From: Mike Galbraith <efault@gmx.de>
To: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>,
Ingo Molnar <mingo@elte.hu>,
Arjan van de Ven <arjan@linux.jf.intel.com>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>,
Yanmin Zhang <yanmin_zhang@linux.jf.intel.com>,
Gautham R Shenoy <ego@in.ibm.com>
Subject: Re: [patch 2/2] sched: fix select_idle_sibling() logic in select_task_rq_fair()
Date: Tue, 09 Mar 2010 14:27:00 +0100 [thread overview]
Message-ID: <1268141220.28032.12.camel@marge.simson.net> (raw)
In-Reply-To: <1268114749.6349.20.camel@marge.simson.net>
On Tue, 2010-03-09 at 07:05 +0100, Mike Galbraith wrote:
> On Mon, 2010-03-08 at 14:24 -0800, Suresh Siddha wrote:
> > Ok. I dropped that part in v2 patches that I just posted.
>
> Yeah, I see mails (ramble ramble) crossed in the night. I'll take them
> out for a spin, see if your box can get fixed up without busting mine :)
Ok, done, pretty darn boring testing. My worries wrt ramp didn't
materialize, and without the sync+1=idle bit, box no longer hates your
patch ;) It adds a tad of overhead though.
I tested it in my tinkered tree due to some things in virgin tip that I
think would have upset results.
tip = v2.6.33-6681-g91259c4
tip-x = tip + my patches
tip-xx = tip-x + your patches
netperf TCP_RR
unpinned
tip 64942.13 52203.28 92852.34 avg 69999.25 1.000
tip-x 107387.04 107334.34 107495.54 avg 107405.64 1.534 1.000
tip-xx 104602.54 104183.84 104095.35 avg 104293.91 1.489 .971
pinned
tip 101044.77 99478.75 100091.53 avg 100205.01 1.000
tip-x 103767.53 103138.71 103459.92 avg 103455.38 1.032 1.000
tip-xx 103465.68 103306.90 103189.46 avg 103320.68 1.031 .998
tbench 8
tip 1181.10 1173.81 1179.42 avg 1178.11 1.000
tip-x 1212.81 1214.73 1212.99 avg 1213.51 1.030 1.000
tip-xx 1212.56 1208.96 1207.57 avg 1209.69 1.026 .996
mysql+oltp
clients 1 2 4 8 16 32 64 128 256
tip 10364.73 19210.43 37055.69 36724.26 35797.53 34806.00 32533.02 27463.10 21331.35
8499.12 19438.84 36873.36 36574.78 35712.63 34763.84 32086.15 27325.47 20845.97
9707.81 19431.68 36940.02 36495.55 35590.46 34612.24 32125.14 27124.94 21228.64
tip avg 9523.88 19360.31 36956.35 36598.19 35700.20 34727.36 32248.10 27304.50 21135.32
tip-x 11143.63 20398.50 36954.97 36515.57 35877.48 35101.78 32974.35 28018.32 22172.95
11134.06 20615.39 36873.60 36541.21 35957.70 34954.68 32613.53 27737.31 21792.48
11156.34 20481.79 36773.47 36473.38 35778.64 34867.88 32485.94 27705.44 22024.18
tip-x avg 11144.67 20498.56 36867.34 36510.05 35871.27 34974.78 32691.27 27820.35 21996.53
vs tip 1.170 1.058 .997 .997 1.004 1.007 1.013 1.018 1.040
tip-xx 11208.50 20952.28 36842.91 36495.95 35642.27 35049.26 32566.01 27096.20 22058.94
11176.96 21123.91 37334.72 36849.41 36298.53 35269.21 32625.55 27248.00 21448.19
11112.04 21042.49 37317.42 36835.94 36147.29 35144.11 32695.83 27393.21 21881.37
