Greeting,
FYI, we noticed a 5.6% improvement of filebench.sum_operations/s due to commit:
commit: 260916b537317f1215edd2d94be48b86a1efcfa1 ("sched: Add support for lazy
preemption")
https://git.kernel.org/cgit/linux/kernel/git/rt/linux-rt-devel.git linux-5.14.y-rt-rebase
in testcase: filebench
on test machine: 272 threads 2 sockets Intel(R) Xeon Phi(TM) CPU 7255 @ 1.10GHz with 112G
memory
with following parameters:
disk: 1HDD
fs: f2fs
test: filemicro_writefsync.f
cpufreq_governor: performance
ucode: 0x11
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 # generate the yaml file for lkp run
bin/lkp run generated-yaml-file
=========================================================================================
compiler/cpufreq_governor/disk/fs/kconfig/rootfs/tbox_group/test/testcase/ucode:
gcc-9/performance/1HDD/f2fs/x86_64-rhel-8.3/debian-10.4-x86_64-20200603.cgz/lkp-knm02/filemicro_writefsync.f/filebench/0x11
commit:
e4d90d8a96 ("x86: Enable RT also on 32bit")
260916b537 ("sched: Add support for lazy preemption")
e4d90d8a96548e17 260916b537317f1215edd2d94be
---------------- ---------------------------
%stddev %change %stddev
\ | \
53.90 +5.6% 56.90 filebench.sum_bytes_mb/s
6904 +5.6% 7288 filebench.sum_operations/s
0.14 -2.3% 0.13 filebench.sum_time_ms/op
6898 +5.6% 7281 filebench.sum_writes/s
40413 +4.0% 42029 vmstat.io.bo
4010 ± 2% +3.8% 4161 vmstat.system.cs
119271 -4.8% 113529 proc-vmstat.nr_inactive_file
119271 -4.8% 113529 proc-vmstat.nr_zone_inactive_file
359912 -1.0% 356298 proc-vmstat.pgfault
352.20 +0.8% 355.11 perf-stat.i.cpu-migrations
13056 +3.0% 13452 perf-stat.i.minor-faults
13057 +3.0% 13452 perf-stat.i.page-faults
12404 +2.8% 12753 perf-stat.ps.minor-faults
12404 +2.8% 12753 perf-stat.ps.page-faults
0.33 ± 60% +99.9% 0.66 ± 11%
perf-sched.sch_delay.avg.ms.smpboot_thread_fn.kthread.ret_from_fork
76.41 ± 25% -48.3% 39.47 ± 38% perf-sched.total_wait_and_delay.average.ms
3504 ± 7% -12.2% 3077 ± 3% perf-sched.total_wait_and_delay.count.ms
76.05 ± 25% -48.9% 38.88 ± 38% perf-sched.total_wait_time.average.ms
201.91 ± 31% -56.8% 87.30 ± 60%
perf-sched.wait_and_delay.avg.ms.smpboot_thread_fn.kthread.ret_from_fork
1016 -25.0% 762.17 ± 7%
perf-sched.wait_and_delay.count.smpboot_thread_fn.kthread.ret_from_fork
305.49 ± 73% -85.7% 43.75 ±223%
perf-sched.wait_time.avg.ms.do_nanosleep.hrtimer_nanosleep.__x64_sys_nanosleep.do_syscall_64
17.35 ± 43% -44.1% 9.69
perf-sched.wait_time.avg.ms.schedule_hrtimeout_range_clock.poll_schedule_timeout.constprop.0.do_select
201.58 ± 31% -57.0% 86.64 ± 61%
perf-sched.wait_time.avg.ms.smpboot_thread_fn.kthread.ret_from_fork
7.92 ± 14% -2.1 5.78 ± 17%
perf-profile.calltrace.cycles-pp.__libc_start_main
7.92 ± 14% -2.1 5.78 ± 17%
perf-profile.calltrace.cycles-pp.main.__libc_start_main
7.92 ± 14% -2.1 5.78 ± 17%
perf-profile.calltrace.cycles-pp.run_builtin.main.__libc_start_main
5.85 ± 15% -1.6 4.28 ± 19%
perf-profile.calltrace.cycles-pp.cmd_record.cmd_sched.run_builtin.main.__libc_start_main
5.85 ± 15% -1.6 4.28 ± 19%
perf-profile.calltrace.cycles-pp.cmd_sched.run_builtin.main.__libc_start_main
5.85 ± 15% -1.6 4.28 ± 19%
perf-profile.calltrace.cycles-pp.__evlist__enable.cmd_record.cmd_sched.run_builtin.main
7.92 ± 14% -2.1 5.78 ± 17%
perf-profile.children.cycles-pp.__libc_start_main
7.92 ± 14% -2.1 5.78 ± 17% perf-profile.children.cycles-pp.main
7.92 ± 14% -2.1 5.78 ± 17% perf-profile.children.cycles-pp.run_builtin
7.92 ± 14% -2.1 5.78 ± 17% perf-profile.children.cycles-pp.cmd_record
5.85 ± 15% -1.6 4.28 ± 19% perf-profile.children.cycles-pp.cmd_sched
5.85 ± 15% -1.6 4.28 ± 19%
perf-profile.children.cycles-pp.__evlist__enable
1.05 ± 34% -0.6 0.44 ± 64%
perf-profile.children.cycles-pp.dequeue_task_fair
0.15 ±100% +0.3 0.48 ± 59%
perf-profile.children.cycles-pp.tlb_flush_mmu
0.39 ± 90% +0.4 0.77 ± 32%
perf-profile.children.cycles-pp.timerqueue_add
0.11 ±141% +0.4 0.54 ± 79%
perf-profile.children.cycles-pp.schedule_hrtimeout_range_clock
0.11 ±141% +0.5 0.57 ± 88%
perf-profile.children.cycles-pp.sd_event_run
0.17 ±107% +0.5 0.64 ± 43% perf-profile.children.cycles-pp.error_entry
0.17 ±107% +0.5 0.64 ± 43% perf-profile.self.cycles-pp.error_entry
0.20 ±111% +0.5 0.67 ± 53% perf-profile.self.cycles-pp.update_rq_clock
filebench.sum_operations_s
7300 +--------------------------------------------------------------------+
| |
7250 |-+ |
7200 |-+ |
| |
7150 |-+ |
| |
7100 |-+ |
| |
7050 |-+ |
7000 |-+ |
| |
6950 |-+ |
| |
6900 +--------------------------------------------------------------------+
[*] 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/lkp@lists.01.org Intel Corporation
Thanks,
Oliver Sang