* Re: [LKP] [lkp-robot] [mm] 9092c71bb7: blogbench.write_score -12.3% regression [not found] <20180408015739.GN3845@yexl-desktop> @ 2018-05-29 7:30 ` Huang, Ying 2018-06-05 4:58 ` Huang, Ying 0 siblings, 1 reply; 11+ messages in thread From: Huang, Ying @ 2018-05-29 7:30 UTC (permalink / raw) To: Josef Bacik Cc: kernel test robot, lkp@01.org, Chris Mason, David Sterba, linux-btrfs Hi, Josef, Do you have time to take a look at the regression? kernel test robot <xiaolong.ye@intel.com> writes: > Greeting, > > FYI, we noticed a -12.3% regression of blogbench.write_score and a +9.6% improvement > of blogbench.read_score due to commit: > > > commit: 9092c71bb724dba2ecba849eae69e5c9d39bd3d2 ("mm: use sc->priority for slab shrink targets") > https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master > > in testcase: blogbench > on test machine: 16 threads Intel(R) Xeon(R) CPU D-1541 @ 2.10GHz with 8G memory > with following parameters: > > disk: 1SSD > fs: btrfs > cpufreq_governor: performance > > test-description: Blogbench is a portable filesystem benchmark that tries to reproduce the load of a real-world busy file server. > test-url: https://www.pureftpd.org/project/blogbench > > > > 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 run job.yaml > > ========================================================================================= > compiler/cpufreq_governor/disk/fs/kconfig/rootfs/tbox_group/testcase: > gcc-7/performance/1SSD/btrfs/x86_64-rhel-7.2/debian-x86_64-2016-08-31.cgz/lkp-bdw-de1/blogbench > > commit: > fcb2b0c577 ("mm: show total hugetlb memory consumption in /proc/meminfo") > 9092c71bb7 ("mm: use sc->priority for slab shrink targets") > > fcb2b0c577f145c7 9092c71bb724dba2ecba849eae > ---------------- -------------------------- > %stddev %change %stddev > \ | \ > 3256 -12.3% 2854 blogbench.write_score > 1235237 2% +9.6% 1354163 blogbench.read_score > 28050912 -10.1% 25212230 blogbench.time.file_system_outputs > 6481995 3% +25.0% 8105320 2% blogbench.time.involuntary_context_switches > 906.00 +13.7% 1030 blogbench.time.percent_of_cpu_this_job_got > 2552 +14.0% 2908 blogbench.time.system_time > 173.80 +8.4% 188.32 blogbench.time.user_time > 19353936 +3.6% 20045728 blogbench.time.voluntary_context_switches > 8719514 +13.0% 9850451 softirqs.RCU > 2.97 5% -0.7 2.30 3% mpstat.cpu.idle% > 24.92 -6.5 18.46 mpstat.cpu.iowait% > 0.65 2% +0.1 0.75 mpstat.cpu.soft% > 67.76 +6.7 74.45 mpstat.cpu.sys% > 50206 -10.7% 44858 vmstat.io.bo > 49.25 -9.1% 44.75 2% vmstat.procs.b > 224125 -1.8% 220135 vmstat.system.cs > 48903 +10.7% 54134 vmstat.system.in > 3460654 +10.8% 3834883 meminfo.Active > 3380666 +11.0% 3752872 meminfo.Active(file) > 1853849 -17.4% 1530415 meminfo.Inactive > 1836507 -17.6% 1513054 meminfo.Inactive(file) > 551311 -10.3% 494265 meminfo.SReclaimable > 196525 -12.6% 171775 meminfo.SUnreclaim > 747837 -10.9% 666040 meminfo.Slab > 8.904e+08 -24.9% 6.683e+08 cpuidle.C1.time > 22971020 -12.8% 20035820 cpuidle.C1.usage > 2.518e+08 3% -31.7% 1.72e+08 cpuidle.C1E.time > 821393 2% -33.3% 548003 cpuidle.C1E.usage > 75460078 2% -23.3% 57903768 2% cpuidle.C3.time > 136506 3% -25.3% 101956 3% cpuidle.C3.usage > 56892498 4% -23.3% 43608427 4% cpuidle.C6.time > 85034 3% -33.9% 56184 3% cpuidle.C6.usage > 24373567 -24.5% 18395538 cpuidle.POLL.time > 449033 2% -10.8% 400493 cpuidle.POLL.usage > 1832 +9.3% 2002 turbostat.Avg_MHz > 22967645 -12.8% 20032521 turbostat.C1 > 18.43 -4.6 13.85 turbostat.C1% > 821328 2% -33.3% 547948 turbostat.C1E > 5.21 3% -1.6 3.56 turbostat.C1E% > 136377 3% -25.3% 101823 3% turbostat.C3 > 1.56 2% -0.4 1.20 3% turbostat.C3% > 84404 3% -34.0% 55743 3% turbostat.C6 > 1.17 4% -0.3 0.90 4% turbostat.C6% > 25.93 -26.2% 19.14 turbostat.CPU%c1 > 0.12 3% -19.1% 0.10 9% turbostat.CPU%c3 > 14813304 +10.7% 16398388 turbostat.IRQ > 38.19 +3.6% 39.56 turbostat.PkgWatt > 4.51 +4.5% 4.71 turbostat.RAMWatt > 8111200 13% -63.2% 2986242 48% proc-vmstat.compact_daemon_free_scanned > 1026719 30% -81.2% 193485 30% proc-vmstat.compact_daemon_migrate_scanned > 2444 21% -63.3% 897.50 20% proc-vmstat.compact_daemon_wake > 8111200 13% -63.2% 2986242 48% proc-vmstat.compact_free_scanned > 755491 32% -81.6% 138856 28% proc-vmstat.compact_isolated > 1026719 30% -81.2% 193485 30% proc-vmstat.compact_migrate_scanned > 137.75 34% +2.8e+06% 3801062 2% proc-vmstat.kswapd_inodesteal > 6749 20% -53.6% 3131 12% proc-vmstat.kswapd_low_wmark_hit_quickly > 844991 +11.2% 939487 proc-vmstat.nr_active_file > 3900576 -10.5% 3490567 proc-vmstat.nr_dirtied > 459789 -17.8% 377930 proc-vmstat.nr_inactive_file > 137947 -10.3% 123720 proc-vmstat.nr_slab_reclaimable > 49165 -12.6% 42989 proc-vmstat.nr_slab_unreclaimable > 1382 11% -26.2% 1020 20% proc-vmstat.nr_writeback > 3809266 -10.7% 3403350 proc-vmstat.nr_written > 844489 +11.2% 938974 proc-vmstat.nr_zone_active_file > 459855 -17.8% 378121 proc-vmstat.nr_zone_inactive_file > 7055 18% -52.0% 3389 11% proc-vmstat.pageoutrun > 33764911 2% +21.3% 40946445 proc-vmstat.pgactivate > 42044161 2% +12.1% 47139065 proc-vmstat.pgdeactivate > 92153 20% -69.1% 28514 24% proc-vmstat.pgmigrate_success > 15212270 -10.7% 13591573 proc-vmstat.pgpgout > 42053817 2% +12.1% 47151755 proc-vmstat.pgrefill > 11297 107% +1025.4% 127138 21% proc-vmstat.pgscan_direct > 19930162 -24.0% 15141439 proc-vmstat.pgscan_kswapd > 19423629 -24.0% 14758807 proc-vmstat.pgsteal_kswapd > 10868768 +184.8% 30950752 proc-vmstat.slabs_scanned The slab scan number increased a lot. > 3361780 3% -22.9% 2593327 3% proc-vmstat.workingset_activate > 4994722 2% -43.2% 2835020 2% proc-vmstat.workingset_refault > 316427 -9.3% 286844 slabinfo.Acpi-Namespace.active_objs > 3123 -9.4% 2829 slabinfo.Acpi-Namespace.active_slabs > 318605 -9.4% 288623 slabinfo.Acpi-Namespace.num_objs > 3123 -9.4% 2829 slabinfo.Acpi-Namespace.num_slabs > 220514 -40.7% 130747 slabinfo.btrfs_delayed_node.active_objs > 9751 -25.3% 7283 slabinfo.btrfs_delayed_node.active_slabs > 263293 -25.3% 196669 slabinfo.btrfs_delayed_node.num_objs > 9751 -25.3% 7283 slabinfo.btrfs_delayed_node.num_slabs > 6383 8% -12.0% 5615 2% slabinfo.btrfs_delayed_ref_head.num_objs > 9496 +15.5% 10969 slabinfo.btrfs_extent_buffer.active_objs > 9980 +20.5% 12022 slabinfo.btrfs_extent_buffer.num_objs > 260933 -10.7% 233136 slabinfo.btrfs_extent_map.active_objs > 9392 -10.6% 8396 slabinfo.btrfs_extent_map.active_slabs > 263009 -10.6% 235107 slabinfo.btrfs_extent_map.num_objs > 9392 -10.6% 8396 slabinfo.btrfs_extent_map.num_slabs > 271938 -10.3% 243802 slabinfo.btrfs_inode.active_objs > 9804 -10.6% 8768 slabinfo.btrfs_inode.active_slabs > 273856 -10.4% 245359 slabinfo.btrfs_inode.num_objs > 9804 -10.6% 8768 slabinfo.btrfs_inode.num_slabs > 7085 5% -5.5% 6692 2% slabinfo.btrfs_path.num_objs > 311936 -16.4% 260797 slabinfo.dentry.active_objs > 7803 -9.6% 7058 slabinfo.dentry.active_slabs > 327759 -9.6% 296439 slabinfo.dentry.num_objs > 7803 -9.6% 7058 slabinfo.dentry.num_slabs > 2289 -23.3% 1755 6% slabinfo.proc_inode_cache.active_objs > 2292 -19.0% 1856 6% slabinfo.proc_inode_cache.num_objs > 261546 -12.3% 229485 slabinfo.radix_tree_node.active_objs > 9404 -11.9% 8288 slabinfo.radix_tree_node.active_slabs > 263347 -11.9% 232089 slabinfo.radix_tree_node.num_objs > 9404 -11.9% 8288 slabinfo.radix_tree_node.num_slabs The slab size decreased with the new commit. >From perf-profile result, 26.81 ± 2% -6.5 20.35 ± 2% perf-profile.calltrace.cycles-pp.secondary_startup_64 24.48 ± 2% -5.8 18.73 perf-profile.calltrace.cycles-pp.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64 24.48 ± 2% -5.8 18.73 perf-profile.calltrace.cycles-pp.cpu_startup_entry.start_secondary.secondary_startup_64 24.48 ± 2% -5.8 18.73 perf-profile.calltrace.cycles-pp.start_secondary.secondary_startup_64 22.80 ± 2% -5.5 17.30 ± 2% perf-profile.calltrace.cycles-pp.cpuidle_enter_state.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64 20.20 ± 2% -4.3 15.85 perf-profile.calltrace.cycles-pp.intel_idle.cpuidle_enter_state.do_idle.cpu_startup_entry.start_secondary 23.03 ± 2% -2.6 20.42 perf-profile.calltrace.cycles-pp.sys_rename.entry_SYSCALL_64_fastpath 17.02 ± 2% -1.8 15.17 ± 2% perf-profile.calltrace.cycles-pp.btrfs_rename.vfs_rename.sys_rename.entry_SYSCALL_64_fastpath 17.03 ± 2% -1.8 15.19 ± 2% perf-profile.calltrace.cycles-pp.vfs_rename.sys_rename.entry_SYSCALL_64_fastpath 13.70 ± 2% -1.2 12.47 ± 3% perf-profile.calltrace.cycles-pp.__btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename.entry_SYSCALL_64_fastpath 5.06 ± 3% -1.1 3.96 ± 2% perf-profile.calltrace.cycles-pp.btrfs_async_run_delayed_root.normal_work_helper.process_one_work.worker_thread.kthread 7.37 ± 4% -0.9 6.49 ± 2% perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.finish_wait.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot 7.32 ± 4% -0.9 6.45 ± 2% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.finish_wait.btrfs_tree_lock.btrfs_lock_root_node 1.54 ± 4% -0.7 0.81 ± 7% perf-profile.calltrace.cycles-pp.poll_idle.cpuidle_enter_state.do_idle.cpu_startup_entry.start_secondary 2.33 ± 2% -0.7 1.62 ± 4% perf-profile.calltrace.cycles-pp.do_idle.cpu_startup_entry.start_kernel.secondary_startup_64 2.33 ± 2% -0.7 1.62 ± 4% perf-profile.calltrace.cycles-pp.cpu_startup_entry.start_kernel.secondary_startup_64 2.33 ± 2% -0.7 1.62 ± 4% perf-profile.calltrace.cycles-pp.start_kernel.secondary_startup_64 2.23 ± 3% -0.7 1.53 ± 4% perf-profile.calltrace.cycles-pp.cpuidle_enter_state.do_idle.cpu_startup_entry.start_kernel.secondary_startup_64 2.13 ± 2% -0.7 1.46 ± 4% perf-profile.calltrace.cycles-pp.intel_idle.cpuidle_enter_state.do_idle.cpu_startup_entry.start_kernel 5.59 -0.7 4.94 ± 3% perf-profile.calltrace.cycles-pp.__dentry_kill.dput.sys_rename.entry_SYSCALL_64_fastpath 5.60 -0.7 4.94 ± 3% perf-profile.calltrace.cycles-pp.dput.sys_rename.entry_SYSCALL_64_fastpath 6.96 -0.7 6.31 ± 3% perf-profile.calltrace.cycles-pp.btrfs_del_inode_ref.__btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename 5.59 -0.7 4.94 ± 3% perf-profile.calltrace.cycles-pp.evict.__dentry_kill.dput.sys_rename.entry_SYSCALL_64_fastpath 5.58 -0.6 4.94 ± 3% perf-profile.calltrace.cycles-pp.btrfs_evict_inode.evict.__dentry_kill.dput.sys_rename 6.94 ± 2% -0.6 6.30 ± 3% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_del_inode_ref.__btrfs_unlink_inode.btrfs_rename.vfs_rename 6.66 ± 4% -0.6 6.10 ± 3% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_rename.vfs_rename 6.66 ± 4% -0.6 6.10 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename 3.38 ± 3% -0.5 2.84 ± 3% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_delete_delayed_items.btrfs_async_run_delayed_root.normal_work_helper.process_one_work 3.40 ± 3% -0.5 2.86 ± 3% perf-profile.calltrace.cycles-pp.btrfs_delete_delayed_items.btrfs_async_run_delayed_root.normal_work_helper.process_one_work.worker_thread 7.30 ± 4% -0.4 6.86 ± 2% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item.__btrfs_unlink_inode 1.18 ± 4% -0.4 0.76 ± 2% perf-profile.calltrace.cycles-pp.__btrfs_update_delayed_inode.btrfs_async_run_delayed_root.normal_work_helper.process_one_work.worker_thread 1.16 ± 4% -0.4 0.74 ± 2% perf-profile.calltrace.cycles-pp.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_async_run_delayed_root.normal_work_helper.process_one_work 5.96 ± 2% -0.4 5.54 ± 3% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_inode_ref.__btrfs_unlink_inode 5.96 ± 2% -0.4 5.54 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_inode_ref.__btrfs_unlink_inode.btrfs_rename 1.16 ± 4% -0.4 0.74 ± 2% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_async_run_delayed_root.normal_work_helper 2.95 ± 3% -0.4 2.54 ± 3% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items.btrfs_async_run_delayed_root 2.95 ± 3% -0.4 2.54 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items.btrfs_async_run_delayed_root.normal_work_helper 0.94 ± 2% -0.3 0.59 ± 4% perf-profile.calltrace.cycles-pp.shrink_inactive_list.shrink_node_memcg.shrink_node.kswapd.kthread 1.25 -0.3 0.91 ± 2% perf-profile.calltrace.cycles-pp.shrink_node_memcg.shrink_node.kswapd.kthread.ret_from_fork 0.84 ± 2% -0.3 0.52 ± 3% perf-profile.calltrace.cycles-pp.shrink_page_list.shrink_inactive_list.shrink_node_memcg.shrink_node.kswapd 1.01 -0.3 0.71 ± 3% perf-profile.calltrace.cycles-pp.btrfs_create.path_openat.do_filp_open.do_sys_open.entry_SYSCALL_64_fastpath 0.58 ± 4% -0.2 0.34 ± 70% perf-profile.calltrace.cycles-pp.queued_write_lock_slowpath.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items 1.73 ± 2% -0.2 1.50 ± 2% perf-profile.calltrace.cycles-pp.finish_wait.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_inode_ref 1.67 ± 4% -0.2 1.44 ± 3% perf-profile.calltrace.cycles-pp.finish_wait.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item 1.89 -0.2 1.67 ± 2% perf-profile.calltrace.cycles-pp.btrfs_commit_inode_delayed_inode.btrfs_evict_inode.evict.__dentry_kill.dput 1.88 -0.2 1.66 ± 2% perf-profile.calltrace.cycles-pp.__btrfs_update_delayed_inode.btrfs_commit_inode_delayed_inode.btrfs_evict_inode.evict.__dentry_kill 1.90 -0.2 1.68 ± 2% perf-profile.calltrace.cycles-pp.btrfs_truncate_inode_items.btrfs_evict_inode.evict.__dentry_kill.dput 1.78 -0.2 1.57 ± 5% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_del_orphan_item.btrfs_orphan_del.btrfs_evict_inode.evict 1.88 -0.2 1.67 ± 2% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_truncate_inode_items.btrfs_evict_inode.evict.__dentry_kill 1.78 -0.2 1.57 ± 5% perf-profile.calltrace.cycles-pp.btrfs_orphan_del.btrfs_evict_inode.evict.__dentry_kill.dput 1.78 -0.2 1.57 ± 5% perf-profile.calltrace.cycles-pp.btrfs_del_orphan_item.btrfs_orphan_del.btrfs_evict_inode.evict.__dentry_kill 1.87 -0.2 1.66 ± 2% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_commit_inode_delayed_inode.btrfs_evict_inode 1.87 -0.2 1.66 ± 2% perf-profile.calltrace.cycles-pp.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_commit_inode_delayed_inode.btrfs_evict_inode.evict 0.75 ± 11% -0.2 0.57 ± 7% perf-profile.calltrace.cycles-pp.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent.__btrfs_drop_extents 0.75 ± 11% -0.2 0.57 ± 7% perf-profile.calltrace.cycles-pp.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent.__btrfs_drop_extents.insert_reserved_file_extent 1.39 ± 3% -0.2 1.23 ± 4% perf-profile.calltrace.cycles-pp.prepare_to_wait_event.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items 1.84 -0.2 1.68 ± 3% perf-profile.calltrace.cycles-pp.__btrfs_unlink_inode.btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename 1.84 -0.2 1.68 ± 3% perf-profile.calltrace.cycles-pp.btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename.entry_SYSCALL_64_fastpath 1.62 -0.2 1.46 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_truncate_inode_items.btrfs_evict_inode.evict 0.87 ± 5% -0.2 0.72 ± 5% perf-profile.calltrace.cycles-pp.finish_wait.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items 1.81 -0.2 1.66 ± 3% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_unlink_inode.btrfs_rename 1.81 -0.2 1.66 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_unlink_inode.btrfs_rename.vfs_rename 1.62 -0.2 1.46 ± 3% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_truncate_inode_items.btrfs_evict_inode 1.69 ± 2% -0.1 1.55 ± 2% perf-profile.calltrace.cycles-pp.end_bio_extent_readpage.normal_work_helper.process_one_work.worker_thread.kthread 1.58 -0.1 1.44 ± 2% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_inode.__btrfs_update_delayed_inode 1.58 -0.1 1.44 ± 2% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_commit_inode_delayed_inode 2.89 ± 3% -0.1 2.77 ± 2% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath.queued_write_lock_slowpath.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot 1.51 -0.1 1.39 ± 5% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_orphan_item.btrfs_orphan_del 1.51 -0.1 1.39 ± 5% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_orphan_item.btrfs_orphan_del.btrfs_evict_inode 0.94 -0.1 0.82 perf-profile.calltrace.cycles-pp.schedule_idle.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64 0.93 -0.1 0.81 perf-profile.calltrace.cycles-pp.__schedule.schedule_idle.do_idle.cpu_startup_entry.start_secondary 1.52 -0.1 1.44 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_unlink_inode 0.71 ± 3% -0.1 0.66 ± 2% perf-profile.calltrace.cycles-pp.prepare_to_wait_event.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_truncate_inode_items 0.60 ± 2% -0.0 0.56 ± 2% perf-profile.calltrace.cycles-pp.ttwu_do_activate.try_to_wake_up.autoremove_wake_function.__wake_up_common.wake_up_page_bit 1.61 ± 2% +0.1 1.67 ± 2% perf-profile.calltrace.cycles-pp.kswapd.kthread.ret_from_fork 1.61 ± 2% +0.1 1.67 ± 2% perf-profile.calltrace.cycles-pp.shrink_node.kswapd.kthread.ret_from_fork 0.55 +0.1 0.68 ± 4% perf-profile.calltrace.cycles-pp.find_get_entry.pagecache_get_page.generic_file_read_iter.__vfs_read.vfs_read 0.57 +0.1 0.70 ± 4% perf-profile.calltrace.cycles-pp.pagecache_get_page.generic_file_read_iter.__vfs_read.vfs_read.sys_read 0.59 ± 3% +0.3 0.87 perf-profile.calltrace.cycles-pp.__account_scheduler_latency.enqueue_entity.enqueue_task_fair.ttwu_do_activate.try_to_wake_up 1.29 ± 3% +0.3 1.58 perf-profile.calltrace.cycles-pp.enqueue_task_fair.ttwu_do_activate.try_to_wake_up.autoremove_wake_function.__wake_up_common 0.54 +0.3 0.83 ± 2% perf-profile.calltrace.cycles-pp.queued_read_lock_slowpath.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent 0.84 +0.3 1.18 ± 3% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.prepare_to_wait_event.btrfs_tree_read_lock.btrfs_read_lock_root_node 0.71 ± 4% +0.3 1.05 perf-profile.calltrace.cycles-pp.ttwu_do_activate.try_to_wake_up.autoremove_wake_function.__wake_up_common.__wake_up_common_lock 0.86 +0.4 1.21 ± 3% perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.prepare_to_wait_event.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot 0.90 +0.4 1.28 ± 3% perf-profile.calltrace.cycles-pp.prepare_to_wait_event.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item 0.35 ± 71% +0.4 0.79 ± 3% perf-profile.calltrace.cycles-pp.queued_read_lock_slowpath.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_inode 0.98 ± 4% +0.4 1.41 ± 2% perf-profile.calltrace.cycles-pp.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget 0.98 ± 5% +0.4 1.43 ± 2% perf-profile.calltrace.cycles-pp.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget.btrfs_lookup_dentry 0.94 ± 3% +0.4 1.39 perf-profile.calltrace.cycles-pp.try_to_wake_up.autoremove_wake_function.__wake_up_common.__wake_up_common_lock.btrfs_clear_path_blocking 0.96 ± 3% +0.5 1.42 perf-profile.calltrace.cycles-pp.autoremove_wake_function.__wake_up_common.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot 1.05 ± 2% +0.5 1.52 ± 4% perf-profile.calltrace.cycles-pp.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent 1.06 ± 3% +0.5 1.53 ± 4% perf-profile.calltrace.cycles-pp.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent.__do_readpage 0.97 ± 3% +0.5 1.45 perf-profile.calltrace.cycles-pp.__wake_up_common.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item 0.62 ± 4% +0.5 1.13 perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath.queued_read_lock_slowpath.btrfs_clear_lock_blocking_rw.btrfs_clear_path_blocking.btrfs_search_slot 0.65 ± 4% +0.5 1.19 perf-profile.calltrace.cycles-pp.queued_read_lock_slowpath.btrfs_clear_lock_blocking_rw.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item 0.17 ±141% +0.6 0.73 perf-profile.calltrace.cycles-pp.save_stack_trace_tsk.__account_scheduler_latency.enqueue_entity.enqueue_task_fair.ttwu_do_activate 0.00 +0.6 0.57 ± 2% perf-profile.calltrace.cycles-pp.task_work_run.exit_to_usermode_loop.syscall_return_slowpath.entry_SYSCALL_64_fastpath 0.00 +0.6 0.59 ± 2% perf-profile.calltrace.cycles-pp.exit_to_usermode_loop.syscall_return_slowpath.entry_SYSCALL_64_fastpath 0.00 +0.6 0.60 ± 2% perf-profile.calltrace.cycles-pp.syscall_return_slowpath.entry_SYSCALL_64_fastpath 0.56 ± 4% +0.6 1.17 perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_file_extent 1.17 +0.6 1.78 ± 2% perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.finish_wait.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot 1.16 +0.6 1.77 ± 2% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.finish_wait.btrfs_tree_read_lock.btrfs_read_lock_root_node 1.18 +0.6 1.80 ± 2% perf-profile.calltrace.cycles-pp.finish_wait.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item 0.57 ± 7% +0.6 1.21 ± 2% perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_inode 0.75 ± 4% +0.6 1.38 perf-profile.calltrace.cycles-pp.btrfs_clear_lock_blocking_rw.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry 1.95 +0.7 2.61 perf-profile.calltrace.cycles-pp.copy_user_enhanced_fast_string.copyout.copy_page_to_iter.generic_file_read_iter.__vfs_read 1.96 +0.7 2.62 perf-profile.calltrace.cycles-pp.copyout.copy_page_to_iter.generic_file_read_iter.__vfs_read.vfs_read 0.00 +0.7 0.70 perf-profile.calltrace.cycles-pp.__save_stack_trace.save_stack_trace_tsk.__account_scheduler_latency.enqueue_entity.enqueue_task_fair 2.10 +0.7 2.80 perf-profile.calltrace.cycles-pp.copy_page_to_iter.generic_file_read_iter.__vfs_read.vfs_read.sys_read 0.85 ± 5% +0.7 1.58 perf-profile.calltrace.cycles-pp.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent 0.84 ± 6% +0.7 1.57 perf-profile.calltrace.cycles-pp.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget 0.00 +0.8 0.75 ± 4% perf-profile.calltrace.cycles-pp.super_cache_scan.shrink_slab.shrink_node.kswapd.kthread 0.00 +0.8 0.75 ± 5% perf-profile.calltrace.cycles-pp.shrink_slab.shrink_node.kswapd.kthread.ret_from_fork 1.07 ± 5% +0.9 1.98 ± 2% perf-profile.calltrace.cycles-pp.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent.__do_readpage 1.09 ± 7% +0.9 2.01 ± 2% perf-profile.calltrace.cycles-pp.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget.btrfs_lookup_dentry 4.90 ± 2% +1.3 6.19 perf-profile.calltrace.cycles-pp.__do_page_cache_readahead.ondemand_readahead.generic_file_read_iter.__vfs_read.vfs_read 4.90 ± 2% +1.3 6.20 perf-profile.calltrace.cycles-pp.ondemand_readahead.generic_file_read_iter.__vfs_read.vfs_read.sys_read 4.44 ± 2% +1.3 5.74 perf-profile.calltrace.cycles-pp.extent_readpages.__do_page_cache_readahead.ondemand_readahead.generic_file_read_iter.__vfs_read 2.87 ± 3% +1.4 4.29 perf-profile.calltrace.cycles-pp.__extent_readpages.extent_readpages.__do_page_cache_readahead.ondemand_readahead.generic_file_read_iter 2.22 ± 4% +1.4 3.65 perf-profile.calltrace.cycles-pp.btrfs_lookup_file_extent.btrfs_get_extent.__do_readpage.__extent_readpages.extent_readpages 2.22 ± 4% +1.4 3.65 perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent.__do_readpage.__extent_readpages 2.27 ± 6% +1.4 3.72 perf-profile.calltrace.cycles-pp.btrfs_lookup_inode.btrfs_iget.btrfs_lookup_dentry.btrfs_lookup.path_openat 2.27 ± 6% +1.4 3.72 perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget.btrfs_lookup_dentry.btrfs_lookup 2.72 ± 3% +1.4 4.17 perf-profile.calltrace.cycles-pp.__do_readpage.__extent_readpages.extent_readpages.