tip-xx avg 11165.83 21039.56 37165.01 36727.10 36029.36 35154.19 32629.13 27245.80 21796.16
vs tip 1.172 1.086 1.005 1.003 1.009 1.012 1.011 .997 1.031
vs tip-x 1.001 1.026 1.008 1.005 1.004 1.005 .998 .979 .990
pgsql+oltp
clients 1 2 4 8 16 32 64 128 256
tip 15408.13 30181.76 53244.49 52684.00 52306.08 51115.08 50156.59 48907.81 45657.03
15658.69 30379.33 53308.72 52879.11 52169.98 51080.25 50023.30 48758.86 46009.55
15534.79 30343.67 53030.92 52437.59 52063.99 50897.84 50073.97 48720.65 45990.12
tip avg 15533.87 30301.58 53194.71 52666.90 52180.01 51031.05 50084.62 48795.77 45885.56
tip-x 15897.83 31418.82 54116.88 53712.74 52628.43 51310.64 50624.15 49291.35 46430.55
15848.06 30702.00 53166.82 52935.46 51859.48 51024.90 50123.57 48739.02 46033.40
16043.14 30919.32 53576.71 53400.79 52525.99 51635.50 50515.64 49095.82 46374.42
tip-x avg 15929.67 31013.38 53620.13 53349.66 52337.96 51323.68 50421.12 49042.06 46279.45
vs tip 1.025 1.023 1.007 1.012 1.003 1.005 1.006 1.005 1.008
tip-xx 16071.66 31145.83 53953.67 53280.52 52476.55 51472.32 50219.63 49016.45 46280.48
15899.41 30769.93 53051.25 53006.46 52000.32 51151.68 50062.48 48539.12 45634.23
16019.03 31035.88 53850.74 53086.54 52136.27 51246.17 50432.08 48544.87 45693.85
tip-xx avg 15996.70 30983.88 53618.55 53124.50 52204.38 51290.05 50238.06 48700.14 45869.52
vs tip 1.075 1.012 .879 1.001 1.006 1.022 1.010 1.013 1.074
vs tip-x 1.004 .999 .999 .995 .997 .999 .996 .993 .991
Context switching - times in microseconds - smaller is better
-------------------------------------------------------------------------
Host OS 2p/0K 2p/16K 2p/64K 8p/16K 8p/64K 16p/16K 16p/64K
ctxsw ctxsw ctxsw ctxsw ctxsw ctxsw ctxsw
--------- ------------- ------ ------ ------ ------ ------ ------- -------
marge 2.6.34-tip-smp 1.3400 1.4400 1.4000 2.1200 1.6400 2.26000 1.68000
marge 2.6.34-tip-smp 1.3600 1.3700 1.2700 2.1000 1.5600 2.26000 1.61000
marge 2.6.34-tip-smp 1.3800 1.3400 1.4000 2.1000 1.6400 2.29000 1.83000
marge 2.6.34-tip-smpx 1.0600 1.0900 1.0800 1.8200 1.3000 1.82000 1.33000
marge 2.6.34-tip-smpx 1.0700 1.1000 1.1200 1.8000 1.3100 1.83000 1.37000
marge 2.6.34-tip-smpx 1.0600 1.0900 1.1100 1.7800 1.2600 1.83000 1.38000
marge 2.6.34-tip-smpxx 1.1000 1.0900 1.1100 1.8000 1.3100 1.86000 1.37000
marge 2.6.34-tip-smpxx 1.1000 1.1000 1.1000 1.8000 1.3100 1.84000 1.35000
marge 2.6.34-tip-smpxx 1.1000 1.0800 1.1100 1.8100 1.3000 1.85000 1.36000
*Local* Communication latencies in microseconds - smaller is better
---------------------------------------------------------------------
Host OS 2p/0K Pipe AF UDP RPC/ TCP RPC/ TCP
ctxsw UNIX UDP TCP conn
--------- ------------- ----- ----- ---- ----- ----- ----- ----- ----
marge 2.6.34-tip-smp 1.340 4.179 4.42 8.959 12.1 27.4 14.9 18.
