FYI, we noticed a 10.6% improvement of hackbench.throughput due to commit: commit f54c5d4e28da93ffb92c40f84f6f6e6db41d652e ("sched/fair: Do not decay new task load on first enqueue") https://github.com/0day-ci/linux Matt-Fleming/sched-fair-Do-not-decay-new-task-load-on-first-enqueue/20160924-014929 in testcase: hackbench on test machine: 32 threads Intel(R) Xeon(R) CPU E5-2680 0 @ 2.70GHz with 32G memory with following parameters: cpuset.mems: 0-$((nr_node-1)) cpuset.cpus: 0-$((nr_cpu-1)) nr_threads: 1600% mode: threads ipc: pipe cpufreq_governor: performance Hackbench is both a benchmark and a stress test for the Linux kernel scheduler. 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/cpuset.cpus/cpuset.mems/ipc/kconfig/mode/nr_threads/rootfs/tbox_group/testcase: gcc-6/performance/0-$((nr_cpu-1))/0-$((nr_node-1))/pipe/x86_64-kexec/threads/1600%/debian-x86_64-2016-08-31.cgz/lkp-snb01/hackbench commit: c1fad9ef7e ("objtool: Add do_task_dead() to global noreturn list") f54c5d4e28 ("sched/fair: Do not decay new task load on first enqueue") c1fad9ef7ed14aad f54c5d4e28da93ffb92c40f84f ---------------- -------------------------- %stddev %change %stddev \ | \ 248098 ± 1% +10.6% 274415 ± 0% hackbench.throughput 8.706e+08 ± 1% +21.7% 1.06e+09 ± 0% hackbench.time.involuntary_context_switches 1674142 ± 1% +10.2% 1844981 ± 0% hackbench.time.minor_page_faults 1547 ± 1% +9.7% 1696 ± 0% hackbench.time.user_time 2.287e+09 ± 2% -11.8% 2.017e+09 ± 0% hackbench.time.voluntary_context_switches 5198633 ± 9% -24.8% 3908585 ± 17% interrupts.CAL:Function_call_interrupts 217794 ± 2% -14.7% 185716 ± 1% meminfo.KernelStack 105989 ± 8% -20.7% 84099 ± 14% numa-meminfo.node1.KernelStack 1237200 ± 0% -15.9% 1040658 ± 2% softirqs.RCU 540161 ± 4% -53.0% 254100 ± 0% softirqs.SCHED 1569 ± 2% +11.6% 1751 ± 2% vmstat.procs.r 290364 ± 7% -67.9% 93107 ± 1% vmstat.system.in 1.34 ± 2% +88.1% 2.52 ± 0% turbostat.CPU%c1 2.61 ± 4% -34.3% 1.71 ± 1% turbostat.CPU%c6 0.19 ± 14% +61.3% 0.30 ± 19% turbostat.Pkg%pc6 35103736 ± 5% -65.8% 11992857 ± 1% cpuidle.C1.usage 197809 ± 3% +189.3% 572238 ± 0% cpuidle.C2.usage 57732567 ± 10% -92.4% 4375330 ± 24% cpuidle.POLL.time 185806 ± 5% -66.1% 63025 ± 1% cpuidle.POLL.usage 17524006 ± 2% +11.3% 19495992 ± 2% numa-numastat.node0.local_node 17524039 ± 2% +11.3% 19496020 ± 2% numa-numastat.node0.numa_hit 15305415 ± 6% +25.0% 19128452 ± 2% numa-numastat.node1.local_node 15305434 ± 6% +25.0% 19128469 ± 2% numa-numastat.node1.numa_hit 8687240 ± 9% +15.3% 10019684 ± 3% numa-vmstat.node0.numa_hit 8687217 ± 9% +15.3% 10019664 ± 