From: Linus Torvalds <torvalds@linux-foundation.org>
To: Jesper Dangaard Brouer <brouer@redhat.com>
Cc: Dave Chinner <dchinner@redhat.com>,
Mike Snitzer <snitzer@redhat.com>,
Christoph Lameter <cl@linux.com>,
Pekka Enberg <penberg@kernel.org>,
Andrew Morton <akpm@linux-foundation.org>,
David Rientjes <rientjes@google.com>,
Joonsoo Kim <iamjoonsoo.kim@lge.com>,
"dm-devel@redhat.com" <dm-devel@redhat.com>,
Alasdair G Kergon <agk@redhat.com>, Joe Thornber <ejt@redhat.com>,
Mikulas Patocka <mpatocka@redhat.com>,
Vivek Goyal <vgoyal@redhat.com>,
Sami Tolvanen <samitolvanen@google.com>,
Viresh Kumar <viresh.kumar@linaro.org>,
Heinz Mauelshagen <heinzm@redhat.com>,
linux-mm <linux-mm@kvack.org>,
"netdev@vger.kernel.org" <netdev@vger.kernel.org>
Subject: Re: slab-nomerge (was Re: [git pull] device mapper changes for 4.3)
Date: Mon, 7 Sep 2015 13:22:13 -0700 [thread overview]
Message-ID: <CA+55aFym-dM37xtvKjddMheSV9vPUq=tnN9FoFvEgD0QWW22sg@mail.gmail.com> (raw)
In-Reply-To: <20150907113026.5bb28ca3@redhat.com>
On Mon, Sep 7, 2015 at 2:30 AM, Jesper Dangaard Brouer
<brouer@redhat.com> wrote:
>
> The slub allocator have a faster "fastpath", if your workload is
> fast-reusing within the same per-cpu page-slab, but once the workload
> increases you hit the slowpath, and then slab catches up. Slub looks
> great in micro-benchmarking.
>
> And with "slab_nomerge" I get even high performance:
I think those two are related.
Not merging means that effectively the percpu caches end up being
bigger (simply because there are more of them), and so it captures
more of the fastpath cases.
Obviously the percpu queue size is an easy tunable too, but there are
real downsides to that too. I suspect your IP forwarding case isn't so
different from some of the microbenchmarks, it just has more
outstanding work..
And yes, the slow path (ie not hitting in the percpu cache) of SLUB
could hopefully be optimizable too, although maybe the bulk patches
are the way to go (and unrelated to this thread - at least part of
your bulk patches actually got merged last Friday - they were part of
Andrew's patch-bomb).
Linus
WARNING: multiple messages have this Message-ID (diff)
From: Linus Torvalds <torvalds@linux-foundation.org>
To: Jesper Dangaard Brouer <brouer@redhat.com>
Cc: Dave Chinner <dchinner@redhat.com>,
Mike Snitzer <snitzer@redhat.com>,
Christoph Lameter <cl@linux.com>,
Pekka Enberg <penberg@kernel.org>,
Andrew Morton <akpm@linux-foundation.org>,
David Rientjes <rientjes@google.com>,
Joonsoo Kim <iamjoonsoo.kim@lge.com>,
"dm-devel@redhat.com" <dm-devel@redhat.com>,
Alasdair G Kergon <agk@redhat.com>, Joe Thornber <ejt@redhat.com>,
Mikulas Patocka <mpatocka@redhat.com>,
Vivek Goyal <vgoyal@redhat.com>,
Sami Tolvanen <samitolvanen@google.com>,
Viresh Kumar <viresh.kumar@linaro.org>,
Heinz Mauelshagen <heinzm@redhat.com>,
linux-mm <linux-mm@kvack.org>,
"netdev@vger.kernel.org" <netdev@vger.kernel.org>
Subject: Re: slab-nomerge (was Re: [git pull] device mapper changes for 4.3)
Date: Mon, 7 Sep 2015 13:22:13 -0700 [thread overview]
Message-ID: <CA+55aFym-dM37xtvKjddMheSV9vPUq=tnN9FoFvEgD0QWW22sg@mail.gmail.com> (raw)
In-Reply-To: <20150907113026.5bb28ca3@redhat.com>
On Mon, Sep 7, 2015 at 2:30 AM, Jesper Dangaard Brouer
<brouer@redhat.com> wrote:
>
> The slub allocator have a faster "fastpath", if your workload is
> fast-reusing within the same per-cpu page-slab, but once the workload
> increases you hit the slowpath, and then slab catches up. Slub looks
> great in micro-benchmarking.