marge 2.6.34-tip-smp 1.360 4.369 4.61 8.868 12.0 26.9 14.9 18.
marge 2.6.34-tip-smp 1.380 4.364 4.70 8.931 11.9 25.6 14.9 18.
marge 2.6.34-tip-smpx 1.060 3.486 6.16 7.635 10.9 9.422 13.7 18.
marge 2.6.34-tip-smpx 1.070 3.456 5.93 7.677 11.0 9.412 13.7 18.
marge 2.6.34-tip-smpx 1.060 3.474 6.12 7.544 10.9 9.495 13.7 18.
marge 2.6.34-tip-smpxx 1.100 3.556 6.30 7.911 11.3 9.756 14.0 18.
marge 2.6.34-tip-smpxx 1.100 3.563 6.12 7.879 11.4 9.806 14.2 18.
marge 2.6.34-tip-smpxx 1.100 3.553 6.17 7.908 11.3 9.779 14.0 18.
*Local* Communication bandwidths in MB/s - bigger is better
-----------------------------------------------------------------------------
Host OS Pipe AF TCP File Mmap Bcopy Bcopy Mem Mem
UNIX reread reread (libc) (hand) read write
--------- ------------- ---- ---- ---- ------ ------ ------ ------ ---- -----
marge 2.6.34-tip-smp 1756 5168 754. 2818.3 4815.3 1243.7 1233.1 4342 1686.
marge 2.6.34-tip-smp 1778 5193 1152 2809.7 4748.9 1242.3 1236.1 4454 1679.
marge 2.6.34-tip-smp 1763 5161 1141 2801.5 4731.5 1246.6 1233.7 4493 1683.
marge 2.6.34-tip-smpx 2790 5230 2979 2838.2 4833.4 1322.5 1310.3 4488 1749.
marge 2.6.34-tip-smpx 2726 5222 2980 2825.6 4770.3 1336.4 1320.7 4447 1741.
marge 2.6.34-tip-smpx 2800 5214 2973 2817.3 4768.1 1383.3 1359.4 4447 1746.
marge 2.6.34-tip-smpxx 2549 5228 2931 2830.9 4795.9 1236.3 1229.3 4384 1686.
marge 2.6.34-tip-smpxx 2568 5212 740. 2833.0 4782.2 1239.2 1235.3 4462 1683.
marge 2.6.34-tip-smpxx 2608 5219 2949 2845.9 4806.4 1243.5 1232.7 4474 1683.
next prev parent reply other threads:[~2010-03-09 13:27 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-03-05 18:39 [patch 1/2] sched: check for prev_cpu == this_cpu in wake_affine() Suresh Siddha
2010-03-05 18:39 ` [patch 2/2] sched: fix select_idle_sibling() logic in select_task_rq_fair() Suresh Siddha
2010-03-05 20:25 ` Mike Galbraith
2010-03-06 8:21 ` Mike Galbraith
2010-03-08 22:24 ` Suresh Siddha
2010-03-09 6:05 ` Mike Galbraith
2010-03-09 13:27 ` Mike Galbraith [this message]
2010-03-10 21:14 ` Mike Galbraith
2010-03-06 8:36 ` Mike Galbraith
2010-03-05 19:36 ` [patch 1/2] sched: check for prev_cpu == this_cpu in wake_affine() Mike Galbraith
2010-03-08 19:09 ` Suresh Siddha
2010-03-08 22:25 ` Mike Galbraith
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=1268141220.28032.12.camel@marge.simson.net \
--to=efault@gmx.de \
--cc=a.p.zijlstra@chello.nl \
--cc=arjan@linux.jf.intel.com \
--cc=ego@in.ibm.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@elte.hu \
--cc=suresh.b.siddha@intel.com \
--cc=svaidy@linux.vnet.ibm.com \
--cc=yanmin_zhang@linux.jf.intel.com \
/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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
Powered by JetHome