274.916 Elapsed Time 245.026
> User Time 833.63 User Time 832.34
> System Time 73.704 System Time 73.41
> Percent CPU 335.8 Percent CPU 373.6
> Context Switches 12984.8 Context Switches 13427.4
> Sleeps 21459.2 Sleeps 21642
Do you think looks like it may be a CPU scheduling or disk/fs artifact?
Neither user nor system time are significantly worse, while the vanilla
kernel is using a lot more of the CPUs' power (ie waiting for IO less,
or becoming idle less often due to CPU scheduler balancing).
Aside: under-load conditions like this is actually something where the
CPU scheduler doesn't do brilliantly at currently. I attribute this to
probably most "performance tests" loading it up as much as possible.
I am (on and off) looking at improving performance in these conditions,
and am making some inroads.
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"aart@kvack.org"> aart@kvack.org </a>
next prev parent reply other threads:[~2004-11-27 0:37 UTC|newest]
Thread overview: 15+ messages / expand[flat|nested] mbox.gz Atom feed top
2004-11-21 15:44 Nikita Danilov
2004-11-21 21:12 ` Andrew Morton
2004-11-24 10:40 ` Marcelo Tosatti
2004-11-24 16:32 ` Marcelo Tosatti
2004-11-24 21:53 ` Nikita Danilov
2004-11-26 18:58 ` Marcelo Tosatti
2004-11-27 0:37 ` Nick Piggin [this message]
2004-11-30 16:29 ` Marcelo Tosatti
2004-12-01 1:33 ` Andrew Morton
2004-11-30 22:57 ` Marcelo Tosatti
2004-12-01 12:23 ` Nikita Danilov
2004-12-01 18:58 ` Marcelo Tosatti
2004-12-02 1:59 ` Andrew Morton
2004-11-27 10:41 ` Nikita Danilov
2004-11-27 8:19 ` Marcelo Tosatti
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=41A7CC3D.9030405@yahoo.com.au \
--to=nickpiggin@yahoo.com.au \
--cc=AKPM@Osdl.ORG \
--cc=Linux-Kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=marcelo.tosatti@cyclades.com \
--cc=nikita@clusterfs.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox