reaim.time.voluntary_context_switches +110.7% Reply-To: kernel test robot User-Agent: Heirloom mailx 12.5 6/20/10 FYI, we noticed reaim.time.voluntary_context_switches +110.7% due to commit: commit d59dc7bcfa649ef2128a76b6487b16f4b3f14d23 ("sched/numa, mm: Revert to checking pmd/pte_write instead of VMA flags") https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git master in testcase: reaim on test machine: 72 threads Intel(R) Xeon(R) CPU E5-2699 v3 @ 2.30GHz with 128G memory with following parameters: runtime: 300s nr_task: 100% nr_job: 10000 test: short cpufreq_governor: performance REAIM is an updated and improved version of AIM 7 benchmark. 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. Details are as below: --------------------------------------------------------------------------------------------------> To reproduce: git clone git://git.kernel.org/pub/scm/linux/kernel/git/wfg/lkp-tests.git cd lkp-tests bin/lkp install job.yaml # job file is attached in this email bin/lkp run job.yaml ========================================================================================= compiler/cpufreq_governor/kconfig/nr_job/nr_task/rootfs/runtime/tbox_group/test/testcase: gcc-6/performance/x86_64-rhel-7.2/10000/100%/debian-x86_64-2016-08-31.cgz/300s/lkp-hsw-ep2/short/reaim commit: b95202a3b6 ("sched/deadline: Document behavior of sched_yield()") d59dc7bcfa ("sched/numa, mm: Revert to checking pmd/pte_write instead of VMA flags") b95202a3b6bb8715 d59dc7bcfa649ef2128a76b648 ---------------- -------------------------- fail:runs %reproduction fail:runs | | | %stddev %change %stddev \ | \ 2.06 ± 1% -2.7% 2.00 ± 0% reaim.std_dev_percent 328520 ± 6% +110.7% 692172 ± 2% reaim.time.voluntary_context_switches 1078864 ± 1% +332.2% 4662990 ± 1% interrupts.CAL:Function_call_interrupts 20354 ± 7% +37.5% 27995 ± 6% meminfo.Shmem 186030 ± 2% +22.9% 228590 ± 1% softirqs.SCHED 6511 ± 1% +36.6% 8893 ± 1% vmstat.system.cs 77823 ± 0% +16.6% 90739 ± 0% vmstat.system.in 3386 ± 6% -13.6% 2925 ± 4% numa-meminfo.node0.PageTables 819.33 ± 39% +516.0% 5047 ± 81% numa-meminfo.node1.Inactive(anon) 7767 ± 0% +32.0% 10249 ± 23% numa-meminfo.node1.Mapped 2853 ± 8% +16.9% 3336 ± 4% numa-meminfo.node1.PageTables 849.00 ± 6% -13.8% 732.25 ± 4% numa-vmstat.node0.nr_page_table_pages 204.33 ± 39% +517.3% 1261 ± 82% numa-vmstat.node1.nr_inactive_anon 715.33 ± 9% +16.4% 832.75 ± 4% numa-vmstat.node1.nr_page_table_pages 204.33 ± 39% +517.3% 1261 ± 82% numa-vmstat.node1.nr_zone_inactive_anon 1.58 ± 3% +11.1% 1.76 ± 4% perf-profile.children.cycles-pp.__do_page_fault 1.61 ± 2% +11.6% 1.80 ± 5% perf-profile.children.cycles-pp.do_page_fault 1.21 ± 3% +14.6% 1.39 ± 5% perf-profile.children.cycles-pp.handle_mm_fault 1.62 ± 2% +11.6% 1.81 ± 4% perf-profile.children.cycles-pp.page_fault 18555610 ± 5% +142.8% 45052001 ± 2% cpuidle.C1-HSW.time 284802 ± 5% +148.3% 707062 ± 2% cpuidle.C1-HSW.usage 10511856 ± 4% +100.4% 21062895 ± 8% cpuidle.C1E-HSW.time 84157 ± 2% +152.3% 212324 ± 1% cpuidle.C1E-HSW.usage 5259 ± 3% +197.5% 15647 ± 0% cpuidle.POLL.usage 125636 ± 0% +285.1% 483761 ± 1% latency_stats.hits.wait_on_page_bit.__migration_entry_wait.migration_entry_wait.do_swap_page.handle_mm_fault.__do_page_fault.do_page_fault.page_fault 10887 ± 85% -98.7% 141.25 ±109% latency_stats.max.call_rwsem_down_write_failed.unlink_file_vma.free_pgtables.unmap_region.do_munmap.SyS_munmap.do_syscall_64.return_from_SYSCALL_64 2007 ± 64% +281.0% 7649 ± 37% latency_stats.max.wait_on_page_bit.__migration_entry_wait.migration_entry_wait.do_swap_page.handle_mm_fault.__do_page_fault.do_page_fault.page_fault 10975 ± 85% -97.9% 234.75 ±102% latency_stats.sum.call_rwsem_down_write_failed.unlink_file_vma.free_pgtables.unmap_region.do_munmap.SyS_munmap.do_syscall_64.return_from_SYSCALL_64 13269626 ± 0% +293.0% 52145660 ± 1% latency_stats.sum.max 13202213 ± 1% +295.0% 52145660 ± 1% latency_stats.sum.wait_on_page_bit.__migration_entry_wait.migration_entry_wait.do_swap_page.handle_mm_fault.__do_page_fault.do_page_fault.page_fault 1970029 ± 2% +37.0% 2699350 ± 2% perf-stat.context-switches 83062 ± 3% +11.2% 92391 ± 2% perf-stat.cpu-migrations 0.01 ± 0% +5.1% 0.01 ± 3% perf-stat.dTLB-store-miss-rate% 4.265e+09 ± 1% +1.9% 4.345e+09 ± 0% perf-stat.iTLB-loads 2.228e+09 ± 0% +1.9% 2.271e+09 ± 0% perf-stat.node-load-misses 1.113e+09 ± 0% +1.3% 1.127e+09 ± 0% perf-stat.node-store-misses 4.582e+08 ± 1% +1.9% 4.669e+08 ± 0% perf-stat.node-stores 5092 ± 7% +37.3% 6991 ± 6% proc-vmstat.nr_shmem 435033 ± 0% -15.8% 366135 ± 0% proc-vmstat.numa_hint_faults 237335 ± 0% -13.7% 204706 ± 0% proc-vmstat.numa_hint_faults_local 41453 ± 2% +232.2% 137724 ± 1% proc-vmstat.numa_pages_migrated 7184 ± 9% +48.6% 10678 ± 3% proc-vmstat.pgactivate 1193 ± 2% +419.5% 6197 ± 1% proc-vmstat.pgmigrate_fail 41453 ± 2% +232.2% 137724 ± 1% proc-vmstat.pgmigrate_success 0.49 ± 8% -16.1% 0.41 ± 13% sched_debug.cfs_rq:/.nr_running.avg 6.79 ± 7% -19.5% 5.47 ± 12% sched_debug.cfs_rq:/.runnable_load_avg.avg 649539 ± 7% +9.4% 710821 ± 7% sched_debug.cpu.avg_idle.avg 6.62 ± 9% -19.9% 5.31 ± 13% sched_debug.cpu.cpu_load[0].avg 7.18 ± 9% -20.7% 5.70 ± 11% sched_debug.cpu.cpu_load[1].avg 7.10 ± 9% -20.6% 5.64 ± 12% sched_debug.cpu.cpu_load[2].avg 7.01 ± 8% -19.8% 5.62 ± 12% sched_debug.cpu.cpu_load[3].avg 6.96 ± 6% -19.2% 5.62 ± 12% sched_debug.cpu.cpu_load[4].avg 16907 ± 3% -17.0% 14029 ± 16% sched_debug.cpu.load.avg 460630 ± 22% -36.1% 294449 ± 35% sched_debug.cpu.load.max 0.52 ± 9% -18.4% 0.43 ± 14% sched_debug.cpu.nr_running.avg 14012 ± 2% +35.2% 18940 ± 2% sched_debug.cpu.nr_switches.avg 37653 ± 3% +13.4% 42694 ± 6% sched_debug.cpu.nr_switches.max 8357 ± 4% +44.7% 12093 ± 3% sched_debug.cpu.nr_switches.min 0.00 ±353% +650.0% 0.01 ± 67% sched_debug.cpu.nr_uninterruptible.avg 35378 ± 0% +14.0% 40319 ± 0% sched_debug.cpu.sched_count.avg 29104 ± 1% +12.7% 32810 ± 1% sched_debug.cpu.sched_count.min 3146 ± 4% +77.0% 5568 ± 2% sched_debug.cpu.sched_goidle.avg 1931 ± 4% +114.4% 4141 ± 2% sched_debug.cpu.sched_goidle.min 40729 ± 2% +9.8% 44726 ± 4% sched_debug.cpu.ttwu_count.max 0.02 ± 3% +8.0% 0.02 ± 2% sched_debug.rt_rq:/.rt_time.avg perf-stat.context-switches 3e+06 ++----------------------------------------------------------------+ | O O O O O O O O O | 2.5e+06 O+ O O O O | | | | .*. .*..*. .*.. .*..*..*.*..* *.. .*.. | 2e+06 *+. .*..*.*. *. *. * : : * * *..* | *.*. : : : : | 1.5e+06 ++ : : : : | | : : : : | 1e+06 ++ : : : : | | : : : : | | : : : : | 500000 ++ :: :: | | : : | 0 ++-O--------------------------------------------*------------*----+ reaim.time.voluntary_context_switches 800000 ++-----------------------------------------------------------------+ | O O | 700000 O+ O O O O O O O O O O O | | | 600000 ++ | 500000 ++ | | | 400000 ++ .*. | | *..*.. .*..*.. .*.. .*. *..*..* *..*.*..* .* 300000 ++ + * * *. : : : *. | 200000 *+.*.*.. .* : : : : | | *. : : : : | 100000 ++ : : : : | | :: : | 0 ++-O--------------------------------------------*-------------*----+ interrupts.CAL:Function_call_interrupts 5e+06 ++----------------------------------------------------------------+ 4.5e+06 O+ O O O O O O O O O O O O O | | | 4e+06 ++ | 3.5e+06 ++ | | | 3e+06 ++ | 2.5e+06 ++ | 2e+06 ++ | | | 1.5e+06 ++ | 1e+06 *+.*.*..*..*.*..*.*..*..*.*..*..*.*..*..*.*..* *..*.*..* *..* | + : + : | 500000 ++ + : + : | 0 ++-O--------------------------------------------*------------*----+ vmstat.system.in 100000 ++-----------------------------------------------------------------+ 90000 O+ O O O O O O O O O O O O O | | | 80000 *+.*.*..*..*.*..*..*.*..*..*.*..*..*..*.*..*..* *..*.*..* *..* 70000 ++ : : : : | | : : : : | 60000 ++ : : : : | 50000 ++ : : : : | 40000 ++ : : : : | | : : : : | 30000 ++ : : : : | 20000 ++ : : : : | | :: :: | 10000 ++ : : | 0 ++-O--------------------------------------------*-------------*----+ vmstat.system.cs 10000 ++------------------------------------------------------------------+ 9000 ++ O O O O O | O O O O O O O O O | 8000 ++ | 7000 ++ .*.*.. .*..*..*.* *. | *.. .*..*..*.*..*. *..*..*.*. : : *..*..* *..* 6000 ++ *.*. : : : : | 5000 ++ : : : : | 4000 ++ : : : : | | : : : : | 3000 ++ : : : : | 2000 ++ : : : : | | : : :: | 1000 ++ : : | 0 ++-O---------------------------------------------*-------------*----+ [*] bisect-good sample [O] bisect-bad sample Thanks, Xiaolong