__do_page_cache_readahead.ondemand_readahead 2.33 ± 4% +1.5 3.79 perf-profile.calltrace.cycles-pp.btrfs_get_extent.__do_readpage.__extent_readpages.extent_readpages.__do_page_cache_readahead 2.38 +1.6 3.95 perf-profile.calltrace.cycles-pp.queued_read_lock_slowpath.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item 2.63 ± 6% +1.7 4.33 perf-profile.calltrace.cycles-pp.btrfs_iget.btrfs_lookup_dentry.btrfs_lookup.path_openat.do_filp_open 1.68 ± 4% +2.1 3.79 perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item 8.66 +2.2 10.90 perf-profile.calltrace.cycles-pp.generic_file_read_iter.__vfs_read.vfs_read.sys_read.entry_SYSCALL_64_fastpath 8.71 +2.3 10.97 perf-profile.calltrace.cycles-pp.__vfs_read.vfs_read.sys_read.entry_SYSCALL_64_fastpath 8.85 +2.3 11.15 perf-profile.calltrace.cycles-pp.vfs_read.sys_read.entry_SYSCALL_64_fastpath 9.02 +2.3 11.33 perf-profile.calltrace.cycles-pp.sys_read.entry_SYSCALL_64_fastpath 2.71 ± 3% +2.6 5.35 perf-profile.calltrace.cycles-pp.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry 4.63 +2.7 7.29 perf-profile.calltrace.cycles-pp.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry 4.65 +2.7 7.32 perf-profile.calltrace.cycles-pp.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry.btrfs_lookup 1.94 ± 2% +2.8 4.77 perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath.queued_read_lock_slowpath.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot 2.81 ± 4% +3.3 6.15 perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot 3.58 ± 2% +3.4 6.98 perf-profile.calltrace.cycles-pp.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry.btrfs_lookup 8.69 +6.3 15.03 perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry.btrfs_lookup.path_openat 8.75 +6.4 15.12 perf-profile.calltrace.cycles-pp.btrfs_lookup_dir_item.btrfs_lookup_dentry.btrfs_lookup.path_openat.do_filp_open 47.34 +8.1 55.46 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_fastpath 11.47 +8.1 19.60 perf-profile.calltrace.cycles-pp.btrfs_lookup.path_openat.do_filp_open.do_sys_open.entry_SYSCALL_64_fastpath 11.46 +8.1 19.59 perf-profile.calltrace.cycles-pp.btrfs_lookup_dentry.btrfs_lookup.path_openat.do_filp_open.do_sys_open 13.90 +8.3 22.20 perf-profile.calltrace.cycles-pp.path_openat.do_filp_open.do_sys_open.entry_SYSCALL_64_fastpath 13.91 +8.3 22.22 perf-profile.calltrace.cycles-pp.do_filp_open.do_sys_open.entry_SYSCALL_64_fastpath 14.13 +8.4 22.52 perf-profile.calltrace.cycles-pp.do_sys_open.entry_SYSCALL_64_fastpath The cycles for dentry lookup increased much too. Is this the reason why write score decreased? If you need more information, please let me know. Best Regards, Huang, Ying > 1140424 12% +40.2% 1598980 14% sched_debug.cfs_rq:/.MIN_vruntime.max > 790.55 +13.0% 893.20 3% sched_debug.cfs_rq:/.exec_clock.stddev > 1140425 12% +40.2% 1598982 14% sched_debug.cfs_rq:/.max_vruntime.max > 0.83 10% +21.5% 1.00 8% sched_debug.cfs_rq:/.nr_running.avg > 3.30 99% +266.3% 12.09 13% sched_debug.cfs_rq:/.removed.load_avg.avg > 153.02 97% +266.6% 560.96 13% sched_debug.cfs_rq:/.removed.runnable_sum.avg > 569.93 102% +173.2% 1556 14% sched_debug.cfs_rq:/.removed.runnable_sum.stddev > 1.42 60% +501.5% 8.52 34% sched_debug.cfs_rq:/.removed.util_avg.avg > 19.88 59% +288.9% 77.29 16% sched_debug.cfs_rq:/.removed.util_avg.max > 5.05 58% +342.3% 22.32 22% sched_debug.cfs_rq:/.removed.util_avg.stddev > 791.44 3% +47.7% 1168 8% sched_debug.cfs_rq:/.util_avg.avg > 1305 6% +33.2% 1738 5% sched_debug.cfs_rq:/.util_avg.max > 450.25 11% +66.2% 748.17 14% sched_debug.cfs_rq:/.util_avg.min > 220.82 8% +21.1% 267.46 5% sched_debug.cfs_rq:/.util_avg.stddev > 363118 11% -23.8% 276520 11% sched_debug.cpu.avg_idle.avg > 726003 8% -30.8% 502313 4% sched_debug.cpu.avg_idle.max > 202629 3% -32.2% 137429 18% sched_debug.cpu.avg_idle.stddev > 31.96 28% +54.6% 49.42 14% sched_debug.cpu.cpu_load[3].min > 36.21 25% +64.0% 59.38 6% sched_debug.cpu.cpu_load[4].min > 1007 5% +20.7% 1216 7% sched_debug.cpu.curr->pid.avg > 4.50 5% +14.8% 5.17 5% sched_debug.cpu.nr_running.max > 2476195 -11.8% 2185022 sched_debug.cpu.nr_switches.max > 212888 -26.6% 156172 3% sched_debug.cpu.nr_switches.stddev > 3570 2% -58.7% 1474 2% sched_debug.cpu.nr_uninterruptible.max > -803.67 -28.7% -573.38 sched_debug.cpu.nr_uninterruptible.min > 1004 2% -50.4% 498.55 3% sched_debug.cpu.nr_uninterruptible.stddev > 2478809 -11.7% 2189310 sched_debug.cpu.sched_count.max > 214130 -26.5% 157298 3% sched_debug.cpu.sched_count.stddev > 489430 2% -16.6% 408309 2% sched_debug.cpu.sched_goidle.avg > 724333 2% -28.2% 520263 2% sched_debug.cpu.sched_goidle.max > 457611 -18.1% 374746 3% sched_debug.cpu.sched_goidle.min > 62957 2% -47.4% 33138 3% sched_debug.cpu.sched_goidle.stddev > 676053 2% -15.4% 571816 2% sched_debug.cpu.ttwu_local.max > 42669 3% +22.3% 52198 sched_debug.cpu.ttwu_local.min > 151873 2% -18.3% 124118 2% sched_debug.cpu.ttwu_local.stddev > > > > blogbench.write_score > > 3300 +-+------------------------------------------------------------------+ > 3250 +-+ +. .+ +. .+ : : : +. .+ .+.+.+. .| > |: +. .+ +.+.+.+ + + + : +. : : +. + +.+ + + | > 3200 +-+ + +.+ + : + + : + + | > 3150 +-+.+ ++ +.+ | > 3100 +-+ | > 3050 +-+ | > | | > 3000 +-+ | > 2950 +-+ O O | > 2900 +-O O O O | > 2850 +-+ O O O O O O O OO O O O | > | O O O O | > 2800 O-+ O O | > 2750 +-+------------------------------------------------------------------+ > > > [*] 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. > > > Thanks, > Xiaolong ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [LKP] [lkp-robot] [mm] 9092c71bb7: blogbench.write_score -12.3% regression 2018-05-29 7:30 ` [LKP] [lkp-robot] [mm] 9092c71bb7: blogbench.write_score -12.3% regression Huang, Ying @ 2018-06-05 4:58 ` Huang, Ying 2018-06-14 1:37 ` Huang, Ying 0 siblings, 1 reply; 11+ messages in thread From: Huang, Ying @ 2018-06-05 4:58 UTC (permalink / raw) To: Josef Bacik; +Cc: David Sterba, Chris Mason, lkp@01.org, linux-btrfs Ping... "Huang, Ying" <ying.huang@intel.com> writes: > Hi, Josef, > > Do you have time to take a look at the regression? > > kernel test robot <xiaolong.ye@intel.com> writes: > >> Greeting, >> >> FYI, we noticed a -12.3% regression of blogbench.write_score and a +9.6% improvement >> of blogbench.read_score due to commit: >> >> >> commit: 9092c71bb724dba2ecba849eae69e5c9d39bd3d2 ("mm: use sc->priority for slab shrink targets") >> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master >> >> in testcase: blogbench >> on test machine: 16 threads Intel(R) Xeon(R) CPU D-1541 @ 2.10GHz with 8G memory >> with following parameters: >> >> disk: 1SSD >> fs: btrfs >> cpufreq_governor: performance >> >> test-description: Blogbench is a portable filesystem benchmark that tries to reproduce the load of a real-world busy file server. >> test-url: https://www.pureftpd.org/project/blogbench >> >> >> >> 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 run job.yaml >> >> ========================================================================================= >> compiler/cpufreq_governor/disk/fs/kconfig/rootfs/tbox_group/testcase: >> gcc-7/performance/1SSD/btrfs/x86_64-rhel-7.2/debian-x86_64-2016-08-31.cgz/lkp-bdw-de1/blogbench >> >> commit: >> fcb2b0c577 ("mm: show total hugetlb memory consumption in /proc/meminfo") >> 9092c71bb7 ("mm: use sc->priority for slab shrink targets") >> >> fcb2b0c577f145c7 9092c71bb724dba2ecba849eae >> ---------------- -------------------------- >> %stddev %change %stddev >> \ | \ >> 3256 -12.3% 2854 blogbench.write_score >> 1235237 2% +9.6% 1354163 blogbench.read_score >> 28050912 -10.1% 25212230 blogbench.time.file_system_outputs >> 6481995 3% +25.0% 8105320 2% blogbench.time.involuntary_context_switches >> 906.00 +13.7% 1030 blogbench.time.percent_of_cpu_this_job_got >> 2552 +14.0% 2908 blogbench.time.system_time >> 173.80 +8.4% 188.32 blogbench.time.user_time >> 19353936 +3.6% 20045728 blogbench.time.voluntary_context_switches >> 8719514 +13.0% 9850451 softirqs.RCU >> 2.97 5% -0.7 2.30 3% mpstat.cpu.idle% >> 24.92 -6.5 18.46 mpstat.cpu.iowait% >> 0.65 2% +0.1 0.75 mpstat.cpu.soft% >> 67.76 +6.7 74.45 mpstat.cpu.sys% >> 50206 -10.7% 44858 vmstat.io.bo >> 49.25 -9.1% 44.75 2% vmstat.procs.b >> 224125 -1.8% 220135 vmstat.system.cs >> 48903 +10.7% 54134 vmstat.system.in >> 3460654 +10.8% 3834883 meminfo.Active >> 3380666 +11.0% 3752872 meminfo.Active(file) >> 1853849 -17.4% 1530415 meminfo.Inactive >> 1836507 -17.6% 1513054 meminfo.Inactive(file) >> 551311 -10.3% 494265 meminfo.SReclaimable >> 196525 -12.6% 171775 meminfo.SUnreclaim >> 747837 -10.9% 666040 meminfo.Slab >> 8.904e+08 -24.9% 6.683e+08 cpuidle.C1.time >> 22971020 -12.8% 20035820 cpuidle.C1.usage >> 2.518e+08 3% -31.7% 1.72e+08 cpuidle.C1E.time >> 821393 2% -33.3% 548003 cpuidle.C1E.usage >> 75460078 2% -23.3% 57903768 2% cpuidle.C3.time >> 136506 3% -25.3% 101956 3% cpuidle.C3.usage >> 56892498 4% -23.3% 43608427 4% cpuidle.C6.time >> 85034 3% -33.9% 56184 3% cpuidle.C6.usage >> 24373567 -24.5% 18395538 cpuidle.POLL.time >> 449033 2% -10.8% 400493 cpuidle.POLL.usage >> 1832 +9.3% 2002 turbostat.Avg_MHz >> 22967645 -12.8% 20032521 turbostat.C1 >> 18.43 -4.6 13.85 turbostat.C1% >> 821328 2% -33.3% 547948 turbostat.C1E >> 5.21 3% -1.6 3.56 turbostat.C1E% >> 136377 3% -25.3% 101823 3% turbostat.C3 >> 1.56 2% -0.4 1.20 3% turbostat.C3% >> 84404 3% -34.0% 55743 3% turbostat.C6 >> 1.17 4% -0.3 0.90 4% turbostat.C6% >> 25.93 -26.2% 19.14 turbostat.CPU%c1 >> 0.12 3% -19.1% 0.10 9% turbostat.CPU%c3 >> 14813304 +10.7% 16398388 turbostat.IRQ >> 38.19 +3.6% 39.56 turbostat.PkgWatt >> 4.51 +4.5% 4.71 turbostat.RAMWatt >> 8111200 13% -63.2% 2986242 48% proc-vmstat.compact_daemon_free_scanned >> 1026719 30% -81.2% 193485 30% proc-vmstat.compact_daemon_migrate_scanned >> 2444 21% -63.3% 897.50 20% proc-vmstat.compact_daemon_wake >> 8111200 13% -63.2% 2986242 48% proc-vmstat.compact_free_scanned >> 755491 32% -81.6% 138856 28% proc-vmstat.compact_isolated >> 1026719 30% -81.2% 193485 30% proc-vmstat.compact_migrate_scanned >> 137.75 34% +2.8e+06% 3801062 2% proc-vmstat.kswapd_inodesteal >> 6749 20% -53.6% 3131 12% proc-vmstat.kswapd_low_wmark_hit_quickly >> 844991 +11.2% 939487 proc-vmstat.nr_active_file >> 3900576 -10.5% 3490567 proc-vmstat.nr_dirtied >> 459789 -17.8% 377930 proc-vmstat.nr_inactive_file >> 137947 -10.3% 123720 proc-vmstat.nr_slab_reclaimable >> 49165 -12.6% 42989 proc-vmstat.nr_slab_unreclaimable >> 1382 11% -26.2% 1020 20% proc-vmstat.nr_writeback >> 3809266 -10.7% 3403350 proc-vmstat.nr_written >> 844489 +11.2% 938974 proc-vmstat.nr_zone_active_file >> 459855 -17.8% 378121 proc-vmstat.nr_zone_inactive_file >> 7055 18% -52.0% 3389 11% proc-vmstat.pageoutrun >> 33764911 2% +21.3% 40946445 proc-vmstat.pgactivate >> 42044161 2% +12.1% 47139065 proc-vmstat.pgdeactivate >> 92153 20% -69.1% 28514 24% proc-vmstat.pgmigrate_success >> 15212270 -10.7% 13591573 proc-vmstat.pgpgout >> 42053817 2% +12.1% 47151755 proc-vmstat.pgrefill >> 11297 107% +1025.4% 127138 21% proc-vmstat.pgscan_direct >> 19930162 -24.0% 15141439 proc-vmstat.pgscan_kswapd >> 19423629 -24.0% 14758807 proc-vmstat.pgsteal_kswapd >> 10868768 +184.8% 30950752 proc-vmstat.slabs_scanned > > The slab scan number increased a lot. > >> 3361780 3% -22.9% 2593327 3% proc-vmstat.workingset_activate >> 4994722 2% -43.2% 2835020 2% proc-vmstat.workingset_refault >> 316427 -9.3% 286844 slabinfo.Acpi-Namespace.active_objs >> 3123 -9.4% 2829 slabinfo.Acpi-Namespace.active_slabs >> 318605 -9.4% 288623 slabinfo.Acpi-Namespace.num_objs >> 3123 -9.4% 2829 slabinfo.Acpi-Namespace.num_slabs >> 220514 -40.7% 130747 slabinfo.btrfs_delayed_node.active_objs >> 9751 -25.3% 7283 slabinfo.btrfs_delayed_node.active_slabs >> 263293 -25.3% 196669 slabinfo.btrfs_delayed_node.num_objs >> 9751 -25.3% 7283 slabinfo.btrfs_delayed_node.num_slabs >> 6383 8% -12.0% 5615 2% slabinfo.btrfs_delayed_ref_head.num_objs >> 9496 +15.5% 10969 slabinfo.btrfs_extent_buffer.active_objs >> 9980 +20.5% 12022 slabinfo.btrfs_extent_buffer.num_objs >> 260933 -10.7% 233136 slabinfo.btrfs_extent_map.active_objs >> 9392 -10.6% 8396 slabinfo.btrfs_extent_map.active_slabs >> 263009 -10.6% 235107 slabinfo.btrfs_extent_map.num_objs >> 9392 -10.6% 8396 slabinfo.btrfs_extent_map.num_slabs >> 271938 -10.3% 243802 slabinfo.btrfs_inode.active_objs >> 9804 -10.6% 8768 slabinfo.btrfs_inode.active_slabs >> 273856 -10.4% 245359 slabinfo.btrfs_inode.num_objs >> 9804 -10.6% 8768 slabinfo.btrfs_inode.num_slabs >> 7085 5% -5.5% 6692 2% slabinfo.btrfs_path.num_objs >> 311936 -16.4% 260797 slabinfo.dentry.active_objs >> 7803 -9.6% 7058 slabinfo.dentry.active_slabs >> 327759 -9.6% 296439 slabinfo.dentry.num_objs >> 7803 -9.6% 7058 slabinfo.dentry.num_slabs >> 2289 -23.3% 1755 6% slabinfo.proc_inode_cache.active_objs >> 2292 -19.0% 1856 6% slabinfo.proc_inode_cache.num_objs >> 261546 -12.3% 229485 slabinfo.radix_tree_node.active_objs >> 9404 -11.9% 8288 slabinfo.radix_tree_node.active_slabs >> 263347 -11.9% 232089 slabinfo.radix_tree_node.num_objs >> 9404 -11.9% 8288 slabinfo.radix_tree_node.num_slabs > > The slab size decreased with the new commit. > > From perf-profile result, > > 26.81 ± 2% -6.5 20.35 ± 2% perf-profile.calltrace.cycles-pp.secondary_startup_64 > 24.48 ± 2% -5.8 18.73 perf-profile.calltrace.cycles-pp.