From: "Björn Steinbrink" <B.Steinbrink@gmx.de>
To: Steven Rostedt <rostedt@goodmis.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>,
LKML <linux-kernel@vger.kernel.org>,
Christoph Lameter <clameter@sgi.com>, Ingo Molnar <mingo@elte.hu>,
Peter Zijlstra <a.p.zijlstra@chello.nl>,
Christoph Hellwig <hch@infradead.org>
Subject: Re: Major regression on hackbench with SLUB
Date: Mon, 10 Dec 2007 08:38:08 +0100 [thread overview]
Message-ID: <20071210073808.GA32112@atjola.homenet> (raw)
In-Reply-To: <Pine.LNX.4.58.0712081702450.23837@gandalf.stny.rr.com>
On 2007.12.08 17:16:24 -0500, Steven Rostedt wrote:
>
> Hi Linus,
>
> > On Fri, 7 Dec 2007, Linus Torvalds wrote:
> > >
> > > Can you do one run with oprofile, and see exactly where the cost is? It
> > > should hopefully be pretty darn obvious, considering your timing.
>
> The results are here:
>
> http://people.redhat.com/srostedt/slub/results/slab.op
> http://people.redhat.com/srostedt/slub/results/slub.op
Hm, you seem to be hitting the "another_slab" stuff in __slab_alloc
alot. I wonder if !node_match triggers too often. We always start with
the per cpu slab, if that one is on the wrong node, you'll always hit
that "another_slab" path.
After searching for way too long (given that I have no clue about that
stuff anyway and just read the code out of curiousness), I noticed that
the the cpu_to_node stuff on x86_64 seems to be initialized to 0xff
(arch/x86/mm/numa_64.c), and Google brought me this dmesg output [1],
which, AFAICT, shows that the per cpu slab setup is done _before_
cpu_to_node is correctly setup. That would lead to the per cpu slabs all
having node == 0xff, which looks pretty bad.
Disclaimer: I read the slub/numa/$WHATEVER_I_SAW_THERE for the first
time, so this might be total bull ;-)
Björn
[1] http://linux.derkeiler.com/Mailing-Lists/Kernel/2007-10/msg04648.html
next prev parent reply other threads:[~2007-12-10 7:38 UTC|newest]
Thread overview: 93+ messages / expand[flat|nested] mbox.gz Atom feed top
2007-12-07 17:50 Major regression on hackbench with SLUB Steven Rostedt
2007-12-07 17:55 ` Ingo Molnar
2007-12-07 17:59 ` Linus Torvalds
2007-12-07 18:36 ` Linus Torvalds
2007-12-07 18:58 ` Steven Rostedt
2007-12-08 22:16 ` Steven Rostedt
2007-12-10 7:38 ` Björn Steinbrink [this message]
2007-12-10 14:00 ` Steven Rostedt
2007-12-09 0:28 ` Steven Rostedt
2007-12-13 22:11 ` Christoph Lameter
2007-12-11 14:33 ` Major regression on hackbench with SLUB (more numbers) Ingo Molnar
2007-12-13 22:22 ` Christoph Lameter
2007-12-14 4:24 ` Christoph Lameter
2007-12-14 6:15 ` Steven Rostedt
2007-12-21 12:09 ` Ingo Molnar
2007-12-21 12:26 ` Ingo Molnar
2007-12-21 16:26 ` Christoph Lameter
2007-12-21 16:33 ` Ingo Molnar
2007-12-21 21:56 ` Christoph Lameter
2007-12-21 16:18 ` Christoph Lameter
2007-12-21 16:44 ` Linus Torvalds
2007-12-21 22:11 ` Christoph Lameter
2007-12-21 22:16 ` Peter Zijlstra
