All of lore.kernel.org
 help / color / mirror / Atom feed
From: Dietmar Eggemann <dietmar.eggemann@arm.com>
To: lkp@lists.01.org
Subject: Re: [sched/fair] 98d8fd8126: -20.8% hackbench.throughput
Date: Mon, 12 Oct 2015 19:43:46 +0100	[thread overview]
Message-ID: <561BFF62.20503@arm.com> (raw)
In-Reply-To: <87si64mv9a.fsf@yhuang-dev.intel.com>

[-- Attachment #1: Type: text/plain, Size: 5140 bytes --]

Hi Ying,

On 24/09/15 03:00, kernel test robot wrote:
> FYI, we noticed the below changes on
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git sched/core
> commit 98d8fd8126676f7ba6e133e65b2ca4b17989d32c ("sched/fair: Initialize task load and utilization before placing task on rq")
> 
> 
> =========================================================================================
> tbox_group/testcase/rootfs/kconfig/compiler/nr_threads/mode/ipc:
>   lkp-ws02/hackbench/debian-x86_64-2015-02-07.cgz/x86_64-rhel/gcc-4.9/1600%/process/pipe
> 
> commit: 
>   231678b768da07d19ab5683a39eeb0c250631d02
>   98d8fd8126676f7ba6e133e65b2ca4b17989d32c
> 
> 231678b768da07d1 98d8fd8126676f7ba6e133e65b 
> ---------------- -------------------------- 
>          %stddev     %change         %stddev
>              \          |                \  
>     188818 ±  1%     -20.8%     149585 ±  1%  hackbench.throughput

[...]

> 
> lkp-ws02: Westmere-EP
> Memory: 16G
> 
> 
> 
> 
>                       hackbench.time.involuntary_context_switches
> 
>     3e+08 ++------------------------O---------------------------------------+
>           O  O         O   O     O                                          |
>   2.5e+08 ++     O                                                O  O   O  O
>           |         O                   O     O      O                      |
>           |                   O                   O     O   O               |
>     2e+08 ++                               O                                |
>           |                                                    O            |
>   1.5e+08 ++                                                                |
>           |                                                                 |
>     1e+08 ++                                                                |
>           |                                  .*...*..*..*...*..*            |
>           *..*...*..*..*...*..  .*..  ..*..*.                               |
>     5e+07 ++                  *.    *.                                      |
>           |                                                                 |
>         0 ++----------------------------------------------------------------+
> 
>                                   vmstat.system.in
> 
>   300000 ++-----------------------------------------------------------------+
>          |                O     O   O                                       |
>          O            O      O                                       O   O  |
>   250000 ++ O   O                                                 O         O
>          |         O                                                        |
>          |                             O      O  O   O  O  O                |
>   200000 ++                                O                   O            |
>          |                                                                  |
>   150000 ++                                                                 |
>          |                                                                  |
>          |                                                                  |
>   100000 ++                                                                 |
>          |           .*...                   .*..*...*..*..*...*            |
>          *..*...*..*.     *..  .*...  .*...*.                               |
>    50000 ++------------------*------*---------------------------------------+
> 
> 	[*] bisect-good sample
> 	[O] bisect-bad  sample
> 
> To reproduce:
> 
>         git clone git://git.kernel.org/pub/scm/linux/kernel/git/wfg/lkp-tests.git
>         cd lkp-tests
>         bin/lkp install job.yaml  # job file is attached in this email
>         bin/lkp run     job.yaml

I try to recreate this on one of my Intel machines (Xeon CPU E5-2650 v2
@ 2.60GHz) w/ 16 logical cpus. We haven't seen anything near a 20%
performance degradation for hackbench when we were running our hackbench
tests on 5/6 core arm machines as well as on a IVB-EP (2*10*2) Intel
machine.

So I cloned the repo:

# git clone git://git.kernel.org/pub/scm/linux/kernel/git/wfg
/lkp-tests.git lkp-tests

and ran the hackbench example:

root # lkp install $LKP_SRC/jobs/hackbench.yaml

root # lkp split-job $LKP_SRC/jobs/hackbench.yaml

root # lkp run ./hackbench-50%-threads-socket.yaml

2015-10-12 19:27:20 /usr/bin/hackbench -g 8 --threads -l 60000
Running in threaded mode with 8 groups using 40 file descriptors each
(== 320 tasks)
Each sender will pass 60000 messages of 100 bytes
...
wait for background monitors: perf-profile uptime proc-vmstat proc-stat
meminfo slabinfo interrupts softirqs diskstats cpuidle turbostat sched_debug

root # lkp result hackbench

/result/hackbench/50%-threads-socket-50/$MACHINE/ubuntu/defconfig/gcc-4.8/3.16.0-50-generic/0/

But I can't get any statistics out of it?

root # lkp stat hackbench
runs sum average stddev% case

Thanks for your help in advance!

-- Dietmar

[...]


WARNING: multiple messages have this Message-ID (diff)
From: Dietmar Eggemann <dietmar.eggemann@arm.com>
To: kernel test robot <ying.huang@intel.com>,
	Morten Rasmussen <Morten.Rasmussen@arm.com>
Cc: "lkp@01.org" <lkp@01.org>, LKML <linux-kernel@vger.kernel.org>,
	Thomas Gleixner <tglx@linutronix.de>,
	Mike Galbraith <efault@gmx.de>,
	Linus Torvalds <torvalds@linux-foundation.org>,
	Juri Lelli <Juri.Lelli@arm.com>,
	Peter Zijlstra <peterz@infradead.org>,
	Ingo Molnar <mingo@redhat.com>, Ingo Molnar <mingo@kernel.org>
Subject: Re: [lkp] [sched/fair] 98d8fd8126: -20.8% hackbench.throughput
Date: Mon, 12 Oct 2015 19:43:46 +0100	[thread overview]
Message-ID: <561BFF62.20503@arm.com> (raw)
In-Reply-To: <87si64mv9a.fsf@yhuang-dev.intel.com>

Hi Ying,

On 24/09/15 03:00, kernel test robot wrote:
> FYI, we noticed the below changes on
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git sched/core
> commit 98d8fd8126676f7ba6e133e65b2ca4b17989d32c ("sched/fair: Initialize task load and utilization before placing task on rq")
> 
> 
> =========================================================================================
> tbox_group/testcase/rootfs/kconfig/compiler/nr_threads/mode/ipc:
>   lkp-ws02/hackbench/debian-x86_64-2015-02-07.cgz/x86_64-rhel/gcc-4.9/1600%/process/pipe
> 
> commit: 
>   231678b768da07d19ab5683a39eeb0c250631d02
>   98d8fd8126676f7ba6e133e65b2ca4b17989d32c
> 
> 231678b768da07d1 98d8fd8126676f7ba6e133e65b 
> ---------------- -------------------------- 
>          %stddev     %change         %stddev
>              \          |                \  
>     188818 ±  1%     -20.8%     149585 ±  1%  hackbench.throughput

[...]

> 
> lkp-ws02: Westmere-EP
> Memory: 16G
> 
> 
> 
> 
>                       hackbench.time.involuntary_context_switches
> 
>     3e+08 ++------------------------O---------------------------------------+
>           O  O         O   O     O                                          |
>   2.5e+08 ++     O                                                O  O   O  O
>           |         O                   O     O      O                      |
>           |                   O                   O     O   O               |
>     2e+08 ++                               O                                |
>           |                                                    O            |
>   1.5e+08 ++                                                                |
>           |                                                                 |
>     1e+08 ++                                                                |
>           |                                  .*...*..*..*...*..*            |
>           *..*...*..*..*...*..  .*..  ..*..*.                               |
>     5e+07 ++                  *.    *.                                      |
>           |                                                                 |
>         0 ++----------------------------------------------------------------+
> 
>                                   vmstat.system.in
> 
>   300000 ++-----------------------------------------------------------------+
>          |                O     O   O                                       |
>          O            O      O                                       O   O  |
>   250000 ++ O   O                                                 O         O
>          |         O                                                        |
>          |                             O      O  O   O  O  O                |
>   200000 ++                                O                   O            |
>          |                                                                  |
>   150000 ++                                                                 |
>          |                                                                  |
>          |                                                                  |
>   100000 ++                                                                 |
>          |           .*...                   .*..*...*..*..*...*            |
>          *..*...*..*.     *..  .*...  .*...*.                               |
>    50000 ++------------------*------*---------------------------------------+
> 
> 	[*] bisect-good sample
> 	[O] bisect-bad  sample
> 
> To reproduce:
> 
>         git clone git://git.kernel.org/pub/scm/linux/kernel/git/wfg/lkp-tests.git
>         cd lkp-tests
>         bin/lkp install job.yaml  # job file is attached in this email
>         bin/lkp run     job.yaml

I try to recreate this on one of my Intel machines (Xeon CPU E5-2650 v2
@ 2.60GHz) w/ 16 logical cpus. We haven't seen anything near a 20%
performance degradation for hackbench when we were running our hackbench
tests on 5/6 core arm machines as well as on a IVB-EP (2*10*2) Intel
machine.

So I cloned the repo:

# git clone git://git.kernel.org/pub/scm/linux/kernel/git/wfg
/lkp-tests.git lkp-tests

and ran the hackbench example:

root # lkp install $LKP_SRC/jobs/hackbench.yaml

root # lkp split-job $LKP_SRC/jobs/hackbench.yaml

root # lkp run ./hackbench-50%-threads-socket.yaml

2015-10-12 19:27:20 /usr/bin/hackbench -g 8 --threads -l 60000
Running in threaded mode with 8 groups using 40 file descriptors each
(== 320 tasks)
Each sender will pass 60000 messages of 100 bytes
...
wait for background monitors: perf-profile uptime proc-vmstat proc-stat
meminfo slabinfo interrupts softirqs diskstats cpuidle turbostat sched_debug

root # lkp result hackbench

/result/hackbench/50%-threads-socket-50/$MACHINE/ubuntu/defconfig/gcc-4.8/3.16.0-50-generic/0/

But I can't get any statistics out of it?

root # lkp stat hackbench
runs sum average stddev% case

Thanks for your help in advance!

-- Dietmar

[...]


  reply	other threads:[~2015-10-12 18:43 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-09-24  2:00 [sched/fair] 98d8fd8126: -20.8% hackbench.throughput kernel test robot
2015-09-24  2:00 ` [lkp] " kernel test robot
2015-10-12 18:43 ` Dietmar Eggemann [this message]
2015-10-12 18:43   ` Dietmar Eggemann
2015-10-14  1:48   ` Huang, Ying
2015-10-14  1:48     ` [lkp] " Huang, Ying

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=561BFF62.20503@arm.com \
    --to=dietmar.eggemann@arm.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.