From: Fengguang Wu <fengguang.wu@intel.com>
To: lkp@lists.01.org
Subject: 3-5% increased netperf throughput by "sched: Micro-optimize the smart wake-affine logic"
Date: Sat, 07 Sep 2013 20:38:19 +0800 [thread overview]
Message-ID: <20130907123819.GA705@localhost> (raw)
[-- Attachment #1: Type: text/plain, Size: 11249 bytes --]
Hi Peter,
We are glad to report some measurable performance improvements by your
commit
commit 7d9ffa8961482232d964173cccba6e14d2d543b2
Author: Peter Zijlstra <peterz@infradead.org>
Date: Thu Jul 4 12:56:46 2013 +0800
sched: Micro-optimize the smart wake-affine logic
Smart wake-affine is using node-size as the factor currently, but the overhead
of the mask operation is high.
Thus, this patch introduce the 'sd_llc_size' percpu variable, which will record
the highest cache-share domain size, and make it to be the new factor, in order
to reduce the overhead and make it more reasonable.
Tested-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Tested-by: Michael Wang <wangyun@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Acked-by: Michael Wang <wangyun@linux.vnet.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Link: http://lkml.kernel.org/r/51D5008E.6030102(a)linux.vnet.ibm.com
[ Tidied up the changelog. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
:040000 040000 e7c8a8c55bfa1261f3c6b75674a83eb76bb88a3f 129777b8d0b74ce189760ad76d9aaecd65b7ee7f M kernel
bisect run success
# bad: [37570e7ef5be99ba5188bb17ed547ac4bbf65e73] Merge remote-tracking branch 'nfc-next/master' into devel-hourly-2013090406
# good: [6e4664525b1db28f8c4e1130957f70a94c19213e] Linux 3.11
git bisect start '37570e7ef5be99ba5188bb17ed547ac4bbf65e73' '6e4664525b1db28f8c4e1130957f70a94c19213e' '--'
# good: [8bcaa20433634ac70c96d9e5f8ece4b8577c9694] Merge remote-tracking branch 'arm-soc/for-next' into devel-hourly-2013090406
git bisect good 8bcaa20433634ac70c96d9e5f8ece4b8577c9694
# good: [820acdf740b7d04476959189e9a144c2315339a4] drm/i915: do display power state notification on crtc enable/disable
git bisect good 820acdf740b7d04476959189e9a144c2315339a4
# bad: [5bae522a51aa6bbae54bd2d745d0320f74c40b76] Merge remote-tracking branch 'perf/perf/trace.fmt' into devel-hourly-2013090406
git bisect bad 5bae522a51aa6bbae54bd2d745d0320f74c40b76
# bad: [8afb4c018e21c882c8fad196772ef74d494185e2] perf tools: Re-implement debug print function for linking python/perf.so
git bisect bad 8afb4c018e21c882c8fad196772ef74d494185e2
# good: [17f41571bb2c4a398785452ac2718a6c5d77180e] kprobes/x86: Call out into INT3 handler directly instead of using notifier
git bisect good 17f41571bb2c4a398785452ac2718a6c5d77180e
# bad: [34f77abcb34e1da4ee3ca5c5a41b673664eee1fa] perf annotate: Put dso name in symbol annotation title
git bisect bad 34f77abcb34e1da4ee3ca5c5a41b673664eee1fa
# bad: [8404db63461af62025f32f8368861fb33604e62f] perf tests: Add attr record group sampling test
git bisect bad 8404db63461af62025f32f8368861fb33604e62f
# bad: [9a545de019b536771feefb76f85e5038b65c2190] perf: Migrate per cpu event accounting
git bisect bad 9a545de019b536771feefb76f85e5038b65c2190
# good: [62470419e993f8d9d93db0effd3af4296ecb79a5] sched: Implement smarter wake-affine logic
git bisect good 62470419e993f8d9d93db0effd3af4296ecb79a5
# bad: [90983b16078ab0fdc58f0dab3e8e3da79c9579a2] perf: Sanitize get_callchain_buffer()
git bisect bad 90983b16078ab0fdc58f0dab3e8e3da79c9579a2
# bad: [6050cb0b0b366092d1383bc23d7b16cd26db00f0] perf: Fix branch stack refcount leak on callchain init failure
git bisect bad 6050cb0b0b366092d1383bc23d7b16cd26db00f0
# bad: [7d9ffa8961482232d964173cccba6e14d2d543b2] sched: Micro-optimize the smart wake-affine logic
git bisect bad 7d9ffa8961482232d964173cccba6e14d2d543b2
# first bad commit: [7d9ffa8961482232d964173cccba6e14d2d543b2] sched: Micro-optimize the smart wake-affine logic
A comparison of all good commits [*] with all bad commits [o]
(good/bad in the sense of git bisect)
netperf.Throughput_Mbps
208 ++-------------------------------------------------------------------+
206 +OOO O OOO OOOO O O O O O O O |
O O O O OO O O O O O O O O
204 ++ O O O O O OOOO O OOOOO OOO O OO OOO OO|
202 ++ O O |
| O |
200 ++ |
198 ++ |
196 ++ |
| * |
194 ++ ****. *** .**** **.*******.* *** |
192 ++ * * ** :: |
| * *** .* : * |
190 ** *.* **.** **** * |
188 ++------------*------------------------------------------------------+
vmstat.system.in
1640 ++----------O-------------------------------------------------------+
O O OO O O O O |
1620 +O O O O O O O O O O O OO
| O OOO OO OO O O O O O OO O |
1600 ++ O O O O O O OO O O O O |
| O O O O O O O O |
1580 ++ O O O |
| |
1560 ++ * |
| * * * * :* |
1540 ++ * :: ***.* * :**.* ::* * .** :* |
| :+ * * * ** ** ::* * * * * * |
1520 ++* * ::*** +: ** : * |
* : * * :** :: |
1500 +*-------------*----*-----------------------------------------------+
vmstat.system.cs
10000 ++-----------------------------------------------------------------+
*****.***** |
9800 ++ ** .****** |
9600 ++ * :* ** **.* * *** .* * **** |
| * *.** ** ** * * * * *.* |
9400 ++ |
9200 ++ |
| |
9000 ++ |
8800 ++ O |
O O OO O O OOO O O OOO OO OO OO O |
8600 ++O O O O OO OO O O O O OO O O OO O O OO OO
8400 +O O OOO O OO O O OO O O |
| O |
8200 ++-----------------------------------------------------------------+
lock_stat.slock-AF_INET.contentions
110000 ++----------------------------------------------------------------+
| |
105000 ++ O O O O |
OOO O O O O OO O O O O O O O O
| O O O O OOO O O OOOOO O OOO OOO O OOO OOO O|
100000 ++ O O OOO O O O O O |
| O |
95000 ++ |
| |
90000 ++ * * .*** * ** *. * *** *** |
| * ** * * * * * * * ****.* |
| **.* **** ::*. ** : |
85000 *** * * * * * |
| |
80000 ++----------------------------------------------------------------+
lock_stat.slock-AF_INET.contentions.lock_sock_nested
92000 ++-----------------------------------------------------------------+
90000 ++ O O O |
| O O O O O O |
88000 OO O O O OO O O O OO O OO OO O O O OO
86000 ++ O O O OO O OO OO OO O O OOOOOO O |
84000 ++ OO O O O O O O |
82000 ++ O |
| |
80000 ++ |
78000 ++ * *. * * * *. |
76000 ++ * ***.** * * : * * :* **.** *** * * |
74000 ++ **.* * :+ ** : * * * * * |
| * * ***: *** * |
72000 ** * |
70000 ++-----------------------------------------------------------------+
lock_stat.slock-AF_INET.contentions.tcp_v4_rcv
110000 ++----------------------------------------------------------------+
| |
105000 ++ O O O O |
OOO O O O O OO O O O O O O O O
| O O O O OOO O O OOOOO O OOO OOO O OOO OOO O|
100000 ++ O O OOO O O O O O |
| O |
95000 ++ |
| |
90000 ++ * * .*** * ** *. * *** *** |
| * ** * * * * * * * ****.* |
| **.* **** ::*. ** : |
85000 *** * * * * * |
| |
80000 ++----------------------------------------------------------------+
WARNING: multiple messages have this Message-ID (diff)
From: Fengguang Wu <fengguang.wu@intel.com>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Yuanhan Liu <yuanhan.liu@linux.intel.com>,
"Huang, Ying" <ying.huang@intel.com>,
Fengguang Wu <fengguang.wu@intel.com>,
Mike Galbraith <efault@gmx.de>, Ingo Molnar <mingo@kernel.org>,
LKML <linux-kernel@vger.kernel.org>,
lkp@01.org
Subject: 3-5% increased netperf throughput by "sched: Micro-optimize the smart wake-affine logic"
Date: Sat, 7 Sep 2013 20:38:19 +0800 [thread overview]
Message-ID: <20130907123819.GA705@localhost> (raw)
Hi Peter,
We are glad to report some measurable performance improvements by your
commit
commit 7d9ffa8961482232d964173cccba6e14d2d543b2
Author: Peter Zijlstra <peterz@infradead.org>
Date: Thu Jul 4 12:56:46 2013 +0800
sched: Micro-optimize the smart wake-affine logic
Smart wake-affine is using node-size as the factor currently, but the overhead
of the mask operation is high.