2007-12-21 22:17 ` Peter Zijlstra
2007-12-21 22:27 ` Christoph Lameter
2007-12-21 22:54 ` Ingo Molnar
2007-12-21 23:20 ` David Miller
2007-12-22 9:45 ` Ingo Molnar
2007-12-24 19:24 ` Christoph Lameter
2007-12-21 23:27 ` Andrew Morton
2007-12-21 23:51 ` Christoph Lameter
2007-12-22 0:28 ` Andi Kleen
2007-12-22 0:33 ` Chuck Ebbert
2007-12-22 2:19 ` Matt Mackall
2007-12-22 2:44 ` Andi Kleen
2007-12-22 10:03 ` Ingo Molnar
2007-12-22 12:37 ` Andi Kleen
2007-12-22 12:51 ` Pekka Enberg
2007-12-22 12:54 ` Andi Kleen
2007-12-22 13:27 ` Pekka Enberg
2007-12-22 13:01 ` Alexey Dobriyan
2007-12-22 18:01 ` Linus Torvalds
2007-12-22 19:09 ` Ingo Molnar
2007-12-22 19:29 ` Ingo Molnar
2007-12-22 22:52 ` Jason L Tibbitts III
2007-12-24 3:59 ` Alessandro Suardi
2007-12-22 19:25 ` Theodore Tso
2007-12-22 21:00 ` Linus Torvalds
2007-12-22 21:50 ` Al Viro
2007-12-22 23:29 ` Al Viro
2007-12-22 22:10 ` Willy Tarreau
2007-12-23 0:59 ` Steven Rostedt
2007-12-23 5:12 ` Willy Tarreau
2007-12-23 14:15 ` Andi Kleen
2007-12-24 3:45 ` Theodore Tso
2007-12-24 19:21 ` Christoph Lameter
2007-12-24 23:37 ` Theodore Tso
2007-12-25 3:34 ` Andi Kleen
2008-01-01 12:47 ` Theodore Tso
2008-01-01 15:19 ` Andi Kleen
2008-01-01 16:23 ` [patch] slub: provide /proc/slabinfo Ingo Molnar
2008-01-05 0:27 ` Arjan van de Ven
2008-01-05 0:49 ` Linus Torvalds
2007-12-26 21:31 ` Major regression on hackbench with SLUB (more numbers) Christoph Lameter
2007-12-26 22:16 ` Al Viro
2007-12-27 5:51 ` Christoph Lameter
2007-12-27 20:28 ` SLUB sysfs support Christoph Lameter
2007-12-27 22:59 ` Al Viro
2007-12-27 23:22 ` Christoph Lameter
2007-12-27 23:53 ` Christoph Lameter
2007-12-27 23:58 ` Al Viro
2007-12-28 0:02 ` Christoph Lameter
2007-12-28 0:45 ` Al Viro
2007-12-28 2:19 ` Christoph Lameter
2007-12-28 3:26 ` Al Viro
2008-01-02 20:25 ` Christoph Lameter
2007-12-27 23:54 ` Al Viro
2007-12-28 9:00 ` Major regression on hackbench with SLUB (more numbers) Ingo Molnar
2007-12-28 14:52 ` Steven Rostedt
2007-12-22 19:46 ` slabtop replacement was " Andi Kleen
2007-12-22 23:28 ` Karol Swietlicki
2007-12-29 18:08 ` Karol Swietlicki
2007-12-21 22:49 ` Linus Torvalds
2007-12-21 17:44 ` Pekka Enberg
2007-12-21 22:22 ` Christoph Lameter
2007-12-21 22:37 ` Christoph Lameter
2007-12-21 22:51 ` Linus Torvalds
2007-12-21 23:17 ` Ingo Molnar
2007-12-21 23:40 ` Pekka Enberg
2007-12-21 23:54 ` Christoph Lameter
2007-12-22 0:02 ` Chuck Ebbert
2007-12-21 22:52 ` Steven Rostedt
2007-12-21 22:56 ` Ingo Molnar
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=20071210073808.GA32112@atjola.homenet \
--to=b.steinbrink@gmx.de \
--cc=a.p.zijlstra@chello.nl \
--cc=clameter@sgi.com \
--cc=hch@infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@elte.hu \
--cc=rostedt@goodmis.org \
--cc=torvalds@linux-foundation.org \
/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;
as well as URLs for NNTP newsgroup(s).