[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <530C0996.3070405@linaro.org>
Date: Tue, 25 Feb 2014 11:10:14 +0800
From: Alex Shi <alex.shi@...aro.org>
To: Fengguang Wu <fengguang.wu@...el.com>
CC: LKML <linux-kernel@...r.kernel.org>
Subject: Re: [sched/balance] 7511dd0a7: +2.1e+05% context switches
On 02/25/2014 09:33 AM, Fengguang Wu wrote:
> On Thu, Feb 20, 2014 at 03:21:29PM +0800, Alex Shi wrote:
>> On 02/19/2014 09:00 PM, Fengguang Wu wrote:
>>> bc575710efe937e 7511dd0a73aaf2ca4bcd829f9
>>> --------------- -------------------------
>>> 2029 ~ 0% +222.9% 6551 ~17% lkp-snb01/micro/will-it-scale/pthread_mutex2
>>> 143678 ~42% +4.8e+05% 6.927e+08 ~ 0% lkp-snb01/micro/will-it-scale/sched_yield
>>> 145708 ~42% +4.8e+05% 6.927e+08 ~ 0% TOTAL time.involuntary_context_switches
>>
>>
>> Thanks for testing, Fengguang!
>>
>> Does the context switch increasing happen on whole patchset? or just
>> happens on this patch?
>
> Only this patch. Some other patches actually reduce the context switches.
> This is all the changes for branch alexshi/topdown:
Thanks a lot for great testing!
Would you like sharing the every benchmarks run on each of testing
machine, then we not only know the benchmark change on the machine, but
also know what benchmark was not effected on the machine?
And it will be perfect if you like to give a URL for benchmarks which
you are using. :)
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 111372 ~ 1% +60.1% 178267 ~ 1% lkp-ws02/micro/hackbench/1600%-threads-pipe
> 111372 ~ 1% +60.1% 178267 ~ 1% TOTAL hackbench.throughput
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 11862017 ~10% +769.1% 1.031e+08 ~ 9% lkp-ws02/micro/hackbench/1600%-threads-pipe
> 11862017 ~10% +769.1% 1.031e+08 ~ 9% TOTAL numa-numastat.node0.numa_hit
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 11862016 ~10% +769.1% 1.031e+08 ~ 9% lkp-ws02/micro/hackbench/1600%-threads-pipe
> 11862016 ~10% +769.1% 1.031e+08 ~ 9% TOTAL numa-numastat.node0.local_node
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 4629280 ~ 9% +714.0% 37683303 ~ 9% lkp-ws02/micro/hackbench/1600%-threads-pipe
> 4629280 ~ 9% +714.0% 37683303 ~ 9% TOTAL proc-vmstat.pgalloc_dma32
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 6476588 ~11% +751.1% 55123483 ~ 7% lkp-ws02/micro/hackbench/1600%-threads-pipe
> 6476588 ~11% +751.1% 55123483 ~ 7% TOTAL numa-vmstat.node0.numa_hit
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 6380486 ~11% +761.9% 54995307 ~ 7% lkp-ws02/micro/hackbench/1600%-threads-pipe
> 6380486 ~11% +761.9% 54995307 ~ 7% TOTAL numa-vmstat.node0.numa_local
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 24214 ~ 1% -71.8% 6824 ~ 7% lkp-a03/micro/will-it-scale/sched_yield
> 55173 ~ 0% -76.9% 12720 ~ 0% lkp-snb01/micro/will-it-scale/pthread_mutex2
> 55292 ~ 0% -78.9% 11653 ~ 3% lkp-snb01/micro/will-it-scale/sched_yield
> 134680 ~ 0% -76.8% 31198 ~ 3% TOTAL interrupts.RES
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 25891 ~ 1% +99.5% 51664 ~ 0% lkp-a03/micro/will-it-scale/sched_yield
> 25891 ~ 1% +99.5% 51664 ~ 0% TOTAL softirqs.SCHED
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 4282244 ~ 1% -46.6% 2284599 ~ 8% lkp-ws02/micro/hackbench/1600%-threads-pipe
> 4282244 ~ 1% -46.6% 2284599 ~ 8% TOTAL proc-vmstat.pgfault
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 105276 ~ 0% -43.3% 59659 ~ 0% lkp-snb01/micro/will-it-scale/pthread_mutex2
> 105799 ~ 1% -38.0% 65577 ~ 1% lkp-snb01/micro/will-it-scale/sched_yield
> 211075 ~ 0% -40.7% 125236 ~ 0% TOTAL cpuidle.C7-SNB.usage
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 4.174e+08 ~ 0% +31.2% 5.478e+08 ~ 0% lkp-a03/micro/will-it-scale/sched_yield
> 4.174e+08 ~ 0% +31.2% 5.478e+08 ~ 0% TOTAL cpuidle.C4.time
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 20.64 ~ 0% +23.9% 25.57 ~ 0% lkp-snb01/micro/will-it-scale/sched_yield
> 20.64 ~ 0% +23.9% 25.57 ~ 0% TOTAL turbostat.%c1
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 214235 ~ 0% -13.4% 185581 ~ 0% lkp-a03/micro/will-it-scale/sched_yield
> 214235 ~ 0% -13.4% 185581 ~ 0% TOTAL interrupts.LOC
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 133373 ~ 0% -10.8% 119020 ~ 0% lkp-a03/micro/will-it-scale/sched_yield
> 133373 ~ 0% -10.8% 119020 ~ 0% TOTAL softirqs.TIMER
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 1109 ~ 8% +9292.8% 104250 ~ 1% lkp-a03/micro/will-it-scale/sched_yield
> 4060 ~38% +1.2e+05% 4868923 ~ 2% lkp-snb01/micro/will-it-scale/sched_yield
> 5169 ~32% +96094.8% 4973173 ~ 2% TOTAL vmstat.system.cs
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 2045 ~ 1% +221.7% 6578 ~ 2% lkp-snb01/micro/will-it-scale/pthread_mutex2
> 2045 ~ 1% +221.7% 6578 ~ 2% TOTAL time.involuntary_context_switches
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 800 ~ 0% -18.8% 650 ~ 0% lkp-a03/micro/will-it-scale/sched_yield
> 800 ~ 0% -18.8% 650 ~ 0% TOTAL vmstat.system.in
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 50.56 ~ 0% -1.8% 49.64 ~ 0% lkp-snb01/micro/will-it-scale/pthread_mutex2
> 50.55 ~ 0% -9.4% 45.80 ~ 0% lkp-snb01/micro/will-it-scale/sched_yield
> 101.11 ~ 0% -5.6% 95.44 ~ 0% TOTAL turbostat.%c0
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 1518 ~ 0% -4.9% 1443 ~ 0% lkp-snb01/micro/will-it-scale/pthread_mutex2
> 1518 ~ 0% -4.9% 1443 ~ 0% TOTAL time.user_time
>
> v3.14-rc2 d2f8e017fdd913fc150160508
> --------------- -------------------------
> 488 ~ 0% -4.8% 465 ~ 0% lkp-snb01/micro/will-it-scale/pthread_mutex2
> 488 ~ 0% -4.8% 465 ~ 0% TOTAL time.percent_of_cpu_this_job_got
>
> Thanks,
> Fengguang
>
--
Thanks
Alex
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists