2021-03-08 08:47:40

by kernel test robot

[permalink] [raw]
Subject: [sched/fair] 9fe1f127b9: adrestia.wakeup_cost_periodic_us 8.1% regression


Greeting,

FYI, we noticed a 8.1% regression of adrestia.wakeup_cost_periodic_us due to commit:


commit: 9fe1f127b913318c631d0041ecf71486e38c2c2d ("sched/fair: Merge select_idle_core/cpu()")
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master


in testcase: adrestia
on test machine: 256 threads Intel(R) Genuine Intel(R) CPU 0000 @ 1.30GHz with 112G memory
with following parameters:

num_threads: 100
cpufreq_governor: performance
ucode: 0xffff0190

test-description: A micro-benchmark to test scheduler load balancer
test-url: https://github.com/mfleming/adrestia



If you fix the issue, kindly add following tag
Reported-by: kernel test robot <[email protected]>


Details are as below:
-------------------------------------------------------------------------------------------------->


To reproduce:

git clone https://github.com/intel/lkp-tests.git
cd lkp-tests
bin/lkp install job.yaml # job file is attached in this email
bin/lkp split-job --compatible job.yaml
bin/lkp run compatible-job.yaml

=========================================================================================
compiler/cpufreq_governor/kconfig/num_threads/rootfs/tbox_group/testcase/ucode:
gcc-9/performance/x86_64-rhel-8.3/100/debian-10.4-x86_64-20200603.cgz/lkp-knl-f1/adrestia/0xffff0190

commit:
6cd56ef1df ("sched/fair: Remove select_idle_smt()")
9fe1f127b9 ("sched/fair: Merge select_idle_core/cpu()")