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64 > 24.48 ± 2% -5.8 18.73 perf-profile.calltrace.cycles-pp.cpu_startup_entry.start_secondary.secondary_startup_64 > 24.48 ± 2% -5.8 18.73 perf-profile.calltrace.cycles-pp.start_secondary.secondary_startup_64 > 22.80 ± 2% -5.5 17.30 ± 2% perf-profile.calltrace.cycles-pp.cpuidle_enter_state.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64 > 20.20 ± 2% -4.3 15.85 perf-profile.calltrace.cycles-pp.intel_idle.cpuidle_enter_state.do_idle.cpu_startup_entry.start_secondary > 23.03 ± 2% -2.6 20.42 perf-profile.calltrace.cycles-pp.sys_rename.entry_SYSCALL_64_fastpath > 17.02 ± 2% -1.8 15.17 ± 2% perf-profile.calltrace.cycles-pp.btrfs_rename.vfs_rename.sys_rename.entry_SYSCALL_64_fastpath > 17.03 ± 2% -1.8 15.19 ± 2% perf-profile.calltrace.cycles-pp.vfs_rename.sys_rename.entry_SYSCALL_64_fastpath > 13.70 ± 2% -1.2 12.47 ± 3% perf-profile.calltrace.cycles-pp.__btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename.entry_SYSCALL_64_fastpath > 5.06 ± 3% -1.1 3.96 ± 2% perf-profile.calltrace.cycles-pp.btrfs_async_run_delayed_root.normal_work_helper.process_one_work.worker_thread.kthread > 7.37 ± 4% -0.9 6.49 ± 2% perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.finish_wait.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot > 7.32 ± 4% -0.9 6.45 ± 2% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.finish_wait.btrfs_tree_lock.btrfs_lock_root_node > 1.54 ± 4% -0.7 0.81 ± 7% perf-profile.calltrace.cycles-pp.poll_idle.cpuidle_enter_state.do_idle.cpu_startup_entry.start_secondary > 2.33 ± 2% -0.7 1.62 ± 4% perf-profile.calltrace.cycles-pp.do_idle.cpu_startup_entry.start_kernel.secondary_startup_64 > 2.33 ± 2% -0.7 1.62 ± 4% perf-profile.calltrace.cycles-pp.cpu_startup_entry.start_kernel.secondary_startup_64 > 2.33 ± 2% -0.7 1.62 ± 4% perf-profile.calltrace.cycles-pp.start_kernel.secondary_startup_64 > 2.23 ± 3% -0.7 1.53 ± 4% perf-profile.calltrace.cycles-pp.cpuidle_enter_state.do_idle.cpu_startup_entry.start_kernel.secondary_startup_64 > 2.13 ± 2% -0.7 1.46 ± 4% perf-profile.calltrace.cycles-pp.intel_idle.cpuidle_enter_state.do_idle.cpu_startup_entry.start_kernel > 5.59 -0.7 4.94 ± 3% perf-profile.calltrace.cycles-pp.__dentry_kill.dput.sys_rename.entry_SYSCALL_64_fastpath > 5.60 -0.7 4.94 ± 3% perf-profile.calltrace.cycles-pp.dput.sys_rename.entry_SYSCALL_64_fastpath > 6.96 -0.7 6.31 ± 3% perf-profile.calltrace.cycles-pp.btrfs_del_inode_ref.__btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename > 5.59 -0.7 4.94 ± 3% perf-profile.calltrace.cycles-pp.evict.__dentry_kill.dput.sys_rename.entry_SYSCALL_64_fastpath > 5.58 -0.6 4.94 ± 3% perf-profile.calltrace.cycles-pp.btrfs_evict_inode.evict.__dentry_kill.dput.sys_rename > 6.94 ± 2% -0.6 6.30 ± 3% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_del_inode_ref.__btrfs_unlink_inode.btrfs_rename.vfs_rename > 6.66 ± 4% -0.6 6.10 ± 3% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_rename.vfs_rename > 6.66 ± 4% -0.6 6.10 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename > 3.38 ± 3% -0.5 2.84 ± 3% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_delete_delayed_items.btrfs_async_run_delayed_root.normal_work_helper.process_one_work > 3.40 ± 3% -0.5 2.86 ± 3% perf-profile.calltrace.cycles-pp.btrfs_delete_delayed_items.btrfs_async_run_delayed_root.normal_work_helper.process_one_work.worker_thread > 7.30 ± 4% -0.4 6.86 ± 2% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item.__btrfs_unlink_inode > 1.18 ± 4% -0.4 0.76 ± 2% perf-profile.calltrace.cycles-pp.__btrfs_update_delayed_inode.btrfs_async_run_delayed_root.normal_work_helper.process_one_work.worker_thread > 1.16 ± 4% -0.4 0.74 ± 2% perf-profile.calltrace.cycles-pp.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_async_run_delayed_root.normal_work_helper.process_one_work > 5.96 ± 2% -0.4 5.54 ± 3% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_inode_ref.__btrfs_unlink_inode > 5.96 ± 2% -0.4 5.54 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_inode_ref.__btrfs_unlink_inode.btrfs_rename > 1.16 ± 4% -0.4 0.74 ± 2% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_async_run_delayed_root.normal_work_helper > 2.95 ± 3% -0.4 2.54 ± 3% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items.btrfs_async_run_delayed_root > 2.95 ± 3% -0.4 2.54 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items.btrfs_async_run_delayed_root.normal_work_helper > 0.94 ± 2% -0.3 0.59 ± 4% perf-profile.calltrace.cycles-pp.shrink_inactive_list.shrink_node_memcg.shrink_node.kswapd.kthread > 1.25 -0.3 0.91 ± 2% perf-profile.calltrace.cycles-pp.shrink_node_memcg.shrink_node.kswapd.kthread.ret_from_fork > 0.84 ± 2% -0.3 0.52 ± 3% perf-profile.calltrace.cycles-pp.shrink_page_list.shrink_inactive_list.shrink_node_memcg.shrink_node.kswapd > 1.01 -0.3 0.71 ± 3% perf-profile.calltrace.cycles-pp.btrfs_create.path_openat.do_filp_open.do_sys_open.entry_SYSCALL_64_fastpath > 0.58 ± 4% -0.2 0.34 ± 70% perf-profile.calltrace.cycles-pp.queued_write_lock_slowpath.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items > 1.73 ± 2% -0.2 1.50 ± 2% perf-profile.calltrace.cycles-pp.finish_wait.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_inode_ref > 1.67 ± 4% -0.2 1.44 ± 3% perf-profile.calltrace.cycles-pp.finish_wait.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item > 1.89 -0.2 1.67 ± 2% perf-profile.calltrace.cycles-pp.btrfs_commit_inode_delayed_inode.btrfs_evict_inode.evict.__dentry_kill.dput > 1.88 -0.2 1.66 ± 2% perf-profile.calltrace.cycles-pp.__btrfs_update_delayed_inode.btrfs_commit_inode_delayed_inode.btrfs_evict_inode.evict.__dentry_kill > 1.90 -0.2 1.68 ± 2% perf-profile.calltrace.cycles-pp.btrfs_truncate_inode_items.btrfs_evict_inode.evict.__dentry_kill.dput > 1.78 -0.2 1.57 ± 5% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_del_orphan_item.btrfs_orphan_del.btrfs_evict_inode.evict > 1.88 -0.2 1.67 ± 2% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_truncate_inode_items.btrfs_evict_inode.evict.__dentry_kill > 1.78 -0.2 1.57 ± 5% perf-profile.calltrace.cycles-pp.btrfs_orphan_del.btrfs_evict_inode.evict.__dentry_kill.dput > 1.78 -0.2 1.57 ± 5% perf-profile.calltrace.cycles-pp.btrfs_del_orphan_item.btrfs_orphan_del.btrfs_evict_inode.evict.__dentry_kill > 1.87 -0.2 1.66 ± 2% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_commit_inode_delayed_inode.btrfs_evict_inode > 1.87 -0.2 1.66 ± 2% perf-profile.calltrace.cycles-pp.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_commit_inode_delayed_inode.btrfs_evict_inode.evict > 0.75 ± 11% -0.2 0.57 ± 7% perf-profile.calltrace.cycles-pp.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent.__btrfs_drop_extents > 0.75 ± 11% -0.2 0.57 ± 7% perf-profile.calltrace.cycles-pp.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent.__btrfs_drop_extents.insert_reserved_file_extent > 1.39 ± 3% -0.2 1.23 ± 4% perf-profile.calltrace.cycles-pp.prepare_to_wait_event.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items > 1.84 -0.2 1.68 ± 3% perf-profile.calltrace.cycles-pp.__btrfs_unlink_inode.btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename > 1.84 -0.2 1.68 ± 3% perf-profile.calltrace.cycles-pp.btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename.entry_SYSCALL_64_fastpath > 1.62 -0.2 1.46 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_truncate_inode_items.btrfs_evict_inode.evict > 0.87 ± 5% -0.2 0.72 ± 5% perf-profile.calltrace.cycles-pp.finish_wait.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items > 1.81 -0.2 1.66 ± 3% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_unlink_inode.btrfs_rename > 1.81 -0.2 1.66 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_unlink_inode.btrfs_rename.vfs_rename > 1.62 -0.2 1.46 ± 3% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_truncate_inode_items.btrfs_evict_inode > 1.69 ± 2% -0.1 1.55 ± 2% perf-profile.calltrace.cycles-pp.end_bio_extent_readpage.normal_work_helper.process_one_work.worker_thread.kthread > 1.58 -0.1 1.44 ± 2% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_inode.__btrfs_update_delayed_inode > 1.58 -0.1 1.44 ± 2% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_commit_inode_delayed_inode > 2.89 ± 3% -0.1 2.77 ± 2% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath.queued_write_lock_slowpath.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot > 1.51 -0.1 1.39 ± 5% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_orphan_item.btrfs_orphan_del > 1.51 -0.1 1.39 ± 5% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_orphan_item.btrfs_orphan_del.btrfs_evict_inode > 0.94 -0.1 0.82 perf-profile.calltrace.cycles-pp.schedule_idle.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64 > 0.93 -0.1 0.81 perf-profile.calltrace.cycles-pp.__schedule.schedule_idle.do_idle.cpu_startup_entry.start_secondary > 1.52 -0.1 1.44 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_unlink_inode > 0.71 ± 3% -0.1 0.66 ± 2% perf-profile.calltrace.cycles-pp.prepare_to_wait_event.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_truncate_inode_items > 0.60 ± 2% -0.0 0.56 ± 2% perf-profile.calltrace.cycles-pp.ttwu_do_activate.try_to_wake_up.autoremove_wake_function.__wake_up_common.wake_up_page_bit > 1.61 ± 2% +0.1 1.67 ± 2% perf-profile.calltrace.cycles-pp.kswapd.kthread.ret_from_fork > 1.61 ± 2% +0.1 1.67 ± 2% perf-profile.calltrace.cycles-pp.shrink_node.kswapd.kthread.ret_from_fork > 0.55 +0.1 0.68 ± 4% perf-profile.calltrace.cycles-pp.find_get_entry.pagecache_get_page.generic_file_read_iter.__vfs_read.vfs_read > 0.57 +0.1 0.70 ± 4% perf-profile.calltrace.cycles-pp.pagecache_get_page.generic_file_read_iter.__vfs_read.vfs_read.sys_read > 0.59 ± 3% +0.3 0.87 perf-profile.calltrace.cycles-pp.