From mboxrd@z Thu Jan 1 00:00:00 1970 Content-Type: multipart/mixed; boundary="===============4008738076651830146==" MIME-Version: 1.0 From: Ye Xiaolong To: lkp@lists.01.org Subject: Re: [lkp-robot] [irq/affinity] 13c024422c: fsmark.files_per_sec -4.3% regression Date: Thu, 13 Apr 2017 09:06:04 +0800 Message-ID: <20170413010604.GH31394@yexl-desktop> In-Reply-To: List-Id: --===============4008738076651830146== Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable On 04/12, Busch, Keith wrote: >Hi Xiaolong, > >Thanks for letting me know of the regression. This change wasn't supposed = to affect anything outside the unbalanced numa case, so if it's indeed the = culprit, I'll need to fix that. > >>From the machine that demonstrates the regression, could you send me 'lscp= u'? And the output from the following commands, both with and without the p= atch applied? Okay, when should I run the `lscpu` and below cmds, before, after or during= the test? Thanks, Xiaolong > ># cat /proc/interrupts ># for i in $(grep nvme0q /proc/interrupts | cut -d":" -f1 | sed "s/ //g")= ; do echo "irq $i smp affinity: `cat /proc/irq/$i/smp_affinity`"; done > > >-----Original Message----- >From: lkp-robot-request(a)eclists.intel.com [mailto:lkp-robot-request(a)ec= lists.intel.com] On Behalf Of Ye, Xiaolong >Sent: Tuesday, April 11, 2017 7:33 PM >To: Busch, Keith >Cc: Thomas Gleixner ; linux-kernel(a)vger.kernel.org= ; linux-nvme(a)lists.infradead.org; Christoph Hellwig ; lkp(a= )01.org >Subject: [lkp-robot] [irq/affinity] 13c024422c: fsmark.files_per_sec -4.3%= regression > > >Greeting, > >FYI, we noticed a -4.3% regression of fsmark.files_per_sec due to commit: > > >commit: 13c024422cbb6dcc513667be9a2613b0f0de781a ("irq/affinity: Assign al= l CPUs a vector") >url: https://github.com/0day-ci/linux/commits/Keith-Busch/irq-affinity-Ass= ign-all-CPUs-a-vector/20170401-035036 > > >in testcase: fsmark >on test machine: 72 threads Intel(R) Xeon(R) CPU E5-2699 v3 @ 2.30GHz with= 128G memory >with following parameters: > > iterations: 8 > disk: 1SSD > nr_threads: 4 > fs: btrfs > filesize: 9B > test_size: 16G > sync_method: fsyncBeforeClose > nr_directories: 16d > nr_files_per_directory: 256fpd > cpufreq_governor: performance > >test-description: The fsmark is a file system benchmark to test synchronou= s write workloads, for example, mail servers workload. >test-url: https://sourceforge.net/projects/fsmark/ > > >Details are as below: >--------------------------------------------------------------------------= ------------------------> > > >To reproduce: > > git clone https://github.com/01org/lkp-tests.git > cd lkp-tests > bin/lkp install job.yaml # job file is attached in this email > bin/lkp run job.yaml > >testcase/path_params/tbox_group/run: fsmark/8-1SSD-4-btrfs-9B-16G-fsyncBef= oreClose-16d-256fpd-performance/lkp-hsw-ep4 > >45e5202213ae6541 13c024422cbb6dcc513667be9a = >---------------- -------------------------- = > %stddev change %stddev > \ | \ = > 13888 -4% 13293 fsmark.files_per_sec > 302 4% 315 fsmark.time.elapsed_time > 302 4% 315 fsmark.time.elapsed_time.max > 2.452e+08 3% 2.528e+08 fsmark.time.file_system_outp= uts > 20265209 19926430 fsmark.time.voluntary_contex= t_switches > 206 -4% 198 fsmark.time.percent_of_cpu_t= his_job_got > 707811 -65% 246894 fsmark.time.involuntary_cont= ext_switches > 383249 =C2=B1 13% -36% 246054 interrupts.CAL:Function= _call_interrupts > 149418 -5% 141286 vmstat.system.cs > 190 -7% 178 turbostat.Avg_MHz > 6.25 -5% 5.93 turbostat.