3% numa-vmstat.node0.numa_local 8239375 ± 3% +19.6% 9854917 ± 2% numa-vmstat.node1.numa_hit 8239361 ± 3% +19.6% 9854904 ± 2% numa-vmstat.node1.numa_local 217952 ± 1% -14.9% 185402 ± 0% proc-vmstat.nr_kernel_stack 32832102 ± 2% +17.6% 38626359 ± 0% proc-vmstat.numa_hit 32832049 ± 2% +17.6% 38626314 ± 0% proc-vmstat.numa_local 36089336 ± 2% +16.9% 42180382 ± 0% proc-vmstat.pgalloc_normal 2547440 ± 1% -11.6% 2252216 ± 0% proc-vmstat.pgfault 36004406 ± 2% +16.9% 42080130 ± 0% proc-vmstat.pgfree 741.00 ± 11% -28.9% 526.50 ± 11% slabinfo.bdev_cache.active_objs 741.00 ± 11% -28.9% 526.50 ± 11% slabinfo.bdev_cache.num_objs 858.00 ± 10% -28.5% 613.25 ± 10% slabinfo.file_lock_cache.active_objs 858.00 ± 10% -28.5% 613.25 ± 10% slabinfo.file_lock_cache.num_objs 23908 ± 2% -8.5% 21874 ± 4% slabinfo.kmalloc-192.active_objs 24087 ± 2% -8.5% 22049 ± 4% slabinfo.kmalloc-192.num_objs 33276 ± 3% -11.0% 29622 ± 3% slabinfo.kmalloc-256.active_objs 33566 ± 3% -10.8% 29930 ± 3% slabinfo.kmalloc-256.num_objs 14696 ± 2% -14.8% 12527 ± 1% slabinfo.task_struct.active_objs 1724 ± 2% -11.9% 1518 ± 1% slabinfo.task_struct.active_slabs 15523 ± 2% -11.9% 13674 ± 1% slabinfo.task_struct.num_objs 1724 ± 2% -11.9% 1518 ± 1% slabinfo.task_struct.num_slabs 60659 ± 0% -11.6% 53599 ± 1% slabinfo.vm_area_struct.active_objs 1381 ± 0% -11.6% 1220 ± 1% slabinfo.vm_area_struct.active_slabs 60797 ± 0% -11.6% 53735 ± 1% slabinfo.vm_area_struct.num_objs 1381 ± 0% -11.6% 1220 ± 1% slabinfo.vm_area_struct.num_slabs 8.002e+12 ± 0% +4.2% 8.338e+12 ± 0% perf-stat.branch-instructions 0.44 ± 2% -17.7% 0.36 ± 0% perf-stat.branch-miss-rate% 3.526e+10 ± 2% -14.2% 3.026e+10 ± 0% perf-stat.branch-misses 6.90 ± 5% +68.6% 11.64 ± 0% perf-stat.cache-miss-rate% 1.28e+10 ± 2% +30.1% 1.665e+10 ± 0% perf-stat.cache-misses 1.856e+11 ± 2% -22.9% 1.43e+11 ± 0% perf-stat.cache-references 85798578 ± 6% -59.5% 34720971 ± 3% perf-stat.cpu-migrations 0.41 ± 4% -52.4% 0.19 ± 14% perf-stat.dTLB-load-miss-rate% 5.162e+10 ± 4% -49.8% 2.592e+10 ± 14% perf-stat.dTLB-load-misses 1.263e+13 ± 0% +5.7% 1.335e+13 ± 0% perf-stat.dTLB-loads 0.23 ± 6% -76.3% 0.05 ± 5% perf-stat.dTLB-store-miss-rate% 2.014e+10 ± 5% -74.8% 5.067e+09 ± 5% perf-stat.dTLB-store-misses 8.666e+12 ± 1% +6.4% 9.219e+12 ± 0% perf-stat.dTLB-stores 37.07 ± 14% -21.5% 29.11 ± 10% perf-stat.iTLB-load-miss-rate% 5.813e+08 ± 25% -45.5% 3.166e+08 ± 12% perf-stat.iTLB-load-misses 9.649e+08 ± 2% -20.3% 7.691e+08 ± 5% perf-stat.iTLB-loads 3.905e+13 ± 0% +5.1% 4.105e+13 ± 0% perf-stat.instructions 71218 ± 22% +85.1% 131838 ± 12% perf-stat.instructions-per-iTLB-miss 0.67 ± 0% +5.3% 0.71 ± 0% perf-stat.ipc 2516529 ± 1% -11.1% 2236475 ± 0% perf-stat.minor-faults 43.75 ± 2% -19.7% 35.13 ± 0% perf-stat.node-load-miss-rate% 5.849e+09 ± 3% -11.0% 5.203e+09 ± 0% perf-stat.node-load-misses 7.515e+09 ± 1% +27.8% 9.607e+09 ± 0% perf-stat.node-loads 22.67 ± 0% +38.6% 31.42 ± 0% perf-stat.node-store-miss-rate% 1.665e+09 ± 1% +109.3% 3.485e+09 ± 0% perf-stat.node-store-misses 5.682e+09 ± 2% +33.9% 7.606e+09 ± 0% perf-stat.node-stores 2516529 ± 1% -11.1% 2236448 ± 0% perf-stat.page-faults 326241 ± 8% -47.5% 171439 ± 8% sched_debug.cfs_rq:/.MIN_vruntime.max 68925 ± 5% -68.6% 21676 ± 31% sched_debug.cfs_rq:/.MIN_vruntime.stddev 38373809 ± 8% -29.1% 27192438 ± 7% sched_debug.cfs_rq:/.load.avg 91907686 ± 14% -42.1% 53227633 ± 10% sched_debug.cfs_rq:/.load.max 27021348 ± 15% -50.8% 13305171 ± 7% sched_debug.cfs_rq:/.load.stddev 38188 ± 8% -23.2% 29335 ± 11% sched_debug.cfs_rq:/.load_avg.avg 80834 ± 14% -44.7% 44719 ± 12% sched_debug.cfs_rq:/.load_avg.max 22512 ± 17% -53.4% 10496 ± 14% sched_debug.cfs_rq:/.load_avg.stddev 326248 ± 8% -47.4% 171447 ± 8% sched_debug.cfs_rq:/.max_vruntime.max 68926 ± 5% -68.5% 21677 ± 31% sched_debug.cfs_rq:/.max_vruntime.stddev 327034 ± 8% -46.7% 174369 ± 5% sched_debug.cfs_rq:/.min_vruntime.max 46773 ± 27% +214.1% 146909 ± 6% sched_debug.cfs_rq:/.min_vruntime.min 65277 ± 6% -90.2% 6372 ± 15% sched_debug.cfs_rq:/.min_vruntime.stddev 36.59 ± 8% -29.3% 25.88 ± 7% sched_debug.cfs_rq:/.nr_running.avg 87.65 ± 14% -44.5% 48.64 ± 6% sched_debug.cfs_rq:/.nr_running.max 25.77 ± 15% -52.1% 12.33 ± 2% sched_debug.cfs_rq:/.nr_running.stddev 23346 ± 12% -50.8% 11487 ± 7% sched_debug.cfs_rq:/.runnable_load_avg.avg 65642 ± 16% -64.9% 23039 ± 7% sched_debug.cfs_rq:/.runnable_load_avg.max 19054 ± 20% -74.1% 4941 ± 3% sched_debug.cfs_rq:/.runnable_load_avg.stddev 141272 ± 30% -93.8% 8802 ±128% sched_debug.cfs_rq:/.spread0.max -138988 ±-18% -86.6% -18657 ±-62% sched_debug.cfs_rq:/.spread0.min 65277 ± 6% -90.2% 6372 ± 15% sched_debug.cfs_rq:/.spread0.stddev 1272 ± 14% -20.8% 1007 ± 5% sched_debug.cfs_rq:/.util_avg.avg 2857 ± 23% -31.0% 1972 ± 4% sched_debug.cfs_rq:/.util_avg.max 555.36 ± 37% -47.6% 290.92 ± 8% sched_debug.cfs_rq:/.util_avg.stddev 4971080 ± 24% -80.3% 978006 ± 3% sched_debug.cpu.avg_idle.max 7448 ± 43% +2715.8% 209722 ± 23% sched_debug.cpu.avg_idle.min 1141706 ± 23% -80.4% 224270 ± 5% sched_debug.cpu.avg_idle.stddev 110.31 ± 30% +54.4% 170.30 ± 23% sched_debug.cpu.clock.stddev 110.31 ± 30% +54.4% 170.30 ± 23% sched_debug.cpu.clock_task.stddev 23385 ± 