__account_scheduler_latency.enqueue_entity.enqueue_task_fair.ttwu_do_activate.try_to_wake_up > 1.29 ± 3% +0.3 1.58 perf-profile.calltrace.cycles-pp.enqueue_task_fair.ttwu_do_activate.try_to_wake_up.autoremove_wake_function.__wake_up_common > 0.54 +0.3 0.83 ± 2% perf-profile.calltrace.cycles-pp.queued_read_lock_slowpath.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent > 0.84 +0.3 1.18 ± 3% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.prepare_to_wait_event.btrfs_tree_read_lock.btrfs_read_lock_root_node > 0.71 ± 4% +0.3 1.05 perf-profile.calltrace.cycles-pp.ttwu_do_activate.try_to_wake_up.autoremove_wake_function.__wake_up_common.__wake_up_common_lock > 0.86 +0.4 1.21 ± 3% perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.prepare_to_wait_event.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot > 0.90 +0.4 1.28 ± 3% perf-profile.calltrace.cycles-pp.prepare_to_wait_event.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item > 0.35 ± 71% +0.4 0.79 ± 3% perf-profile.calltrace.cycles-pp.queued_read_lock_slowpath.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_inode > 0.98 ± 4% +0.4 1.41 ± 2% perf-profile.calltrace.cycles-pp.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget > 0.98 ± 5% +0.4 1.43 ± 2% perf-profile.calltrace.cycles-pp.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget.btrfs_lookup_dentry > 0.94 ± 3% +0.4 1.39 perf-profile.calltrace.cycles-pp.try_to_wake_up.autoremove_wake_function.__wake_up_common.__wake_up_common_lock.btrfs_clear_path_blocking > 0.96 ± 3% +0.5 1.42 perf-profile.calltrace.cycles-pp.autoremove_wake_function.__wake_up_common.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot > 1.05 ± 2% +0.5 1.52 ± 4% perf-profile.calltrace.cycles-pp.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent > 1.06 ± 3% +0.5 1.53 ± 4% perf-profile.calltrace.cycles-pp.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent.__do_readpage > 0.97 ± 3% +0.5 1.45 perf-profile.calltrace.cycles-pp.__wake_up_common.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item > 0.62 ± 4% +0.5 1.13 perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath.queued_read_lock_slowpath.btrfs_clear_lock_blocking_rw.btrfs_clear_path_blocking.btrfs_search_slot > 0.65 ± 4% +0.5 1.19 perf-profile.calltrace.cycles-pp.queued_read_lock_slowpath.btrfs_clear_lock_blocking_rw.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item > 0.17 ±141% +0.6 0.73 perf-profile.calltrace.cycles-pp.save_stack_trace_tsk.__account_scheduler_latency.enqueue_entity.enqueue_task_fair.ttwu_do_activate > 0.00 +0.6 0.57 ± 2% perf-profile.calltrace.cycles-pp.task_work_run.exit_to_usermode_loop.syscall_return_slowpath.entry_SYSCALL_64_fastpath > 0.00 +0.6 0.59 ± 2% perf-profile.calltrace.cycles-pp.exit_to_usermode_loop.syscall_return_slowpath.entry_SYSCALL_64_fastpath > 0.00 +0.6 0.60 ± 2% perf-profile.calltrace.cycles-pp.syscall_return_slowpath.entry_SYSCALL_64_fastpath > 0.56 ± 4% +0.6 1.17 perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_file_extent > 1.17 +0.6 1.78 ± 2% perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.finish_wait.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot > 1.16 +0.6 1.77 ± 2% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.finish_wait.btrfs_tree_read_lock.btrfs_read_lock_root_node > 1.18 +0.6 1.80 ± 2% perf-profile.calltrace.cycles-pp.finish_wait.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item > 0.57 ± 7% +0.6 1.21 ± 2% perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_inode > 0.75 ± 4% +0.6 1.38 perf-profile.calltrace.cycles-pp.btrfs_clear_lock_blocking_rw.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry > 1.95 +0.7 2.61 perf-profile.calltrace.cycles-pp.copy_user_enhanced_fast_string.copyout.copy_page_to_iter.generic_file_read_iter.__vfs_read > 1.96 +0.7 2.62 perf-profile.calltrace.cycles-pp.copyout.copy_page_to_iter.generic_file_read_iter.__vfs_read.vfs_read > 0.00 +0.7 0.70 perf-profile.calltrace.cycles-pp.__save_stack_trace.save_stack_trace_tsk.__account_scheduler_latency.enqueue_entity.enqueue_task_fair > 2.10 +0.7 2.80 perf-profile.calltrace.cycles-pp.copy_page_to_iter.generic_file_read_iter.__vfs_read.vfs_read.sys_read > 0.85 ± 5% +0.7 1.58 perf-profile.calltrace.cycles-pp.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent > 0.84 ± 6% +0.7 1.57 perf-profile.calltrace.cycles-pp.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget > 0.00 +0.8 0.75 ± 4% perf-profile.calltrace.cycles-pp.super_cache_scan.shrink_slab.shrink_node.kswapd.kthread > 0.00 +0.8 0.75 ± 5% perf-profile.calltrace.cycles-pp.shrink_slab.shrink_node.kswapd.kthread.ret_from_fork > 1.07 ± 5% +0.9 1.98 ± 2% perf-profile.calltrace.cycles-pp.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent.__do_readpage > 1.09 ± 7% +0.9 2.01 ± 2% perf-profile.calltrace.cycles-pp.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget.btrfs_lookup_dentry > 4.90 ± 2% +1.3 6.19 perf-profile.calltrace.cycles-pp.__do_page_cache_readahead.ondemand_readahead.generic_file_read_iter.__vfs_read.vfs_read > 4.90 ± 2% +1.3 6.20 perf-profile.calltrace.cycles-pp.ondemand_readahead.generic_file_read_iter.__vfs_read.vfs_read.sys_read > 4.44 ± 2% +1.3 5.74 perf-profile.calltrace.cycles-pp.extent_readpages.__do_page_cache_readahead.ondemand_readahead.generic_file_read_iter.__vfs_read > 2.87 ± 3% +1.4 4.29 perf-profile.calltrace.cycles-pp.__extent_readpages.extent_readpages.__do_page_cache_readahead.ondemand_readahead.generic_file_read_iter > 2.22 ± 4% +1.4 3.65 perf-profile.calltrace.cycles-pp.btrfs_lookup_file_extent.btrfs_get_extent.__do_readpage.__extent_readpages.extent_readpages > 2.22 ± 4% +1.4 3.65 perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent.__do_readpage.__extent_readpages > 2.27 ± 6% +1.4 3.72 perf-profile.calltrace.cycles-pp.btrfs_lookup_inode.btrfs_iget.btrfs_lookup_dentry.btrfs_lookup.path_openat > 2.27 ± 6% +1.4 3.72 perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget.btrfs_lookup_dentry.btrfs_lookup > 2.72 ± 3% +1.4 4.17 perf-profile.calltrace.cycles-pp.__do_readpage.__extent_readpages.extent_readpages.__do_page_cache_readahead.ondemand_readahead > 2.33 ± 4% +1.5 3.79 perf-profile.calltrace.cycles-pp.btrfs_get_extent.__do_readpage.__extent_readpages.extent_readpages.__do_page_cache_readahead > 2.38 +1.6 3.95 perf-profile.calltrace.cycles-pp.queued_read_lock_slowpath.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item > 2.63 ± 6% +1.7 4.33 perf-profile.calltrace.cycles-pp.btrfs_iget.btrfs_lookup_dentry.btrfs_lookup.path_openat.do_filp_open > 1.68 ± 4% +2.1 3.79 perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item > 8.66 +2.2 10.90 perf-profile.calltrace.cycles-pp.generic_file_read_iter.__vfs_read.vfs_read.sys_read.entry_SYSCALL_64_fastpath > 8.71 +2.3 10.97 perf-profile.calltrace.cycles-pp.__vfs_read.vfs_read.sys_read.entry_SYSCALL_64_fastpath > 8.85 +2.3 11.15 perf-profile.calltrace.cycles-pp.vfs_read.sys_read.entry_SYSCALL_64_fastpath > 9.02 +2.3 11.33 perf-profile.calltrace.cycles-pp.sys_read.entry_SYSCALL_64_fastpath > 2.71 ± 3% +2.6 5.35 perf-profile.calltrace.cycles-pp.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry > 4.63 +2.7 7.29 perf-profile.calltrace.cycles-pp.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry > 4.65 +2.7 7.32 perf-profile.calltrace.cycles-pp.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry.btrfs_lookup > 1.94 ± 2% +2.8 4.77 perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath.queued_read_lock_slowpath.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot > 2.81 ± 4% +3.3 6.15 perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot > 3.58 ± 2% +3.4 6.98 perf-profile.calltrace.cycles-pp.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry.btrfs_lookup > 8.69 +6.3 15.03 perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry.btrfs_lookup.path_openat > 8.75 +6.4 15.12 perf-profile.calltrace.cycles-pp.btrfs_lookup_dir_item.btrfs_lookup_dentry.btrfs_lookup.path_openat.do_filp_open > 47.34 +8.1 55.46 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_fastpath > 11.47 +8.1 19.60 perf-profile.calltrace.cycles-pp.btrfs_lookup.path_openat.do_filp_open.do_sys_open.entry_SYSCALL_64_fastpath > 11.46 +8.1 19.59 perf-profile.calltrace.cycles-pp.btrfs_lookup_dentry.btrfs_lookup.path_openat.do_filp_open.do_sys_open > 13.90 +8.3 22.20 perf-profile.calltrace.cycles-pp.path_openat.do_filp_open.do_sys_open.entry_SYSCALL_64_fastpath > 13.91 +8.3 22.22 perf-profile.calltrace.cycles-pp.do_filp_open.do_sys_open.entry_SYSCALL_64_fastpath > 14.13 +8.4 22.52 perf-profile.calltrace.cycles-pp.do_sys_open.entry_SYSCALL_64_fastpath > > The cycles for dentry lookup increased much too. Is this the reason why > write score decreased? > > If you need more information, please let me know. > > Best Regards, > Huang, Ying > >> 1140424 12% +40.2% 1598980 14% sched_debug.cfs_rq:/.MIN_vruntime.max >> 790.55 +13.0% 893.20 3% sched_debug.cfs_rq:/.exec_clock.stddev >> 1140425 12% +40.2% 1598982 14% sched_debug.cfs_rq:/.max_vruntime.max >> 0.83 10% +21.5% 1.00 8% sched_debug.cfs_rq:/.nr_running.avg >> 3.30 99% +266.3% 12.09 13% sched_debug.cfs_rq:/.removed.load_avg.avg >> 153.02 97% +266.6% 560.96 13% sched_debug.cfs_rq:/.removed.runnable_sum.avg >> 569.93 102% +173.2% 1556 14% sched_debug.cfs_rq:/.removed.runnable_sum.stddev >> 1.42 60% +501.5% 8.52 34% sched_debug.cfs_rq:/.removed.util_avg.avg >> 19.88 59% +288.9% 77.29 16% sched_debug.cfs_rq:/.removed.util_avg.max >> 5.05 58% +342.3% 22.32 22% sched_debug.cfs_rq:/.removed.util_avg.stddev >> 791.44 3% +47.7% 1168 8% sched_debug.cfs_rq:/.util_avg.avg >> 1305 6% +33.2% 1738 5% sched_debug.cfs_rq:/.util_avg.max >> 450.25 11% +66.2% 748.17 14% sched_debug.cfs_rq:/.util_avg.min >> 220.82 8% +21.1% 267.46 5% sched_debug.cfs_rq:/.util_avg.stddev >> 363118 11% -23.8% 276520 11% sched_debug.cpu.avg_idle.avg >> 726003 8% -30.8% 502313 4% sched_debug.cpu.avg_idle.max >> 202629 3% -32.2% 137429 18% sched_debug.cpu.avg_idle.stddev >> 31.96 28% +54.6% 49.42 14% sched_debug.cpu.cpu_load[3].min >> 36.21 25% +64.0% 59.38 6% sched_debug.cpu.cpu_load[4].min >> 1007 5% +20.7% 1216 7% sched_debug.cpu.curr->pid.avg >> 4.50 5% +14.8% 5.17 5% sched_debug.cpu.nr_running.max >> 2476195 -11.8% 2185022 sched_debug.cpu.nr_switches.max >> 212888 -26.6% 156172 3% sched_debug.cpu.nr_switches.stddev >> 3570 2% -58.7% 1474 2% sched_debug.cpu.nr_uninterruptible.max >> -803.67 -28.7% -573.38 sched_debug.cpu.nr_uninterruptible.min >> 1004 2% -50.4% 498.55 3% sched_debug.cpu.nr_uninterruptible.stddev >> 2478809 -11.7% 2189310 sched_debug.cpu.sched_count.max >> 214130 -26.5% 157298 3% sched_debug.cpu.sched_count.stddev >> 489430 2% -16.6% 408309 2% sched_debug.cpu.sched_goidle.avg >> 724333 2% -28.2% 520263 2% sched_debug.cpu.sched_goidle.max >> 457611 -18.1% 374746 3% sched_debug.cpu.sched_goidle.min >> 62957 2% -47.4% 33138 3% sched_debug.cpu.sched_goidle.stddev >> 676053 2% -15.4% 571816 2% sched_debug.cpu.ttwu_local.max >> 42669 3% +22.3% 52198 sched_debug.cpu.ttwu_local.min >> 151873 2% -18.3% 124118 2% sched_debug.cpu.ttwu_local.stddev >> >> >> >> blogbench.write_score >> >> 3300 +-+------------------------------------------------------------------+ >> 3250 +-+ +. .+ +. .+ : : : +. .+ .+.+.+. .| >> |: +. .+ +.+.+.+ + + + : +. : : +. + +.+ + + | >> 3200 +-+ + +.+ + : + + : + + | >> 3150 +-+.+ ++ +.