6cd56ef1df399a00 9fe1f127b913318c631d0041ecf
---------------- ---------------------------
%stddev %change %stddev
\ | \
0.64 ? 9% -0.3 0.38 ? 6% perf-profile.children.cycles-pp.run_posix_cpu_timers
0.31 ? 25% -0.1 0.21 ? 5% perf-profile.children.cycles-pp.rb_erase
0.14 ? 8% -0.1 0.06 ? 11% perf-profile.children.cycles-pp.perf_pmu_disable
0.23 ? 9% -0.1 0.17 ? 13% perf-profile.children.cycles-pp.profile_tick
0.23 ? 7% -0.1 0.17 ? 6% perf-profile.children.cycles-pp.cpuidle_not_available
0.64 ? 9% -0.3 0.38 ? 6% perf-profile.self.cycles-pp.run_posix_cpu_timers
0.29 ? 26% -0.1 0.20 ? 6% perf-profile.self.cycles-pp.rb_erase
0.14 ? 11% -0.1 0.05 ? 46% perf-profile.self.cycles-pp.perf_pmu_disable
0.23 ? 9% -0.1 0.17 ? 13% perf-profile.self.cycles-pp.profile_tick
0.22 ? 7% -0.1 0.16 ? 7% perf-profile.self.cycles-pp.cpuidle_not_available
1409 +2.4% 1444 adrestia.time.elapsed_time
1409 +2.4% 1444 adrestia.time.elapsed_time.max
2533 ? 8% +26.7% 3209 ? 5% adrestia.time.involuntary_context_switches
120.33 +1.2% 121.83 adrestia.time.percent_of_cpu_this_job_got
1425 +4.0% 1482 adrestia.time.system_time
24.83 +8.1% 26.83 adrestia.wakeup_cost_periodic_us
87301 -2.3% 85295 vmstat.system.cs
1168 ? 5% +12.5% 1314 ? 4% slabinfo.Acpi-Parse.active_objs
1168 ? 5% +12.5% 1314 ? 4% slabinfo.Acpi-Parse.num_objs
3281229 +2.2% 3352839 proc-vmstat.numa_hit
3281222 +2.2% 3352834 proc-vmstat.numa_local
3515688 +2.0% 3587516 proc-vmstat.pgalloc_normal
3922169 +2.3% 4012192 proc-vmstat.pgfault
3419014 +2.1% 3489343 proc-vmstat.pgfree
10845440 +2.3% 11090816 proc-vmstat.unevictable_pgs_scanned
959.20 ? 31% +104.1% 1957 ? 27% sched_debug.cfs_rq:/.MIN_vruntime.avg
54895 ? 37% +96.5% 107863 ? 28% sched_debug.cfs_rq:/.MIN_vruntime.max
7048 ? 34% +100.5% 14129 ? 27% sched_debug.cfs_rq:/.MIN_vruntime.stddev
959.20 ? 31% +104.1% 1957 ? 27% sched_debug.cfs_rq:/.max_vruntime.avg
54895 ? 37% +96.5% 107863 ? 28% sched_debug.cfs_rq:/.max_vruntime.max
7048 ? 34% +100.5% 14129 ? 27% sched_debug.cfs_rq:/.max_vruntime.stddev
10.80 ? 7% +0.6 11.40 perf-stat.i.cache-miss-rate%
87364 -2.3% 85340 perf-stat.i.context-switches
500.05 ? 6% -13.7% 431.48 ? 3% perf-stat.i.cpu-migrations
0.36 -2.3% 0.36 perf-stat.i.metric.K/sec
10.52 ? 7% +0.6 11.11 perf-stat.overall.cache-miss-rate%
2.33 ? 2% +0.1 2.40 ? 3% perf-stat.overall.iTLB-load-miss-rate%
87319 -2.3% 85301 perf-stat.ps.context-switches
499.70 ? 6% -13.7% 431.29 ? 3% perf-stat.ps.cpu-migrations
993654 ? 8% +10.4% 1097144 interrupts.CAL:Function_call_interrupts
29716 ? 47% +67.8% 49867 ? 21% interrupts.CPU10.RES:Rescheduling_interrupts
7417 ? 42% -51.9% 3568 ? 31% interrupts.CPU102.CAL:Function_call_interrupts
476.17 ? 38% -55.9% 210.00 ? 2% interrupts.CPU102.NMI:Non-maskable_interrupts
476.17 ? 38% -55.9% 210.00 ? 2% interrupts.CPU102.PMI:Performance_monitoring_interrupts
5236 ? 16% +79.4% 9393 ? 55% interrupts.CPU120.CAL:Function_call_interrupts
93.67 ? 39% -51.6% 45.33 ? 32% interrupts.CPU124.TLB:TLB_shootdowns
1963 ? 15% +44.6% 2840 ? 16% interrupts.CPU125.CAL:Function_call_interrupts
11320 ? 26% +152.8% 28624 ? 41% interrupts.CPU14.RES:Rescheduling_interrupts
16173 ? 30% +196.6% 47975 ? 39% interrupts.CPU16.RES:Rescheduling_interrupts
44.00 ?113% -70.1% 13.17 ? 17% interrupts.CPU165.TLB:TLB_shootdowns
486.83 ? 52% -49.6% 245.33 ? 30% interrupts.CPU166.NMI:Non-maskable_interrupts
486.83 ? 52% -49.6% 245.33 ? 30% interrupts.CPU166.PMI:Performance_monitoring_interrupts
2013 ? 59% +270.0% 7450 ? 52% interrupts.CPU185.RES:Rescheduling_interrupts
53.67 ? 10% +122.7% 119.50 ? 25% interrupts.CPU190.TLB:TLB_shootdowns
5289 ? 13% +56.4% 8274 ? 16% interrupts.CPU20.CAL:Function_call_interrupts
13005 ? 33% +217.5% 41297 ? 39% interrupts.CPU20.RES:Rescheduling_interrupts
2309 ? 20% -33.1% 1545 ? 8% interrupts.CPU206.CAL:Function_call_interrupts
5633 ? 50% -82.1% 1007 ?111% interrupts.CPU206.RES:Rescheduling_interrupts
7619 ? 53% -89.7% 788.33 ? 45% interrupts.CPU208.RES:Rescheduling_interrupts
2186 ? 49% -71.5% 623.33 ?102% interrupts.CPU209.RES:Rescheduling_interrupts
5147 ?151% -72.4% 1419 ? 3% interrupts.CPU210.CAL:Function_call_interrupts
12.83 ? 20% +674.0% 99.33 ?169% interrupts.CPU210.TLB:TLB_shootdowns
4747 ? 79% -66.6% 1586 ? 12% interrupts.CPU218.CAL:Function_call_interrupts
13195 ? 52% +112.4% 28025 ? 28% interrupts.CPU22.RES:Rescheduling_interrupts
