From: Andrew Morton <akpm@linux-foundation.org>
To: "Wang, Yalin" <Yalin.Wang@sonymobile.com>
Cc: "'Kirill A. Shutemov'" <kirill@shutemov.name>,
"'arnd@arndb.de'" <arnd@arndb.de>,
"'linux-arch@vger.kernel.org'" <linux-arch@vger.kernel.org>,
"'linux-kernel@vger.kernel.org'" <linux-kernel@vger.kernel.org>,
"'linux@arm.linux.org.uk'" <linux@arm.linux.org.uk>,
"'linux-arm-kernel@lists.infradead.org'"
<linux-arm-kernel@lists.infradead.org>
Subject: Re: [RFC] change non-atomic bitops method
Date: Mon, 2 Feb 2015 22:38:51 -0800 [thread overview]
Message-ID: <20150202223851.f30768d0.akpm@linux-foundation.org> (raw)
In-Reply-To: <35FD53F367049845BC99AC72306C23D1044A02027E0C@CNBJMBX05.corpusers.net>
On Tue, 3 Feb 2015 13:42:45 +0800 "Wang, Yalin" <Yalin.Wang@sonymobile.com> wrote:
>
> ...
>
> #ifdef CHECK_BEFORE_SET
> if (p[i] != times)
> #endif
>
> ...
>
> ----
> One run on CPU0, reader thread run on CPU1,
> Test result:
> sudo ./cache_test
> reader:8.426228173
> 8.672198335
>
> With -DCHECK_BEFORE_SET
> sudo ./cache_test_check
> reader:7.537036819
> 10.799746531
>
You aren't measuring the right thing. You should compare
if (p[i] != x)
p[i] = x;
versus
p[i] = x;
and you should do this for two cases:
a) p[i] == x
b) p[i] != x
The first code sequence will be slower when (p[i] != x) and faster when
(p[i] == x).
Next, we should instrument the kernel to work out the frequency of
set_bit on an already-set bit.
It is only with both these ratios that we can work out whether the
patch is a net gain. My suspicion is that set_bit on an already-set
bit is so rare that the patch will be a loss.
WARNING: multiple messages have this Message-ID (diff)
From: akpm@linux-foundation.org (Andrew Morton)
To: linux-arm-kernel@lists.infradead.org
Subject: [RFC] change non-atomic bitops method
Date: Mon, 2 Feb 2015 22:38:51 -0800 [thread overview]
Message-ID: <20150202223851.f30768d0.akpm@linux-foundation.org> (raw)
In-Reply-To: <35FD53F367049845BC99AC72306C23D1044A02027E0C@CNBJMBX05.corpusers.net>
On Tue, 3 Feb 2015 13:42:45 +0800 "Wang, Yalin" <Yalin.Wang@sonymobile.com> wrote:
>
> ...
>
> #ifdef CHECK_BEFORE_SET
> if (p[i] != times)
> #endif
>
> ...
>
> ----
> One run on CPU0, reader thread run on CPU1,
> Test result:
> sudo ./cache_test
> reader:8.426228173
> 8.672198335
>
> With -DCHECK_BEFORE_SET
> sudo ./cache_test_check
> reader:7.537036819
> 10.799746531
>
You aren't measuring the right thing. You should compare
if (p[i] != x)
p[i] = x;
versus
p[i] = x;
and you should do this for two cases:
a) p[i] == x
b) p[i] != x
The first code sequence will be slower when (p[i] != x) and faster when
(p[i] == x).
Next, we should instrument the kernel to work out the frequency of
set_bit on an already-set bit.
It is only with both these ratios that we can work out whether the
patch is a net gain. My suspicion is that set_bit on an already-set
bit is so rare that the patch will be a loss.
next prev parent reply other threads:[~2015-02-03 6:39 UTC|newest]
Thread overview: 49+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-02-02 3:55 [RFC] change non-atomic bitops method Wang, Yalin
2015-02-02 3:55 ` Wang, Yalin
2015-02-02 18:53 ` Laura Abbott
2015-02-02 18:53 ` Laura Abbott
2015-02-02 19:31 ` Uwe Kleine-König
2015-02-02 19:31 ` Uwe Kleine-König
2015-02-03 15:14 ` David Howells
2015-02-03 15:14 ` David Howells
2015-02-03 15:14 ` David Howells
2015-02-03 19:10 ` Uwe Kleine-König
2015-02-03 19:10 ` Uwe Kleine-König
2015-02-02 23:29 ` Andrew Morton
2015-02-02 23:29 ` Andrew Morton
2015-02-02 23:31 ` Russell King - ARM Linux
2015-02-02 23:31 ` Russell King - ARM Linux
2015-02-03 1:17 ` Kirill A. Shutemov
2015-02-03 1:17 ` Kirill A. Shutemov
2015-02-03 2:13 ` Wang, Yalin
2015-02-03 2:13 ` Wang, Yalin
2015-02-03 5:42 ` Wang, Yalin
2015-02-03 5:42 ` Wang, Yalin
2015-02-03 6:38 ` Andrew Morton [this message]
2015-02-03 6:38 ` Andrew Morton
2015-02-03 7:03 ` Wang, Yalin
2015-02-03 7:03 ` Wang, Yalin
2015-02-03 8:42 ` Wang, Yalin
2015-02-03 8:42 ` Wang, Yalin
2015-02-03 10:59 ` Andrew Morton
2015-02-03 10:59 ` Andrew Morton
2015-02-09 8:18 ` Wang, Yalin
2015-02-09 8:18 ` Wang, Yalin
2015-02-09 20:34 ` Andrew Morton
2015-02-09 20:34 ` Andrew Morton
2015-02-10 7:05 ` Wang, Yalin
2015-02-10 7:05 ` Wang, Yalin
2015-02-09 21:42 ` Rasmus Villemoes
2015-02-09 21:42 ` Rasmus Villemoes
2015-02-09 21:42 ` Rasmus Villemoes
2015-02-03 8:40 ` David Miller
2015-02-03 8:40 ` David Miller
2015-02-03 8:48 ` Andrew Morton
2015-02-03 8:48 ` Andrew Morton
2015-02-03 9:34 ` Rasmus Villemoes
2015-02-03 9:34 ` Rasmus Villemoes
2015-02-03 9:34 ` Rasmus Villemoes
2015-02-03 9:41 ` Wang, Yalin
2015-02-03 9:41 ` Wang, Yalin
2015-02-03 10:39 ` Kirill A. Shutemov
2015-02-03 10:39 ` Kirill A. Shutemov
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=20150202223851.f30768d0.akpm@linux-foundation.org \
--to=akpm@linux-foundation.org \
--cc=Yalin.Wang@sonymobile.com \
--cc=arnd@arndb.de \
--cc=kirill@shutemov.name \
--cc=linux-arch@vger.kernel.org \
--cc=linux-arm-kernel@lists.infradead.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux@arm.linux.org.uk \
/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.