13% -50.3% 11614 ± 7% sched_debug.cpu.cpu_load[0].avg 66355 ± 16% -65.2% 23117 ± 8% sched_debug.cpu.cpu_load[0].max 19176 ± 20% -74.1% 4975 ± 3% sched_debug.cpu.cpu_load[0].stddev 23612 ± 13% -50.8% 11609 ± 7% sched_debug.cpu.cpu_load[1].avg 65359 ± 18% -66.6% 21840 ± 5% sched_debug.cpu.cpu_load[1].max 18875 ± 20% -75.4% 4639 ± 3% sched_debug.cpu.cpu_load[1].stddev 23709 ± 13% -50.9% 11631 ± 6% sched_debug.cpu.cpu_load[2].avg 65102 ± 18% -68.1% 20773 ± 5% sched_debug.cpu.cpu_load[2].max 18761 ± 21% -76.4% 4431 ± 2% sched_debug.cpu.cpu_load[2].stddev 23784 ± 13% -50.8% 11705 ± 6% sched_debug.cpu.cpu_load[3].avg 64669 ± 19% -69.2% 19945 ± 4% sched_debug.cpu.cpu_load[3].max 18579 ± 21% -77.2% 4242 ± 3% sched_debug.cpu.cpu_load[3].stddev 23900 ± 13% -50.4% 11860 ± 5% sched_debug.cpu.cpu_load[4].avg 63732 ± 19% -69.8% 19251 ± 4% sched_debug.cpu.cpu_load[4].max 18348 ± 20% -77.9% 4051 ± 3% sched_debug.cpu.cpu_load[4].stddev 38363160 ± 8% -29.0% 27246596 ± 7% sched_debug.cpu.load.avg 92064972 ± 14% -42.3% 53093649 ± 10% sched_debug.cpu.load.max 27008459 ± 15% -50.8% 13275018 ± 6% sched_debug.cpu.load.stddev 951016 ± 15% -47.4% 500000 ± 0% sched_debug.cpu.max_idle_balance_cost.avg 4258173 ± 22% -88.3% 500000 ± 0% sched_debug.cpu.max_idle_balance_cost.max 877921 ± 22% -100.0% 0.00 ± -1% sched_debug.cpu.max_idle_balance_cost.stddev 0.00 ± 27% +50.7% 0.00 ± 19% sched_debug.cpu.next_balance.stddev 36.58 ± 8% -29.2% 25.90 ± 7% sched_debug.cpu.nr_running.avg 87.85 ± 14% -44.7% 48.54 ± 6% sched_debug.cpu.nr_running.max 25.77 ± 15% -52.3% 12.30 ± 2% sched_debug.cpu.nr_running.stddev 2586247 ± 8% -33.6% 1716571 ± 6% sched_debug.cpu.nr_switches.stddev 2.67 ± 68% -83.7% 0.44 ± 77% sched_debug.cpu.nr_uninterruptible.avg 404.75 ± 11% -48.2% 209.69 ± 13% sched_debug.cpu.nr_uninterruptible.max 210.44 ± 6% -44.7% 116.41 ± 14% sched_debug.cpu.nr_uninterruptible.stddev 1.31 ± 4% +18.1% 1.55 ± 3% perf-profile.calltrace.cycles.__fget.__fget_light.__fdget.__fdget_pos.sys_read 1.42 ± 4% +17.8% 1.67 ± 3% perf-profile.calltrace.cycles.__fget_light.__fdget.__fdget_pos.sys_read.entry_SYSCALL_64_fastpath 4.95 ± 24% -47.8% 2.58 ± 8% perf-profile.calltrace.cycles.__schedule.schedule.pipe_wait.pipe_read.__vfs_read 9.88 ± 24% -46.3% 5.30 ± 3% perf-profile.calltrace.cycles.__wake_up_common.__wake_up_sync_key.pipe_write.__vfs_write.vfs_write 13.58 ± 20% -36.6% 8.60 ± 2% perf-profile.calltrace.cycles.__wake_up_sync_key.pipe_write.__vfs_write.vfs_write.sys_write 2.27 ± 16% -22.8% 1.75 ± 5% perf-profile.calltrace.cycles._raw_spin_lock_irqsave.__wake_up_sync_key.pipe_write.