From mboxrd@z Thu Jan  1 00:00:00 1970
Content-Type: multipart/mixed; boundary="===============4396096619684650181=="
MIME-Version: 1.0
From: Ye Xiaolong <xiaolong.ye@intel.com>
To: lkp@lists.01.org
Subject: Re: [x86/copy_user] adb402cd14: will-it-scale.per_process_ops -12.7% regression
Date: Wed, 16 Nov 2016 20:38:51 +0800
Message-ID: <20161116123851.GA12465@yexl-desktop>
In-Reply-To: <20161116095821.k4iico5a45f4mql7@pd.tnic>
List-Id: <oe-lkp.lists.linux.dev>

--===============4396096619684650181==
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable

On 11/16, Borislav Petkov wrote:
>On Mon, Nov 14, 2016 at 10:19:31PM +0800, Ye Xiaolong wrote:
>> On 11/15, Borislav Petkov wrote:
>> >On Mon, Nov 14, 2016 at 10:37:11PM +0800, Ye Xiaolong wrote:
>> >> Below cmd should help on it:
>> >
>> >Right, before I go do this, can you please retest without
>> >
>> >CONFIG_DEBUG_ATOMIC_SLEEP=3Dy
>> >
>> >?
>> =

>> Sorry for the late, here is the comparison, disable
>> CONFIG_DEBUG_ATOMIC_SLEEP helps to restore the performance back
>> according to the result.
>> =

>> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
>> commit/compiler/cpufreq_governor/rootfs/tbox_group/test/testcase:
>>   adb402cd1461eef6e1a21db4532a3b9e6a6be853/gcc-6/performance/debian-x86_=
64-2016-08-31.cgz/xps/poll1/will-it-scale
>> =

>> kconfig:
>>   x86_64-rhel-7.2
>>   x86_64-rhel-7.2-CONFIG_DEBUG_ATOMIC_SLEEP
>> =

>>  x86_64-rhel-7.2 x86_64-rhel-7.2-CONFIG_DEB
>> ---------------- --------------------------
>>          %stddev     %change         %stddev
>>              \          |                \
>>    5899566 =C2=B1  2%     +28.5%    7583094 =C2=B1  0%  will-it-scale.pe=
r_process_ops
>>    5121156 =C2=B1  0%     +21.1%    6203340 =C2=B1  0%  will-it-scale.pe=
r_thread_ops
>
>Btw, just to make sure I understand the graphs correctly: that config above
>
>"x86_64-rhel-7.2-CONFIG_DEBUG_ATOMIC_SLEEP"
>
>is .config *without* CONFIG_DEBUG_ATOMIC_SLEEP, yes? I.e., config
>"minus" CONFIG_DEBUG_ATOMIC_SLEEP.

Yes, you are right, '-' means 'minus' here.

Thanks,
Xiaolong
>
>Right?
>
>Thanks.
>
>-- =

>Regards/Gruss,
>    Boris.
>
>Good mailing practices for 400: avoid top-posting and trim the reply.

--===============4396096619684650181==--