Thus, this patch introduce the 'sd_llc_size' percpu variable, which will record
the highest cache-share domain size, and make it to be the new factor, in order
to reduce the overhead and make it more reasonable.
Tested-by: Davidlohr Bueso <davidlohr.bueso@hp.com>
Tested-by: Michael Wang <wangyun@linux.vnet.ibm.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Acked-by: Michael Wang <wangyun@linux.vnet.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Link: http://lkml.kernel.org/r/51D5008E.6030102@linux.vnet.ibm.com
[ Tidied up the changelog. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
:040000 040000 e7c8a8c55bfa1261f3c6b75674a83eb76bb88a3f 129777b8d0b74ce189760ad76d9aaecd65b7ee7f M kernel
bisect run success
# bad: [37570e7ef5be99ba5188bb17ed547ac4bbf65e73] Merge remote-tracking branch 'nfc-next/master' into devel-hourly-2013090406
# good: [6e4664525b1db28f8c4e1130957f70a94c19213e] Linux 3.11
git bisect start '37570e7ef5be99ba5188bb17ed547ac4bbf65e73' '6e4664525b1db28f8c4e1130957f70a94c19213e' '--'
# good: [8bcaa20433634ac70c96d9e5f8ece4b8577c9694] Merge remote-tracking branch 'arm-soc/for-next' into devel-hourly-2013090406
git bisect good 8bcaa20433634ac70c96d9e5f8ece4b8577c9694
# good: [820acdf740b7d04476959189e9a144c2315339a4] drm/i915: do display power state notification on crtc enable/disable
git bisect good 820acdf740b7d04476959189e9a144c2315339a4
# bad: [5bae522a51aa6bbae54bd2d745d0320f74c40b76] Merge remote-tracking branch 'perf/perf/trace.fmt' into devel-hourly-2013090406
git bisect bad 5bae522a51aa6bbae54bd2d745d0320f74c40b76
# bad: [8afb4c018e21c882c8fad196772ef74d494185e2] perf tools: Re-implement debug print function for linking python/perf.so
git bisect bad 8afb4c018e21c882c8fad196772ef74d494185e2
# good: [17f41571bb2c4a398785452ac2718a6c5d77180e] kprobes/x86: Call out into INT3 handler directly instead of using notifier
git bisect good 17f41571bb2c4a398785452ac2718a6c5d77180e
# bad: [34f77abcb34e1da4ee3ca5c5a41b673664eee1fa] perf annotate: Put dso name in symbol annotation title
git bisect bad 34f77abcb34e1da4ee3ca5c5a41b673664eee1fa
# bad: [8404db63461af62025f32f8368861fb33604e62f] perf tests: Add attr record group sampling test
git bisect bad 8404db63461af62025f32f8368861fb33604e62f
# bad: [9a545de019b536771feefb76f85e5038b65c2190] perf: Migrate per cpu event accounting
git bisect bad 9a545de019b536771feefb76f85e5038b65c2190
# good: [62470419e993f8d9d93db0effd3af4296ecb79a5] sched: Implement smarter wake-affine logic
git bisect good 62470419e993f8d9d93db0effd3af4296ecb79a5
# bad: [90983b16078ab0fdc58f0dab3e8e3da79c9579a2] perf: Sanitize get_callchain_buffer()
git bisect bad 90983b16078ab0fdc58f0dab3e8e3da79c9579a2
# bad: [6050cb0b0b366092d1383bc23d7b16cd26db00f0] perf: Fix branch stack refcount leak on callchain init failure
git bisect bad 6050cb0b0b366092d1383bc23d7b16cd26db00f0
# bad: [7d9ffa8961482232d964173cccba6e14d2d543b2] sched: Micro-optimize the smart wake-affine logic
git bisect bad 7d9ffa8961482232d964173cccba6e14d2d543b2
# first bad commit: [7d9ffa8961482232d964173cccba6e14d2d543b2] sched: Micro-optimize the smart wake-affine logic
A comparison of all good commits [*] with all bad commits [o]
(good/bad in the sense of git bisect)
netperf.Throughput_Mbps