%Busy > 8994 =C2=B1 3% 199% 26910 =C2=B1 13% perf-stat.cpu-migr= ations > 45475990 44940538 perf-stat.context-switches > 0.42 12% 0.48 perf-stat.dTLB-load-miss-rat= e% > 0.09 10% 0.09 perf-stat.dTLB-store-miss-ra= te% > 2.162e+08 7% 2.324e+08 perf-stat.dTLB-store-misses > 674521 4% 703995 perf-stat.page-faults > 674521 4% 703995 perf-stat.minor-faults > 1.46 4% 1.52 perf-stat.branch-miss-rate% > 1.445e+09 3% 1.494e+09 perf-stat.iTLB-loads > > > > fsmark.time.elapsed_time > > 400 ++------------------------------------------------------------------= --+ > | O = | > 350 ++ = | > 300 O+O.O.O O.O.O.*.O..O.O.O O O.O.O.O O O.O.O.O.O.O.O.*.. .*.*.*.*. .*.= *.* > | *.* * : * : * * = | > 250 ++ : : : : = | > | : : : : = | > 200 ++ : : : : = | > | : : : : = | > 150 ++ : : : : = | > 100 ++ : : : : = | > | : : : : = | > 50 ++ : : = | > | : : = | > 0 ++-----------------------*---------*--------------------------------= --+ > > > fsmark.time.elapsed_time.max > > 400 ++------------------------------------------------------------------= --+ > | O = | > 350 ++ = | > 300 O+O.O.O O.O.O.*.O..O.O.O O O.O.O.O O O.O.O.O.O.O.O.*.. .*.*.*.*. .*.= *.* > | *.* * : * : * * = | > 250 ++ : : : : = | > | : : : : = | > 200 ++ : : : : = | > | : : : : = | > 150 ++ : : : : = | > 100 ++ : : : : = | > | : : : : = | > 50 ++ : : = | > | : : = | > 0 ++-----------------------*---------*--------------------------------= --+ > > > fsmark.time.involuntary_context_switches > > 800000 ++---------------------------------------------------------------= --+ > *.*.*.*.*. .*. .* .*.* .*. .*. .*. = .* > 700000 ++ * *.*.*.* : *.* : *.* *.*.*.*.*.*.*.* * *.= * | > | : : : : = | > 600000 ++ : : : : = | > 500000 ++ : : : : = | > | : : : : = | > 400000 ++ : : : : = | > | O : : : : = | > 300000 ++ : : :: = | > 200000 O+O O O O O O O O O O:O:O O O O:OO O O O O O O = | > | :: : = | > 100000 ++ : : = | > | : : = | > 0 ++----------------------*---------*------------------------------= --+ > > > fsmark.files_per_sec > > 16000 ++----------------------------------------------------------------= --+ > | = | > 14000 O+*.O.O.O.O.O.*.O.O.O.O O O.O.O.O O O.O.O.O.O.O.O.*.*.*.*.*.*.*.*.= *.* > 12000 ++O : : : : = | > | O : : : : = | > 10000 ++ : : : : = | > | : : : : = | > 8000 ++ : : : : = | > | : : : : = | > 6000 ++ : : : : = | > 4000 ++ : : : : = | > | :: :: = | > 2000 ++ : : = | > | : : = | > 0 ++----------------------*---------*-------------------------------= --+ > > [*] bisect-good sample > [O] bisect-bad sample > > >Disclaimer: >Results have been estimated based on internal Intel analysis and are provi= ded >for informational purposes only. Any difference in system hardware or soft= ware >design or configuration may affect actual performance. > > >Thanks, >Xiaolong --===============4008738076651830146==--