__vfs_write.vfs_write 0.74 ± 24% +53.5% 1.14 ± 4% perf-profile.calltrace.cycles._raw_spin_unlock_irqrestore.__wake_up_sync_key.pipe_write.__vfs_write.vfs_write 1.39 ± 11% +36.5% 1.89 ± 4% perf-profile.calltrace.cycles.atime_needs_update.touch_atime.pipe_read.__vfs_read.vfs_read 9.27 ± 25% -48.8% 4.75 ± 3% perf-profile.calltrace.cycles.autoremove_wake_function.__wake_up_common.__wake_up_sync_key.pipe_write.__vfs_write 3.79 ± 6% +14.7% 4.34 ± 3% perf-profile.calltrace.cycles.copy_page_from_iter.pipe_write.__vfs_write.vfs_write.sys_write 3.85 ± 7% +21.5% 4.67 ± 3% perf-profile.calltrace.cycles.copy_page_to_iter.pipe_read.__vfs_read.vfs_read.sys_read 3.65 ± 2% +9.6% 4.00 ± 3% perf-profile.calltrace.cycles.copy_user_generic_string.pipe_write.__vfs_write.vfs_write.sys_write 1.61 ± 31% -46.7% 0.86 ± 22% perf-profile.calltrace.cycles.deactivate_task.__schedule.schedule.pipe_wait.pipe_read 9.21 ± 25% -48.9% 4.71 ± 3% perf-profile.calltrace.cycles.default_wake_function.autoremove_wake_function.__wake_up_common.__wake_up_sync_key.pipe_write 1.82 ± 7% +25.4% 2.28 ± 3% perf-profile.calltrace.cycles.file_update_time.pipe_write.__vfs_write.vfs_write.sys_write 1.81 ± 7% +19.2% 2.16 ± 3% perf-profile.calltrace.cycles.mutex_lock.pipe_read.__vfs_read.vfs_read.sys_read 2.15 ± 16% +80.3% 3.87 ± 4% perf-profile.calltrace.cycles.mutex_unlock.__vfs_write.vfs_write.sys_write.entry_SYSCALL_64_fastpath 6.50 ± 23% -46.5% 3.48 ± 8% perf-profile.calltrace.cycles.pipe_wait.pipe_read.__vfs_read.vfs_read.sys_read 2.31 ± 11% -19.3% 1.86 ± 5% perf-profile.calltrace.cycles.pipe_wait.pipe_write.__vfs_write.vfs_write.sys_write 5.25 ± 24% -47.5% 2.75 ± 9% perf-profile.calltrace.cycles.schedule.pipe_wait.pipe_read.__vfs_read.vfs_read 1.97 ± 11% +36.6% 2.70 ± 3% perf-profile.calltrace.cycles.touch_atime.pipe_read.__vfs_read.vfs_read.sys_read 9.98 ± 24% -41.9% 5.80 ± 6% perf-profile.calltrace.cycles.try_to_wake_up.default_wake_function.autoremove_wake_function.__wake_up_common.__wake_up_sync_key 3.50 ± 6% +23.8% 4.34 ± 2% perf-profile.children.cycles.___might_sleep 0.81 ± 27% -46.4% 0.43 ± 3% perf-profile.children.cycles.___perf_sw_event 2.23 ± 6% +18.0% 2.63 ± 3% perf-profile.children.cycles.__might_fault 5.59 ± 6% +19.6% 6.68 ± 3% perf-profile.children.cycles.__might_sleep 0.85 ± 6% +21.0% 1.02 ± 4% perf-profile.children.cycles.__sb_end_write 0.88 ± 5% +19.4% 1.04 ± 4% perf-profile.children.cycles.__sb_start_write 8.50 ± 21% -37.1% 5.34 ± 3% perf-profile.children.cycles.__schedule 11.44 ± 20% -40.3% 6.83 ± 2% perf-profile.children.cycles.