2276 ? 31% -32.6% 1533 ? 9% interrupts.CPU224.CAL:Function_call_interrupts
7091 ? 97% -80.0% 1421 ? 68% interrupts.CPU224.RES:Rescheduling_interrupts
11595 ? 41% +123.2% 25884 ? 45% interrupts.CPU23.RES:Rescheduling_interrupts
505.17 ? 39% -52.1% 242.00 ? 32% interrupts.CPU230.NMI:Non-maskable_interrupts
505.17 ? 39% -52.1% 242.00 ? 32% interrupts.CPU230.PMI:Performance_monitoring_interrupts
10032 ? 66% -75.9% 2417 ?105% interrupts.CPU232.RES:Rescheduling_interrupts
1805 ? 2% -18.9% 1463 ? 4% interrupts.CPU233.CAL:Function_call_interrupts
3665 ? 79% -57.3% 1563 ? 13% interrupts.CPU234.CAL:Function_call_interrupts
9647 ? 75% -80.7% 1862 ? 91% interrupts.CPU234.RES:Rescheduling_interrupts
9675 ? 60% -81.3% 1811 ?129% interrupts.CPU238.RES:Rescheduling_interrupts
6183 ? 85% -70.6% 1816 ? 18% interrupts.CPU240.CAL:Function_call_interrupts
13421 ? 42% -74.3% 3448 ? 80% interrupts.CPU240.RES:Rescheduling_interrupts
1750 ? 15% -16.1% 1467 ? 4% interrupts.CPU245.CAL:Function_call_interrupts
325.17 ? 30% -35.7% 209.00 ? 29% interrupts.CPU25.NMI:Non-maskable_interrupts
325.17 ? 30% -35.7% 209.00 ? 29% interrupts.CPU25.PMI:Performance_monitoring_interrupts
11691 ? 32% -94.3% 666.83 ? 45% interrupts.CPU250.RES:Rescheduling_interrupts
3546 ? 84% -52.1% 1699 ? 15% interrupts.CPU252.CAL:Function_call_interrupts
321.83 ? 32% -46.3% 172.83 interrupts.CPU252.NMI:Non-maskable_interrupts
321.83 ? 32% -46.3% 172.83 interrupts.CPU252.PMI:Performance_monitoring_interrupts
5324 ? 38% -74.5% 1357 ? 67% interrupts.CPU252.RES:Rescheduling_interrupts
16026 ? 38% +122.8% 35706 ? 37% interrupts.CPU26.RES:Rescheduling_interrupts
17753 ? 32% +57.7% 27993 ? 20% interrupts.CPU3.RES:Rescheduling_interrupts
6198 ? 32% +67.7% 10393 ? 39% interrupts.CPU30.CAL:Function_call_interrupts
14613 ? 36% +154.4% 37173 ? 33% interrupts.CPU30.RES:Rescheduling_interrupts
279.17 ? 20% +26.4% 353.00 ? 9% interrupts.CPU32.TLB:TLB_shootdowns
17828 ? 44% +121.0% 39400 ? 47% interrupts.CPU36.RES:Rescheduling_interrupts
15987 ? 24% +80.8% 28899 ? 28% interrupts.CPU38.RES:Rescheduling_interrupts
22415 ? 28% +60.9% 36065 ? 28% interrupts.CPU42.RES:Rescheduling_interrupts
5940 ? 16% +83.8% 10918 ? 19% interrupts.CPU46.CAL:Function_call_interrupts
24275 ? 29% +64.1% 39834 ? 21% interrupts.CPU46.RES:Rescheduling_interrupts
22740 ? 24% +67.3% 38035 ? 13% interrupts.CPU57.RES:Rescheduling_interrupts
17326 ? 13% +46.3% 25341 ? 13% interrupts.CPU63.RES:Rescheduling_interrupts
11967 ? 25% +95.6% 23414 ? 33% interrupts.CPU64.RES:Rescheduling_interrupts
2214 ? 16% +42.7% 3159 ? 20% interrupts.CPU67.CAL:Function_call_interrupts
7823 ? 32% +96.7% 15390 ? 34% interrupts.CPU67.RES:Rescheduling_interrupts
6129 ? 62% +139.3% 14671 ? 37% interrupts.CPU69.RES:Rescheduling_interrupts
30081 ? 32% +67.9% 50505 ? 28% interrupts.CPU74.RES:Rescheduling_interrupts
16041 ? 43% +129.9% 36885 ? 29% interrupts.CPU8.RES:Rescheduling_interrupts
5641 ? 46% +75.0% 9875 ? 41% interrupts.CPU80.CAL:Function_call_interrupts
428.83 ? 49% -44.6% 237.50 ? 31% interrupts.CPU81.NMI:Non-maskable_interrupts
428.83 ? 49% -44.6% 237.50 ? 31% interrupts.CPU81.PMI:Performance_monitoring_interrupts
6573 ? 33% +164.8% 17408 ? 39% interrupts.CPU81.RES:Rescheduling_interrupts
2992007 ? 10% +37.4% 4111178 ? 4% interrupts.RES:Rescheduling_interrupts



adrestia.wakeup_cost_periodic_us

28 +--------------------------------------------------------------------+
| |
27.5 |-+ |
27 |-O O O O O O O O O O O O O O O O O O O |
| |
26.5 |-+ |
| |
26 |-+ O O O O O O O O O O |
| |
25.5 |-+ |
25 |.+.+.+.+.+.+.+ + +.+.+.+.+.+..+.+.+.+ +.+.+.+.+.+.+ +.+ |
| : :: : : : : : |
24.5 |-+ : : : : : : : : |
| :: :: :: :: |
24 +--------------------------------------------------------------------+


[*] bisect-good sample
[O] bisect-bad sample



Disclaimer:
Results have been estimated based on internal Intel analysis and are provided
for informational purposes only. Any difference in system hardware or software
design or configuration may affect actual performance.


---
0DAY/LKP+ Test Infrastructure Open Source Technology Center
https://lists.01.org/hyperkitty/list/[email protected] Intel Corporation

Thanks,
Oliver Sang


Attachments:
(No filename) (13.05 kB)
config-5.11.0-00019-g9fe1f127b913 (175.11 kB)
job-script (7.61 kB)
job.yaml (5.21 kB)
reproduce (308.00 B)
Download all attachments