208 ++-------------------------------------------------------------------+
206 +OOO O OOO OOOO O O O O O O O |
O O O O OO O O O O O O O O
204 ++ O O O O O OOOO O OOOOO OOO O OO OOO OO|
202 ++ O O |
| O |
200 ++ |
198 ++ |
196 ++ |
| * |
194 ++ ****. *** .**** **.*******.* *** |
192 ++ * * ** :: |
| * *** .* : * |
190 ** *.* **.** **** * |
188 ++------------*------------------------------------------------------+
vmstat.system.in
1640 ++----------O-------------------------------------------------------+
O O OO O O O O |
1620 +O O O O O O O O O O O OO
| O OOO OO OO O O O O O OO O |
1600 ++ O O O O O O OO O O O O |
| O O O O O O O O |
1580 ++ O O O |
| |
1560 ++ * |
| * * * * :* |
1540 ++ * :: ***.* * :**.* ::* * .** :* |
| :+ * * * ** ** ::* * * * * * |
1520 ++* * ::*** +: ** : * |
* : * * :** :: |
1500 +*-------------*----*-----------------------------------------------+
vmstat.system.cs
10000 ++-----------------------------------------------------------------+
*****.***** |
9800 ++ ** .****** |
9600 ++ * :* ** **.* * *** .* * **** |
| * *.** ** ** * * * * *.* |
9400 ++ |
9200 ++ |
| |
9000 ++ |
8800 ++ O |
O O OO O O OOO O O OOO OO OO OO O |
8600 ++O O O O OO OO O O O O OO O O OO O O OO OO
8400 +O O OOO O OO O O OO O O |
| O |
8200 ++-----------------------------------------------------------------+
lock_stat.slock-AF_INET.contentions
110000 ++----------------------------------------------------------------+
| |
105000 ++ O O O O |
OOO O O O O OO O O O O O O O O
| O O O O OOO O O OOOOO O OOO OOO O OOO OOO O|
100000 ++ O O OOO O O O O O |
| O |
95000 ++ |
| |
90000 ++ * * .*** * ** *. * *** *** |
| * ** * * * * * * * ****.* |
| **.* **** ::*. ** : |
85000 *** * * * * * |
| |
80000 ++----------------------------------------------------------------+
lock_stat.slock-AF_INET.contentions.lock_sock_nested
92000 ++-----------------------------------------------------------------+
90000 ++ O O O |
| O O O O O O |
88000 OO O O O OO O O O OO O OO OO O O O OO
86000 ++ O O O OO O OO OO OO O O OOOOOO O |
84000 ++ OO O O O O O O |
82000 ++ O |
| |
80000 ++ |
78000 ++ * *. * * * *. |
76000 ++ * ***.** * * : * * :* **.** *** * * |
74000 ++ **.* * :+ ** : * * * * * |
| * * ***: *** * |
72000 ** * |
70000 ++-----------------------------------------------------------------+
lock_stat.slock-AF_INET.contentions.tcp_v4_rcv
110000 ++----------------------------------------------------------------+
| |
105000 ++ O O O O |
OOO O O O O OO O O O O O O O O
| O O O O OOO O O OOOOO O OOO OOO O OOO OOO O|
100000 ++ O O OOO O O O O O |
| O |
95000 ++ |
| |
90000 ++ * * .*** * ** *. * *** *** |
| * ** * * * * * * * ****.* |
| **.* **** ::*. ** : |
85000 *** * * * * * |
| |
80000 ++----------------------------------------------------------------+
next reply other threads:[~2013-09-07 12:38 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-09-07 12:38 Fengguang Wu [this message]
2013-09-07 12:38 ` 3-5% increased netperf throughput by "sched: Micro-optimize the smart wake-affine logic" Fengguang Wu
[not found] <20130907123251.GA32057@localhost>
2013-09-07 12:37 ` Fengguang Wu
2013-09-07 12:37 ` Fengguang Wu
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=20130907123819.GA705@localhost \
--to=fengguang.wu@intel.com \
--cc=lkp@lists.01.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.