__wake_up_common 15.42 ± 18% -33.0% 10.33 ± 2% perf-profile.children.cycles.__wake_up_sync_key 1.75 ± 37% -81.7% 0.32 ± 13% perf-profile.children.cycles._raw_spin_lock 3.42 ± 16% -19.6% 2.75 ± 13% perf-profile.children.cycles._raw_spin_lock_irqsave 3.12 ± 18% -35.6% 2.01 ± 4% perf-profile.children.cycles.activate_task 1.79 ± 13% +32.1% 2.37 ± 3% perf-profile.children.cycles.atime_needs_update 10.67 ± 22% -42.6% 6.12 ± 3% perf-profile.children.cycles.autoremove_wake_function 0.85 ± 27% -56.7% 0.37 ± 5% perf-profile.children.cycles.check_preempt_curr 3.88 ± 6% +15.3% 4.47 ± 3% perf-profile.children.cycles.copy_page_from_iter 4.04 ± 5% +18.2% 4.77 ± 3% perf-profile.children.cycles.copy_page_to_iter 8.24 ± 2% +9.3% 9.01 ± 3% perf-profile.children.cycles.copy_user_generic_string 1.46 ± 9% +36.9% 2.00 ± 2% perf-profile.children.cycles.current_fs_time 2.27 ± 18% -35.3% 1.47 ± 3% perf-profile.children.cycles.deactivate_task 10.60 ± 21% -42.7% 6.08 ± 3% perf-profile.children.cycles.default_wake_function 1.44 ± 16% -33.2% 0.97 ± 2% perf-profile.children.cycles.dequeue_task_fair 2.26 ± 19% -36.0% 1.45 ± 3% perf-profile.children.cycles.enqueue_task_fair 2.45 ± 8% +25.6% 3.08 ± 3% perf-profile.children.cycles.entry_SYSCALL_64 2.27 ± 5% +18.9% 2.70 ± 3% perf-profile.children.cycles.entry_SYSCALL_64_after_swapgs 1.66 ± 26% -35.3% 1.07 ± 6% perf-profile.children.cycles.exit_to_usermode_loop 1.89 ± 8% +25.8% 2.38 ± 2% perf-profile.children.cycles.file_update_time 1.80 ± 2% +16.1% 2.09 ± 2% perf-profile.children.cycles.fput 1.63 ± 5% +20.6% 1.96 ± 3% perf-profile.children.cycles.fsnotify 1.82 ± 24% -38.4% 1.12 ± 3% perf-profile.children.cycles.idle_cpu 0.79 ± 7% +22.7% 0.97 ± 3% perf-profile.children.cycles.iov_iter_init 3.44 ± 8% +48.8% 5.12 ± 3% perf-profile.children.cycles.mutex_unlock 1.18 ± 17% -31.9% 0.80 ± 5% perf-profile.children.cycles.native_write_msr 2.27 ± 20% -35.0% 1.48 ± 4% perf-profile.children.cycles.pick_next_task_fair 8.83 ± 20% -37.8% 5.50 ± 3% perf-profile.children.cycles.pipe_wait 0.84 ± 24% -49.6% 0.42 ± 3% perf-profile.children.cycles.prepare_to_wait 2.35 ± 42% -83.4% 0.39 ± 85% perf-profile.children.cycles.queued_spin_lock_slowpath 8.56 ± 20% -37.2% 5.38 ± 3% perf-profile.children.cycles.schedule 3.50 ± 22% -38.3% 2.16 ± 3% perf-profile.children.cycles.select_task_rq_fair 0.93 ± 16% -30.8% 0.64 ± 7% perf-profile.children.cycles.set_next_entity 0.80 ± 8% +30.4% 1.04 ± 5% perf-profile.children.cycles.timespec_trunc 2.09 ± 10% +33.2% 2.79 ± 3% perf-profile.children.cycles.touch_atime 10.31 ± 21% -42.6% 5.92 ± 