>
> And with "slab_nomerge" I get even high performance:
I think those two are related.
Not merging means that effectively the percpu caches end up being
bigger (simply because there are more of them), and so it captures
more of the fastpath cases.
Obviously the percpu queue size is an easy tunable too, but there are
real downsides to that too. I suspect your IP forwarding case isn't so
different from some of the microbenchmarks, it just has more
outstanding work..
And yes, the slow path (ie not hitting in the percpu cache) of SLUB
could hopefully be optimizable too, although maybe the bulk patches
are the way to go (and unrelated to this thread - at least part of
your bulk patches actually got merged last Friday - they were part of
Andrew's patch-bomb).
Linus
--
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:"dont@kvack.org"> email@kvack.org </a>
next prev parent reply other threads:[~2015-09-07 20:22 UTC|newest]
Thread overview: 42+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-09-02 23:13 slab-nomerge (was Re: [git pull] device mapper changes for 4.3) Linus Torvalds
2015-09-03 0:48 ` Andrew Morton
2015-09-03 0:53 ` Mike Snitzer
2015-09-03 0:51 ` Mike Snitzer
2015-09-03 0:51 ` Mike Snitzer
2015-09-03 1:21 ` Linus Torvalds
2015-09-03 2:31 ` Mike Snitzer
2015-09-03 3:10 ` Christoph Lameter
2015-09-03 4:55 ` Andrew Morton
2015-09-03 6:09 ` Pekka Enberg
2015-09-03 8:53 ` Dave Chinner
2015-09-03 3:11 ` Linus Torvalds
2015-09-03 6:02 ` Dave Chinner
2015-09-03 6:13 ` Pekka Enberg
2015-09-03 10:29 ` Jesper Dangaard Brouer
2015-09-03 16:19 ` Christoph Lameter
2015-09-04 9:10 ` Jesper Dangaard Brouer
2015-09-04 14:13 ` Christoph Lameter
2015-09-04 6:35 ` Sergey Senozhatsky
2015-09-04 7:01 ` Linus Torvalds
2015-09-04 7:59 ` Sergey Senozhatsky
2015-09-04 9:56 ` Sergey Senozhatsky
2015-09-04 14:05 ` Christoph Lameter
2015-09-04 14:11 ` Linus Torvalds
2015-09-05 2:09 ` Sergey Senozhatsky
2015-09-05 2:09 ` Sergey Senozhatsky
2015-09-05 20:33 ` Linus Torvalds
2015-09-07 8:44 ` Sergey Senozhatsky
2015-09-08 0:22 ` Sergey Senozhatsky
2015-09-03 15:02 ` Linus Torvalds
2015-09-04 3:26 ` Dave Chinner
2015-09-04 3:51 ` Linus Torvalds
2015-09-05 0:36 ` Dave Chinner
2015-09-05 0:36 ` Dave Chinner
2015-09-07 9:30 ` Jesper Dangaard Brouer
2015-09-07 20:22 ` Linus Torvalds [this message]
2015-09-07 20:22 ` Linus Torvalds
2015-09-07 21:17 ` Jesper Dangaard Brouer
2015-09-04 13:55 ` Christoph Lameter
2015-09-04 22:46 ` Dave Chinner
2015-09-05 0:25 ` Christoph Lameter
2015-09-05 1:16 ` Dave Chinner
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='CA+55aFym-dM37xtvKjddMheSV9vPUq=tnN9FoFvEgD0QWW22sg@mail.gmail.com' \
--to=torvalds@linux-foundation.org \
--cc=agk@redhat.com \
--cc=akpm@linux-foundation.org \
--cc=brouer@redhat.com \
--cc=cl@linux.com \
--cc=dchinner@redhat.com \
--cc=dm-devel@redhat.com \
--cc=ejt@redhat.com \
--cc=heinzm@redhat.com \
--cc=iamjoonsoo.kim@lge.com \
--cc=linux-mm@kvack.org \
--cc=mpatocka@redhat.com \
--cc=netdev@vger.kernel.org \
--cc=penberg@kernel.org \
--cc=rientjes@google.com \
--cc=samitolvanen@google.com \
--cc=snitzer@redhat.com \
--cc=vgoyal@redhat.com \
--cc=viresh.kumar@linaro.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.