+ | >> 3100 +-+ | >> 3050 +-+ | >> | | >> 3000 +-+ | >> 2950 +-+ O O | >> 2900 +-O O O O | >> 2850 +-+ O O O O O O O OO O O O | >> | O O O O | >> 2800 O-+ O O | >> 2750 +-+------------------------------------------------------------------+ >> >> >> [*] 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. >> >> >> Thanks, >> Xiaolong > _______________________________________________ > LKP mailing list > LKP@lists.01.org > https://lists.01.org/mailman/listinfo/lkp ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [LKP] [lkp-robot] [mm] 9092c71bb7: blogbench.write_score -12.3% regression 2018-06-05 4:58 ` Huang, Ying @ 2018-06-14 1:37 ` Huang, Ying 2018-06-20 3:51 ` Huang, Ying 0 siblings, 1 reply; 11+ messages in thread From: Huang, Ying @ 2018-06-14 1:37 UTC (permalink / raw) To: Josef Bacik; +Cc: David Sterba, Chris Mason, lkp@01.org, linux-btrfs Ping again ... "Huang, Ying" <ying.huang@intel.com> writes: > Ping... > > "Huang, Ying" <ying.huang@intel.com> writes: > >> Hi, Josef, >> >> Do you have time to take a look at the regression? >> >> kernel test robot <xiaolong.ye@intel.com> writes: >> >>> Greeting, >>> >>> FYI, we noticed a -12.3% regression of blogbench.write_score and a +9.6% improvement >>> of blogbench.read_score due to commit: >>> >>> >>> commit: 9092c71bb724dba2ecba849eae69e5c9d39bd3d2 ("mm: use sc->priority for slab shrink targets") >>> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master >>> >>> in testcase: blogbench >>> on test machine: 16 threads Intel(R) Xeon(R) CPU D-1541 @ 2.10GHz with 8G memory >>> with following parameters: >>> >>> disk: 1SSD >>> fs: btrfs >>> cpufreq_governor: performance >>> >>> test-description: Blogbench is a portable filesystem benchmark that tries to reproduce the load of a real-world busy file server. >>> test-url: https://www.pureftpd.org/project/blogbench >>> >>> >>> >>> 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 run job.yaml >>> >>> ========================================================================================= >>> compiler/cpufreq_governor/disk/fs/kconfig/rootfs/tbox_group/testcase: >>> gcc-7/performance/1SSD/btrfs/x86_64-rhel-7.2/debian-x86_64-2016-08-31.cgz/lkp-bdw-de1/blogbench >>> >>> commit: >>> fcb2b0c577 ("mm: show total hugetlb memory consumption in /proc/meminfo") >>> 9092c71bb7 ("mm: use sc->priority for slab shrink targets") >>> >>> fcb2b0c577f145c7 9092c71bb724dba2ecba849eae >>> ---------------- -------------------------- >>> %stddev %change %stddev >>> \ | \ >>> 3256 -12.3% 2854 blogbench.write_score >>> 1235237 2% +9.6% 1354163 blogbench.read_score >>> 28050912 -10.1% 25212230 blogbench.time.file_system_outputs >>> 6481995 3% +25.0% 8105320 2% blogbench.time.involuntary_context_switches >>> 906.00 +13.7% 1030 blogbench.time.percent_of_cpu_this_job_got >>> 2552 +14.0% 2908 blogbench.time.system_time >>> 173.80 +8.4% 188.32 blogbench.time.user_time >>> 19353936 +3.6% 20045728 blogbench.time.voluntary_context_switches >>> 8719514 +13.0% 9850451 softirqs.RCU >>> 2.97 5% -0.7 2.30 3% mpstat.cpu.idle% >>> 24.92 -6.5 18.46 mpstat.cpu.iowait% >>> 0.65 2% +0.1 0.75 mpstat.cpu.soft% >>> 67.76 +6.7 74.45 mpstat.cpu.sys% >>> 50206 -10.7% 44858 vmstat.io.bo >>> 49.25 -9.1% 44.75 2% vmstat.procs.b >>> 224125 -1.8% 220135 vmstat.system.cs >>> 48903 +10.7% 54134 vmstat.system.in >>> 3460654 +10.8% 3834883 meminfo.Active >>> 3380666 +11.0% 3752872 meminfo.Active(file) >>> 1853849 -17.4% 1530415 meminfo.Inactive >>> 1836507 -17.6% 1513054 meminfo.Inactive(file) >>> 551311 -10.3% 494265 meminfo.SReclaimable >>> 196525 -12.6% 171775 meminfo.SUnreclaim >>> 747837 -10.9% 666040 meminfo.Slab >>> 8.904e+08 -24.9% 6.683e+08 cpuidle.C1.time >>> 22971020 -12.8% 20035820 cpuidle.C1.usage >>> 2.518e+08 3% -31.7% 1.72e+08 cpuidle.C1E.time >>> 821393 2% -33.3% 548003 cpuidle.C1E.usage >>> 75460078 2% -23.3% 57903768 2% cpuidle.C3.time >>> 136506 3% -25.3% 101956 3% cpuidle.C3.usage >>> 56892498 4% -23.3% 43608427 4% cpuidle.C6.time >>> 85034 3% -33.9% 56184 3% cpuidle.C6.usage >>> 24373567 -24.5% 18395538 cpuidle.POLL.time >>> 449033 2% -10.8% 400493 cpuidle.POLL.usage >>> 1832 +9.3% 2002 turbostat.Avg_MHz >>> 22967645 -12.8% 20032521 turbostat.C1 >>> 18.43 -4.6 13.85 turbostat.C1% >>> 821328 2% -33.3% 547948 turbostat.C1E >>> 5.21 3% -1.6 3.56 turbostat.C1E% >>> 136377 3% -25.3% 101823 3% turbostat.C3 >>> 1.56 2% -0.4 1.20 3% turbostat.C3% >>> 84404 3% -34.0% 55743 3% turbostat.C6 >>> 1.17 4% -0.3 0.90 4% turbostat.C6% >>> 25.93 -26.2% 19.14 turbostat.CPU%c1 >>> 0.12 3% -19.1% 0.10 9% turbostat.CPU%c3 >>> 14813304 +10.7% 16398388 turbostat.IRQ >>> 38.19 +3.6% 39.56 turbostat.PkgWatt >>> 4.51 +4.5% 4.71 turbostat.RAMWatt >>> 8111200 13% -63.2% 2986242 48% proc-vmstat.compact_daemon_free_scanned >>> 1026719 30% -81.2% 193485 30% proc-vmstat.compact_daemon_migrate_scanned >>> 2444 21% -63.3% 897.50 20% proc-vmstat.compact_daemon_wake >>> 8111200 13% -63.2% 2986242 48% proc-vmstat.compact_free_scanned >>> 755491 32% -81.6% 138856 28% proc-vmstat.compact_isolated >>> 1026719 30% -81.2% 193485 30% proc-vmstat.compact_migrate_scanned >>> 137.75 34% +2.8e+06% 3801062 2% proc-vmstat.kswapd_inodesteal >>> 6749 20% -53.6% 3131 12% proc-vmstat.kswapd_low_wmark_hit_quickly >>> 844991 +11.2% 939487 proc-vmstat.nr_active_file >>> 3900576 -10.5% 3490567 proc-vmstat.nr_dirtied >>> 459789 -17.8% 377930 proc-vmstat.nr_inactive_file >>> 137947 -10.3% 123720 proc-vmstat.nr_slab_reclaimable >>> 49165 -12.6% 42989 proc-vmstat.nr_slab_unreclaimable >>> 1382 11% -26.2% 1020 20% proc-vmstat.nr_writeback >>> 3809266 -10.7% 3403350 proc-vmstat.nr_written >>> 844489 +11.2% 938974 proc-vmstat.nr_zone_active_file >>> 459855 -17.8% 378121 proc-vmstat.nr_zone_inactive_file >>> 7055 18% -52.0% 3389 11% proc-vmstat.pageoutrun >>> 33764911 2% +21.3% 40946445 proc-vmstat.pgactivate >>> 42044161 2% +12.1% 47139065 proc-vmstat.pgdeactivate >>> 92153 20% -69.1% 28514 24% proc-vmstat.pgmigrate_success >>> 15212270 -10.7% 13591573 proc-vmstat.pgpgout >>> 42053817 2% +12.1% 47151755 proc-vmstat.pgrefill >>> 11297 107% +1025.4% 127138 21% proc-vmstat.pgscan_direct >>> 19930162 -24.0% 15141439 proc-vmstat.pgscan_kswapd >>> 19423629 -24.0% 14758807 proc-vmstat.pgsteal_kswapd >>> 10868768 +184.8% 30950752 proc-vmstat.slabs_scanned >> >> The slab scan number increased a lot. >> >>> 3361780 3% -22.9% 2593327 3% proc-vmstat.workingset_activate >>> 4994722 2% -43.2% 2835020 2% proc-vmstat.workingset_refault >>> 316427 -9.3% 286844 slabinfo.Acpi-Namespace.active_objs >>> 3123 -9.4% 2829 slabinfo.Acpi-Namespace.active_slabs >>> 318605 -9.4% 288623 slabinfo.Acpi-Namespace.num_objs >>> 3123 -9.4% 2829 slabinfo.Acpi-Namespace.num_slabs >>> 220514 -40.7% 130747 slabinfo.btrfs_delayed_node.active_objs >>> 9751 -25.3% 7283 slabinfo.btrfs_delayed_node.active_slabs >>> 263293 -25.3% 196669 slabinfo.btrfs_delayed_node.num_objs >>> 9751 -25.3% 7283 slabinfo.btrfs_delayed_node.num_slabs >>> 6383 8% -12.0% 5615 2% slabinfo.btrfs_delayed_ref_head.num_objs >>> 9496 +15.5% 10969 slabinfo.btrfs_extent_buffer.active_objs >>> 9980 +20.5% 12022 slabinfo.btrfs_extent_buffer.num_objs >>> 260933 -10.7% 233136 slabinfo.btrfs_extent_map.active_objs >>> 9392 -10.6% 8396 slabinfo.btrfs_extent_map.active_slabs >>> 263009 -10.6% 235107 slabinfo.btrfs_extent_map.num_objs >>> 9392 -10.6% 8396 slabinfo.btrfs_extent_map.num_slabs >>> 271938 -10.3% 243802 slabinfo.btrfs_inode.active_objs >>> 9804 -10.6% 8768 slabinfo.btrfs_inode.active_slabs >>> 273856 -10.4% 245359 slabinfo.btrfs_inode.num_objs >>> 9804 -10.6% 8768 slabinfo.btrfs_inode.num_slabs >>> 7085 5% -5.5% 6692 2% slabinfo.btrfs_path.num_objs >>> 311936 -16.4% 260797 slabinfo.dentry.active_objs >>> 7803 -9.6% 7058 slabinfo.dentry.active_slabs >>> 327759 -9.6% 296439 slabinfo.dentry.num_objs >>> 7803 -9.6% 7058 slabinfo.dentry.num_slabs >>> 2289 -23.3% 1755 6% slabinfo.proc_inode_cache.active_objs >>> 2292 -19.0% 1856 6% slabinfo.proc_inode_cache.num_objs >>> 261546 -12.3% 229485 slabinfo.radix_tree_node.active_objs >>> 9404 -11.9% 8288 slabinfo.radix_tree_node.active_slabs >>> 263347 -11.9% 232089 slabinfo.radix_tree_node.num_objs >>> 9404 -11.9% 8288 slabinfo.radix_tree_node.num_slabs >> >> The slab size decreased with the new commit. >> >> From perf-profile result, >> >> 26.81 ± 2% -6.5 20.35 ± 2% perf-profile.calltrace.cycles-pp.secondary_startup_64 >> 24.48 ± 2% -5.8 18.73 perf-profile.calltrace.cycles-pp.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64 >> 24.48 ± 2% -5.8 18.73 perf-profile.calltrace.cycles-pp.cpu_startup_entry.start_secondary.secondary_startup_64 >> 24.48 ± 2% -5.8 18.73 perf-profile.calltrace.cycles-pp.start_secondary.secondary_startup_64 >> 22.80 ± 2% -5.5 17.30 ± 2% perf-profile.calltrace.cycles-pp.cpuidle_enter_state.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64 >> 20.20 ± 2% -4.3 15.85 perf-profile.calltrace.cycles-pp.intel_idle.cpuidle_enter_state.do_idle.cpu_startup_entry.start_secondary >> 23.03 ± 2% -2.6 20.42 perf-profile.calltrace.cycles-pp.sys_rename.entry_SYSCALL_64_fastpath >> 17.02 ± 2% -1.8 15.17 ± 2% perf-profile.calltrace.cycles-pp.btrfs_rename.vfs_rename.sys_rename.entry_SYSCALL_64_fastpath >> 17.03 ± 2% -1.8 15.19 ± 2% perf-profile.calltrace.cycles-pp.vfs_rename.sys_rename.entry_SYSCALL_64_fastpath >> 13.70 ± 2% -1.2 12.47 ± 3% perf-profile.calltrace.cycles-pp.__btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename.entry_SYSCALL_64_fastpath >> 5.06 ± 3% -1.1 3.96 ± 2% perf-profile.calltrace.cycles-pp.btrfs_async_run_delayed_root.normal_work_helper.process_one_work.worker_thread.kthread >> 7.37 ± 4% -0.9 6.49 ± 2% perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.finish_wait.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot >> 7.32 ± 4% -0.9 6.45 ± 2% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.finish_wait.btrfs_tree_lock.btrfs_lock_root_node >> 1.54 ± 4% -0.7 0.81 ± 7% perf-profile.calltrace.cycles-pp.poll_idle.cpuidle_enter_state.do_idle.cpu_startup_entry.start_secondary >> 2.33 ± 2% -0.7 1.62 ± 4% perf-profile.calltrace.cycles-pp.do_idle.cpu_startup_entry.start_kernel.secondary_startup_64 >> 2.33 ± 2% -0.7 1.62 ± 4% perf-profile.calltrace.cycles-pp.cpu_startup_entry.start_kernel.secondary_startup_64 >> 2.33 ± 2% -0.7 1.62 ± 4% perf-profile.calltrace.cycles-pp.start_kernel.secondary_startup_64 >> 2.23 ± 3% -0.7 1.53 ± 4% perf-profile.calltrace.cycles-pp.cpuidle_enter_state.do_idle.cpu_startup_entry.start_kernel.secondary_startup_64 >> 2.13 ± 2% -0.7 1.46 ± 4% perf-profile.calltrace.cycles-pp.intel_idle.cpuidle_enter_state.do_idle.cpu_startup_entry.start_kernel >> 5.59 -0.7 4.94 ± 3% perf-profile.calltrace.cycles-pp.__dentry_kill.dput.sys_rename.entry_SYSCALL_64_fastpath >> 5.60 -0.7 4.94 ± 3% perf-profile.calltrace.cycles-pp.dput.sys_rename.entry_SYSCALL_64_fastpath >> 6.96 -0.7 6.31 ± 3% perf-profile.calltrace.cycles-pp.btrfs_del_inode_ref.__btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename >> 5.59 -0.7 4.94 ± 3% perf-profile.calltrace.cycles-pp.evict.__dentry_kill.dput.sys_rename.entry_SYSCALL_64_fastpath >> 5.58 -0.6 4.94 ± 3% perf-profile.calltrace.cycles-pp.btrfs_evict_inode.evict.__dentry_kill.dput.sys_rename >> 6.94 ± 2% -0.6 6.30 ± 3% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_del_inode_ref.__btrfs_unlink_inode.btrfs_rename.vfs_rename >> 6.66 ± 4% -0.6 6.10 ± 3% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_rename.vfs_rename >> 6.66 ± 4% -0.6 6.10 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename >> 3.38 ± 3% -0.5 2.84 ± 3% > perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_delete_delayed_items.btrfs_async_run_delayed_root.normal_work_helper.process_one_work >> 3.40 ± 3% -0.5 2.86 ± 3% perf-profile.calltrace.cycles-pp.btrfs_delete_delayed_items.btrfs_async_run_delayed_root.normal_work_helper.process_one_work.worker_thread >> 7.30 ± 4% -0.4 6.86 ± 2% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item.