3% perf-profile.children.cycles.try_to_wake_up 4.22 ± 21% -40.8% 2.50 ± 4% perf-profile.children.cycles.ttwu_do_activate 0.96 ± 27% -56.1% 0.42 ± 4% perf-profile.children.cycles.ttwu_do_wakeup 0.78 ± 21% -45.8% 0.42 ± 4% perf-profile.children.cycles.update_curr 1.14 ± 6% +22.2% 1.39 ± 4% perf-profile.children.cycles.update_fast_ctr 0.79 ± 22% -40.5% 0.47 ± 3% perf-profile.children.cycles.update_rq_clock 3.50 ± 6% +23.8% 4.34 ± 2% perf-profile.self.cycles.___might_sleep 0.78 ± 28% -47.9% 0.40 ± 3% perf-profile.self.cycles.___perf_sw_event 2.54 ± 6% +13.7% 2.88 ± 3% perf-profile.self.cycles.__might_sleep 1.57 ± 17% -27.0% 1.15 ± 3% perf-profile.self.cycles.__schedule 2.62 ± 5% +18.9% 3.11 ± 3% perf-profile.self.cycles.__vfs_read 2.45 ± 7% +21.8% 2.98 ± 3% perf-profile.self.cycles.__vfs_write 0.91 ± 16% +27.9% 1.17 ± 3% perf-profile.self.cycles.atime_needs_update 2.09 ± 5% +12.5% 2.35 ± 3% perf-profile.self.cycles.copy_page_from_iter 2.31 ± 5% +14.7% 2.65 ± 3% perf-profile.self.cycles.copy_page_to_iter 8.24 ± 2% +9.3% 9.01 ± 3% perf-profile.self.cycles.copy_user_generic_string 0.85 ± 15% -30.4% 0.59 ± 2% perf-profile.self.cycles.dequeue_task_fair 1.01 ± 17% -31.8% 0.68 ± 3% perf-profile.self.cycles.enqueue_task_fair 2.45 ± 8% +25.6% 3.08 ± 3% perf-profile.self.cycles.entry_SYSCALL_64 2.27 ± 5% +18.9% 2.70 ± 3% perf-profile.self.cycles.entry_SYSCALL_64_after_swapgs 1.17 ± 5% +12.4% 1.32 ± 2% perf-profile.self.cycles.entry_SYSCALL_64_fastpath 0.97 ± 6% +24.5% 1.21 ± 2% perf-profile.self.cycles.file_update_time 1.80 ± 2% +16.1% 2.09 ± 2% perf-profile.self.cycles.fput 1.63 ± 5% +20.6% 1.96 ± 3% perf-profile.self.cycles.fsnotify 1.82 ± 24% -38.4% 1.12 ± 3% perf-profile.self.cycles.idle_cpu 0.79 ± 7% +22.7% 0.97 ± 3% perf-profile.self.cycles.iov_iter_init 3.34 ± 9% +52.6% 5.11 ± 3% perf-profile.self.cycles.mutex_unlock 1.18 ± 17% -31.9% 0.80 ± 5% perf-profile.self.cycles.native_write_msr 2.43 ± 1% +11.9% 2.72 ± 3% perf-profile.self.cycles.pipe_read 3.18 ± 1% +17.2% 3.73 ± 3% perf-profile.self.cycles.pipe_write 2.35 ± 42% -83.4% 0.39 ± 85% perf-profile.self.cycles.queued_spin_lock_slowpath 1.04 ± 20% -39.8% 0.63 ± 3% perf-profile.self.cycles.select_task_rq_fair 0.79 ± 5% +24.8% 0.99 ± 3% perf-profile.self.cycles.sys_read 0.79 ± 7% +22.7% 0.97 ± 3% perf-profile.self.cycles.sys_write 0.80 ± 8% +30.4% 1.04 ± 5% perf-profile.self.cycles.timespec_trunc 1.14 ± 6% +22.2% 1.39 ± 4% perf-profile.self.cycles.update_fast_ctr 1.72 ± 5% +20.8% 2.08 ± 2% perf-profile.self.cycles.vfs_read 1.80 ± 5% +19.0% 2.14 ± 2% perf-profile.self.cycles.vfs_write Thanks, Xiaolong