__btrfs_unlink_inode >> 1.18 ± 4% -0.4 0.76 ± 2% perf-profile.calltrace.cycles-pp.__btrfs_update_delayed_inode.btrfs_async_run_delayed_root.normal_work_helper.process_one_work.worker_thread >> 1.16 ± 4% -0.4 0.74 ± 2% > perf-profile.calltrace.cycles-pp.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_async_run_delayed_root.normal_work_helper.process_one_work >> 5.96 ± 2% -0.4 5.54 ± 3% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_inode_ref.__btrfs_unlink_inode >> 5.96 ± 2% -0.4 5.54 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_inode_ref.__btrfs_unlink_inode.btrfs_rename >> 1.16 ± 4% -0.4 0.74 ± 2% > perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_async_run_delayed_root.normal_work_helper >> 2.95 ± 3% -0.4 2.54 ± 3% > perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items.btrfs_async_run_delayed_root >> 2.95 ± 3% -0.4 2.54 ± 3% > perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items.btrfs_async_run_delayed_root.normal_work_helper >> 0.94 ± 2% -0.3 0.59 ± 4% perf-profile.calltrace.cycles-pp.shrink_inactive_list.shrink_node_memcg.shrink_node.kswapd.kthread >> 1.25 -0.3 0.91 ± 2% perf-profile.calltrace.cycles-pp.shrink_node_memcg.shrink_node.kswapd.kthread.ret_from_fork >> 0.84 ± 2% -0.3 0.52 ± 3% perf-profile.calltrace.cycles-pp.shrink_page_list.shrink_inactive_list.shrink_node_memcg.shrink_node.kswapd >> 1.01 -0.3 0.71 ± 3% perf-profile.calltrace.cycles-pp.btrfs_create.path_openat.do_filp_open.do_sys_open.entry_SYSCALL_64_fastpath >> 0.58 ± 4% -0.2 0.34 ± 70% perf-profile.calltrace.cycles-pp.queued_write_lock_slowpath.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items >> 1.73 ± 2% -0.2 1.50 ± 2% perf-profile.calltrace.cycles-pp.finish_wait.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_inode_ref >> 1.67 ± 4% -0.2 1.44 ± 3% perf-profile.calltrace.cycles-pp.finish_wait.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item >> 1.89 -0.2 1.67 ± 2% perf-profile.calltrace.cycles-pp.btrfs_commit_inode_delayed_inode.btrfs_evict_inode.evict.__dentry_kill.dput >> 1.88 -0.2 1.66 ± 2% perf-profile.calltrace.cycles-pp.__btrfs_update_delayed_inode.btrfs_commit_inode_delayed_inode.btrfs_evict_inode.evict.__dentry_kill >> 1.90 -0.2 1.68 ± 2% perf-profile.calltrace.cycles-pp.btrfs_truncate_inode_items.btrfs_evict_inode.evict.__dentry_kill.dput >> 1.78 -0.2 1.57 ± 5% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_del_orphan_item.btrfs_orphan_del.btrfs_evict_inode.evict >> 1.88 -0.2 1.67 ± 2% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_truncate_inode_items.btrfs_evict_inode.evict.__dentry_kill >> 1.78 -0.2 1.57 ± 5% perf-profile.calltrace.cycles-pp.btrfs_orphan_del.btrfs_evict_inode.evict.__dentry_kill.dput >> 1.78 -0.2 1.57 ± 5% perf-profile.calltrace.cycles-pp.btrfs_del_orphan_item.btrfs_orphan_del.btrfs_evict_inode.evict.__dentry_kill >> 1.87 -0.2 1.66 ± 2% > perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_commit_inode_delayed_inode.btrfs_evict_inode >> 1.87 -0.2 1.66 ± 2% perf-profile.calltrace.cycles-pp.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_commit_inode_delayed_inode.btrfs_evict_inode.evict >> 0.75 ± 11% -0.2 0.57 ± 7% > perf-profile.calltrace.cycles-pp.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent.__btrfs_drop_extents >> 0.75 ± 11% -0.2 0.57 ± 7% > perf-profile.calltrace.cycles-pp.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent.__btrfs_drop_extents.insert_reserved_file_extent >> 1.39 ± 3% -0.2 1.23 ± 4% perf-profile.calltrace.cycles-pp.prepare_to_wait_event.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items >> 1.84 -0.2 1.68 ± 3% perf-profile.calltrace.cycles-pp.__btrfs_unlink_inode.btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename >> 1.84 -0.2 1.68 ± 3% perf-profile.calltrace.cycles-pp.btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename.entry_SYSCALL_64_fastpath >> 1.62 -0.2 1.46 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_truncate_inode_items.btrfs_evict_inode.evict >> 0.87 ± 5% -0.2 0.72 ± 5% perf-profile.calltrace.cycles-pp.finish_wait.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items >> 1.81 -0.2 1.66 ± 3% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_unlink_inode.btrfs_rename >> 1.81 -0.2 1.66 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_unlink_inode.btrfs_rename.vfs_rename >> 1.62 -0.2 1.46 ± 3% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_truncate_inode_items.btrfs_evict_inode >> 1.69 ± 2% -0.1 1.55 ± 2% perf-profile.calltrace.cycles-pp.end_bio_extent_readpage.normal_work_helper.process_one_work.worker_thread.kthread >> 1.58 -0.1 1.44 ± 2% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_inode.__btrfs_update_delayed_inode >> 1.58 -0.1 1.44 ± 2% > perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_commit_inode_delayed_inode >> 2.89 ± 3% -0.1 2.77 ± 2% > perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath.queued_write_lock_slowpath.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot >> 1.51 -0.1 1.39 ± 5% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_orphan_item.btrfs_orphan_del >> 1.51 -0.1 1.39 ± 5% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_orphan_item.btrfs_orphan_del.btrfs_evict_inode >> 0.94 -0.1 0.82 perf-profile.calltrace.cycles-pp.schedule_idle.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64 >> 0.93 -0.1 0.81 perf-profile.calltrace.cycles-pp.__schedule.schedule_idle.do_idle.cpu_startup_entry.start_secondary >> 1.52 -0.1 1.44 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_unlink_inode >> 0.71 ± 3% -0.1 0.66 ± 2% perf-profile.calltrace.cycles-pp.prepare_to_wait_event.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_truncate_inode_items >> 0.60 ± 2% -0.0 0.56 ± 2% perf-profile.calltrace.cycles-pp.ttwu_do_activate.try_to_wake_up.autoremove_wake_function.__wake_up_common.wake_up_page_bit >> 1.61 ± 2% +0.1 1.67 ± 2% perf-profile.calltrace.cycles-pp.kswapd.kthread.ret_from_fork >> 1.61 ± 2% +0.1 1.67 ± 2% perf-profile.calltrace.cycles-pp.shrink_node.kswapd.kthread.ret_from_fork >> 0.55 +0.1 0.68 ± 4% perf-profile.calltrace.cycles-pp.find_get_entry.pagecache_get_page.generic_file_read_iter.__vfs_read.vfs_read >> 0.57 +0.1 0.70 ± 4% perf-profile.calltrace.cycles-pp.pagecache_get_page.generic_file_read_iter.__vfs_read.vfs_read.sys_read >> 0.59 ± 3% +0.3 0.87 perf-profile.calltrace.cycles-pp.__account_scheduler_latency.enqueue_entity.enqueue_task_fair.ttwu_do_activate.try_to_wake_up >> 1.29 ± 3% +0.3 1.58 perf-profile.calltrace.cycles-pp.enqueue_task_fair.ttwu_do_activate.try_to_wake_up.autoremove_wake_function.__wake_up_common >> 0.54 +0.3 0.83 ± 2% > perf-profile.calltrace.cycles-pp.queued_read_lock_slowpath.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent >> 0.84 +0.3 1.18 ± 3% > perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.prepare_to_wait_event.btrfs_tree_read_lock.btrfs_read_lock_root_node >> 0.71 ± 4% +0.3 1.05 perf-profile.calltrace.cycles-pp.ttwu_do_activate.try_to_wake_up.autoremove_wake_function.__wake_up_common.__wake_up_common_lock >> 0.86 +0.4 1.21 ± 3% > perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.prepare_to_wait_event.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot >> 0.90 +0.4 1.28 ± 3% perf-profile.calltrace.cycles-pp.prepare_to_wait_event.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item >> 0.35 ± 71% +0.4 0.79 ± 3% > perf-profile.calltrace.cycles-pp.queued_read_lock_slowpath.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_inode >> 0.98 ± 4% +0.4 1.41 ± 2% perf-profile.calltrace.cycles-pp.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget >> 0.98 ± 5% +0.4 1.43 ± 2% perf-profile.calltrace.cycles-pp.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget.btrfs_lookup_dentry >> 0.94 ± 3% +0.4 1.39 perf-profile.calltrace.cycles-pp.try_to_wake_up.autoremove_wake_function.__wake_up_common.__wake_up_common_lock.btrfs_clear_path_blocking >> 0.96 ± 3% +0.5 1.42 perf-profile.calltrace.cycles-pp.autoremove_wake_function.__wake_up_common.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot >> 1.05 ± 2% +0.5 1.52 ± 4% perf-profile.calltrace.cycles-pp.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent >> 1.06 ± 3% +0.5 1.53 ± 4% perf-profile.calltrace.cycles-pp.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent.__do_readpage >> 0.97 ± 3% +0.5 1.45 perf-profile.calltrace.cycles-pp.__wake_up_common.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item >> 0.62 ± 4% +0.5 1.13 > perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath.queued_read_lock_slowpath.btrfs_clear_lock_blocking_rw.btrfs_clear_path_blocking.btrfs_search_slot >> 0.65 ± 4% +0.5 1.19 > perf-profile.calltrace.cycles-pp.queued_read_lock_slowpath.btrfs_clear_lock_blocking_rw.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item >> 0.17 ±141% +0.6 0.73 perf-profile.calltrace.cycles-pp.save_stack_trace_tsk.__account_scheduler_latency.enqueue_entity.enqueue_task_fair.ttwu_do_activate >> 0.00 +0.6 0.57 ± 2% perf-profile.calltrace.cycles-pp.task_work_run.exit_to_usermode_loop.syscall_return_slowpath.entry_SYSCALL_64_fastpath >> 0.00 +0.6 0.59 ± 2% perf-profile.calltrace.cycles-pp.exit_to_usermode_loop.syscall_return_slowpath.entry_SYSCALL_64_fastpath >> 0.00 +0.6 0.60 ± 2% perf-profile.calltrace.cycles-pp.syscall_return_slowpath.entry_SYSCALL_64_fastpath >> 0.56 ± 4% +0.6 1.17 > perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_file_extent >> 1.17 +0.6 1.78 ± 2% perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.finish_wait.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot >> 1.16 +0.6 1.77 ± 2% > perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.finish_wait.btrfs_tree_read_lock.btrfs_read_lock_root_node >> 1.18 +0.6 1.80 ± 2% perf-profile.calltrace.cycles-pp.finish_wait.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item >> 0.57 ± 7% +0.6 1.21 ± 2% perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_inode >> 0.75 ± 4% +0.6 1.38 > perf-profile.calltrace.cycles-pp.btrfs_clear_lock_blocking_rw.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry >> 1.95 +0.7 2.61 perf-profile.calltrace.cycles-pp.copy_user_enhanced_fast_string.copyout.copy_page_to_iter.generic_file_read_iter.__vfs_read >> 1.96 +0.7 2.62 perf-profile.calltrace.cycles-pp.copyout.copy_page_to_iter.generic_file_read_iter.__vfs_read.vfs_read >> 0.00 +0.7 0.70 perf-profile.calltrace.cycles-pp.__save_stack_trace.save_stack_trace_tsk.__account_scheduler_latency.enqueue_entity.enqueue_task_fair >> 2.10 +0.7 2.80 perf-profile.calltrace.cycles-pp.copy_page_to_iter.generic_file_read_iter.__vfs_read.vfs_read.sys_read >> 0.85 ± 5% +0.7 1.58 perf-profile.calltrace.cycles-pp.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent >> 0.84 ± 6% +0.7 1.57 perf-profile.calltrace.cycles-pp.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget >> 0.00 +0.8 0.75 ± 4% perf-profile.calltrace.cycles-pp.super_cache_scan.shrink_slab.shrink_node.kswapd.kthread >> 0.00 +0.8 0.75 ± 5% perf-profile.calltrace.cycles-pp.shrink_slab.shrink_node.kswapd.kthread.ret_from_fork >> 1.07 ± 5% +0.9 1.98 ± 2% perf-profile.calltrace.cycles-pp.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent.__do_readpage >> 1.09 ± 7% +0.9 2.01 ± 2% perf-profile.calltrace.cycles-pp.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget.btrfs_lookup_dentry >> 4.90 ± 2% +1.3 6.19 perf-profile.calltrace.cycles-pp.__do_page_cache_readahead.ondemand_readahead.generic_file_read_iter.__vfs_read.vfs_read >> 4.90 ± 2% +1.3 6.20 perf-profile.calltrace.cycles-pp.ondemand_readahead.generic_file_read_iter.__vfs_read.vfs_read.sys_read >> 4.44 ± 2% +1.3 5.74 perf-profile.calltrace.cycles-pp.extent_readpages.__do_page_cache_readahead.ondemand_readahead.generic_file_read_iter.__vfs_read >> 2.87 ± 3% +1.4 4.29 perf-profile.calltrace.cycles-pp.__extent_readpages.extent_readpages.__do_page_cache_readahead.ondemand_readahead.generic_file_read_iter >> 2.22 ± 4% +1.4 3.65 perf-profile.calltrace.cycles-pp.btrfs_lookup_file_extent.btrfs_get_extent.__do_readpage.__extent_readpages.extent_readpages >> 2.22 ± 4% +1.4 3.65 perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent.__do_readpage.__extent_readpages >> 2.27 ± 6% +1.4 3.72 perf-profile.calltrace.cycles-pp.btrfs_lookup_inode.btrfs_iget.btrfs_lookup_dentry.btrfs_lookup.path_openat >> 2.27 ± 6% +1.4 3.72 perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget.btrfs_lookup_dentry.btrfs_lookup >> 2.72 ± 3% +1.4 4.17 perf-profile.calltrace.cycles-pp.__do_readpage.__extent_readpages.extent_readpages.__do_page_cache_readahead.ondemand_readahead >> 2.33 ± 4% +1.5 3.79 perf-profile.calltrace.cycles-pp.btrfs_get_extent.__do_readpage.__extent_readpages.extent_readpages.__do_page_cache_readahead >> 2.38 +1.6 3.95 > perf-profile.calltrace.cycles-pp.queued_read_lock_slowpath.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item >> 2.63 ± 6% +1.7 4.33 perf-profile.calltrace.cycles-pp.btrfs_iget.btrfs_lookup_dentry.btrfs_lookup.path_openat.do_filp_open >> 1.68 ± 4% +2.1 3.79 > perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item >> 8.66 +2.2 10.90 perf-profile.calltrace.cycles-pp.generic_file_read_iter.__vfs_read.vfs_read.sys_read.entry_SYSCALL_64_fastpath >> 8.71 +2.3 10.97 perf-profile.calltrace.cycles-pp.__vfs_read.vfs_read.sys_read.entry_SYSCALL_64_fastpath >> 8.85 +2.3 11.15 perf-profile.calltrace.cycles-pp.vfs_read.sys_read.entry_SYSCALL_64_fastpath >> 9.02 +2.3 11.33 perf-profile.calltrace.cycles-pp.sys_read.entry_SYSCALL_64_fastpath >> 2.71 ± 3% +2.6 5.35 perf-profile.calltrace.cycles-pp.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry >> 4.63 +2.7 7.29 perf-profile.calltrace.cycles-pp.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry >> 4.65 +2.7 7.32 perf-profile.calltrace.cycles-pp.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry.btrfs_lookup >> 1.94 ± 2% +2.8 4.77 > perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath.queued_read_lock_slowpath.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot >> 2.81 ± 4% +3.3 6.15 > perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot >> 3.58 ± 2% +3.4 6.98 perf-profile.calltrace.cycles-pp.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry.btrfs_lookup >> 8.69 +6.3 15.03 perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry.btrfs_lookup.path_openat >> 8.75 +6.4 15.12 perf-profile.calltrace.cycles-pp.btrfs_lookup_dir_item.btrfs_lookup_dentry.btrfs_lookup.path_openat.do_filp_open >> 47.34 +8.1 55.46 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_fastpath >> 11.47 +8.1 19.60 perf-profile.calltrace.cycles-pp.btrfs_lookup.path_openat.do_filp_open.do_sys_open.entry_SYSCALL_64_fastpath >> 11.46 +8.1 19.59 perf-profile.calltrace.cycles-pp.btrfs_lookup_dentry.btrfs_lookup.path_openat.do_filp_open.do_sys_open >> 13.90 +8.3 22.20 perf-profile.calltrace.cycles-pp.path_openat.do_filp_open.do_sys_open.entry_SYSCALL_64_fastpath >> 13.91 +8.3 22.22 perf-profile.calltrace.cycles-pp.do_filp_open.do_sys_open.entry_SYSCALL_64_fastpath >> 14.13 +8.4 22.52 perf-profile.calltrace.cycles-pp.do_sys_open.entry_SYSCALL_64_fastpath >> >> The cycles for dentry lookup increased much too. Is this the reason why >> write score decreased? >> >> If you need more information, please let me know. >> >> Best Regards, >> Huang, Ying >> >>> 1140424 12% +40.2% 1598980 14% sched_debug.cfs_rq:/.MIN_vruntime.max >>> 790.55 +13.0% 893.20 3% sched_debug.cfs_rq:/.exec_clock.stddev >>> 1140425 12% +40.2% 1598982 14% sched_debug.cfs_rq:/.max_vruntime.max >>> 0.83 10% +21.5% 1.00 8% sched_debug.cfs_rq:/.nr_running.avg >>> 3.30 99% +266.3% 12.09 13% sched_debug.cfs_rq:/.removed.load_avg.avg >>> 153.02 97% +266.6% 560.96 13% sched_debug.cfs_rq:/.removed.runnable_sum.avg >>> 569.93 102% +173.2% 1556 14% sched_debug.cfs_rq:/.removed.runnable_sum.stddev >>> 1.42 60% +501.5% 8.52 34% sched_debug.cfs_rq:/.removed.util_avg.avg >>> 19.88 59% +288.9% 77.29 16% sched_debug.cfs_rq:/.removed.util_avg.max >>> 5.05 58% +342.3% 22.32 22% sched_debug.cfs_rq:/.removed.util_avg.stddev >>> 791.44 3% +47.7% 1168 8% sched_debug.cfs_rq:/.util_avg.avg >>> 1305 6% +33.2% 1738 5% sched_debug.cfs_rq:/.util_avg.max >>> 450.25 11% +66.2% 748.17 14% sched_debug.cfs_rq:/.util_avg.min >>> 220.82 8% +21.1% 267.46 5% sched_debug.cfs_rq:/.util_avg.stddev >>> 363118 11% -23.8% 276520 11% sched_debug.cpu.avg_idle.avg >>> 726003 8% -30.8% 502313 4% sched_debug.cpu.avg_idle.max >>> 202629 3% -32.2% 137429 18% sched_debug.cpu.avg_idle.stddev >>> 31.96 28% +54.6% 49.42 14% sched_debug.cpu.cpu_load[3].min >>> 36.21 25% +64.0% 59.38 6% sched_debug.cpu.cpu_load[4].min >>> 1007 5% +20.7% 1216 7% sched_debug.cpu.curr->pid.avg >>> 4.50 5% +14.8% 5.17 5% sched_debug.cpu.nr_running.max >>> 2476195 -11.8% 2185022 sched_debug.cpu.nr_switches.max >>> 212888 -26.6% 156172 3% sched_debug.cpu.nr_switches.stddev >>> 3570 2% -58.7% 1474 2% sched_debug.cpu.nr_uninterruptible.max >>> -803.67 -28.7% -573.38 sched_debug.cpu.nr_uninterruptible.min >>> 1004 2% -50.4% 498.55 3% sched_debug.cpu.nr_uninterruptible.stddev >>> 2478809 -11.7% 2189310 sched_debug.cpu.sched_count.max >>> 214130 -26.5% 157298 3% sched_debug.cpu.sched_count.stddev >>> 489430 2% -16.6% 408309 2% sched_debug.cpu.sched_goidle.avg >>> 724333 2% -28.2% 520263 2% sched_debug.cpu.sched_goidle.max >>> 457611 -18.1% 374746 3% sched_debug.cpu.sched_goidle.min >>> 62957 2% -47.4% 33138 3% sched_debug.cpu.sched_goidle.stddev >>> 676053 2% -15.4% 571816 2% sched_debug.cpu.ttwu_local.max >>> 42669 3% +22.3% 52198 sched_debug.cpu.ttwu_local.min >>> 151873 2% -18.3% 124118 2% sched_debug.cpu.ttwu_local.stddev >>> >>> >>> >>> blogbench.write_score >>> >>> 3300 +-+------------------------------------------------------------------+ >>> 3250 +-+ +. .+ +. .+ : : : +. .+ .+.+.+. .| >>> |: +. .+ +.+.+.+ + + + : +. : : +. + +.+ + + | >>> 3200 +-+ + +.+ + : + + : + + | >>> 3150 +-+.+ ++ +.+ | >>> 3100 +-+ | >>> 3050 +-+ | >>> | | >>> 3000 +-+ | >>> 2950 +-+ O O | >>> 2900 +-O O O O | >>> 2850 +-+ O O O O O O O OO O O O | >>> | O O O O | >>> 2800 O-+ O O | >>> 2750 +-+------------------------------------------------------------------+ >>> >>> >>> [*] 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. >>> >>> >>> Thanks, >>> Xiaolong >> _______________________________________________ >> LKP mailing list >> LKP@lists.01.org >> https://lists.01.org/mailman/listinfo/lkp ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [LKP] [lkp-robot] [mm] 9092c71bb7: blogbench.write_score -12.3% regression 2018-06-14 1:37 ` Huang, Ying @ 2018-06-20 3:51 ` Huang, Ying 2018-06-20 12:38 ` Chris Mason 0 siblings, 1 reply; 11+ messages in thread From: Huang, Ying @ 2018-06-20 3:51 UTC (permalink / raw) To: Josef Bacik; +Cc: David Sterba, Chris Mason, lkp@01.org, linux-btrfs Ping * 3 "Huang, Ying" <ying.huang@intel.com> writes: > Ping again ... > > "Huang, Ying" <ying.huang@intel.com> writes: > >> Ping... >> >> "Huang, Ying" <ying.huang@intel.com> writes: >> >>> Hi, Josef, >>> >>> Do you have time to take a look at the regression? >>> >>> kernel test robot <xiaolong.ye@intel.com> writes: >>> >>>> Greeting, >>>> >>>> FYI, we noticed a -12.3% regression of blogbench.write_score and a +9.6% improvement >>>> of blogbench.read_score due to commit: >>>> >>>> >>>> commit: 9092c71bb724dba2ecba849eae69e5c9d39bd3d2 ("mm: use sc->priority for slab shrink targets") >>>> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master >>>> >>>> in testcase: blogbench >>>> on test machine: 16 threads Intel(R) Xeon(R) CPU D-1541 @ 2.10GHz with 8G memory >>>> with following parameters: >>>> >>>> disk: 1SSD >>>> fs: btrfs >>>> cpufreq_governor: performance >>>> >>>> test-description: Blogbench is a portable filesystem benchmark that tries to reproduce the load of a real-world busy file server. >>>> test-url: https://www.pureftpd.org/project/blogbench >>>> >>>> >>>> >>>> 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 run job.yaml >>>> >>>> ========================================================================================= >>>> compiler/cpufreq_governor/disk/fs/kconfig/rootfs/tbox_group/testcase: >>>> gcc-7/performance/1SSD/btrfs/x86_64-rhel-7.2/debian-x86_64-2016-08-31.cgz/lkp-bdw-de1/blogbench >>>> >>>> commit: >>>> fcb2b0c577 ("mm: show total hugetlb memory consumption in /proc/meminfo") >>>> 9092c71bb7 ("mm: use sc->priority for slab shrink targets") >>>> >>>> fcb2b0c577f145c7 9092c71bb724dba2ecba849eae >>>> ---------------- -------------------------- >>>> %stddev %change %stddev >>>> \ | \ >>>> 3256 -12.3% 2854 blogbench.write_score >>>> 1235237 2% +9.6% 1354163 blogbench.read_score >>>> 28050912 -10.1% 25212230 blogbench.time.file_system_outputs >>>> 6481995 3% +25.0% 8105320 2% blogbench.time.involuntary_context_switches >>>> 906.00 +13.7% 1030 blogbench.time.percent_of_cpu_this_job_got >>>> 2552 +14.0% 2908 blogbench.time.system_time >>>> 173.80 +8.4% 188.32 blogbench.time.user_time >>>> 19353936 +3.6% 20045728 blogbench.time.voluntary_context_switches >>>> 8719514 +13.0% 9850451 softirqs.RCU >>>> 2.97 5% -0.7 2.30 3% mpstat.cpu.idle% >>>> 24.92 -6.5 18.46 mpstat.cpu.iowait% >>>> 0.65 2% +0.1 0.75 mpstat.cpu.soft% >>>> 67.76 +6.7 74.45 mpstat.cpu.sys% >>>> 50206 -10.7% 44858 vmstat.io.bo >>>> 49.25 -9.1% 44.75 2% vmstat.procs.b >>>> 224125 -1.8% 220135 vmstat.system.cs >>>> 48903 +10.7% 54134 vmstat.system.in >>>> 3460654 +10.8% 3834883 meminfo.Active >>>> 3380666 +11.0% 3752872 meminfo.Active(file) >>>> 1853849 -17.4% 1530415 meminfo.Inactive >>>> 1836507 -17.6% 1513054 meminfo.Inactive(file) >>>> 551311 -10.3% 494265 meminfo.SReclaimable >>>> 196525 -12.6% 171775 meminfo.SUnreclaim >>>> 747837 -10.9% 666040 meminfo.Slab >>>> 8.904e+08 -24.9% 6.683e+08 cpuidle.C1.time >>>> 22971020 -12.8% 20035820 cpuidle.C1.usage >>>> 2.518e+08 3% -31.7% 1.72e+08 cpuidle.C1E.time >>>> 821393 2% -33.3% 548003 cpuidle.C1E.usage >>>> 75460078 2% -23.3% 57903768 2% cpuidle.C3.time >>>> 136506 3% -25.3% 101956 3% cpuidle.C3.usage >>>> 56892498 4% -23.3% 43608427 4% cpuidle.C6.time >>>> 85034 3% -33.9% 56184 3% cpuidle.C6.usage >>>> 24373567 -24.5% 18395538 cpuidle.POLL.time >>>> 449033 2% -10.8% 400493 cpuidle.POLL.usage >>>> 1832 +9.3% 2002 turbostat.Avg_MHz >>>> 22967645 -12.8% 20032521 turbostat.C1 >>>> 18.43 -4.6 13.85 turbostat.C1% >>>> 821328 2% -33.3% 547948 turbostat.C1E >>>> 5.21 3% -1.6 3.56 turbostat.C1E% >>>> 136377 3% -25.3% 101823 3% turbostat.C3 >>>> 1.56 2% -0.4 1.20 3% turbostat.C3% >>>> 84404 3% -34.0% 55743 3% turbostat.C6 >>>> 1.17 4% -0.3 0.90 4% turbostat.C6% >>>> 25.93 -26.2% 19.14 turbostat.CPU%c1 >>>> 0.12 3% -19.1% 0.10 9% turbostat.CPU%c3 >>>> 14813304 +10.7% 16398388 turbostat.IRQ >>>> 38.19 +3.6% 39.56 turbostat.PkgWatt >>>> 4.51 +4.5% 4.71 turbostat.RAMWatt >>>> 8111200 13% -63.2% 2986242 48% proc-vmstat.compact_daemon_free_scanned >>>> 1026719 30% -81.2% 193485 30% proc-vmstat.compact_daemon_migrate_scanned >>>> 2444 21% -63.3% 897.50 20% proc-vmstat.compact_daemon_wake >>>> 8111200 13% -63.2% 2986242 48% proc-vmstat.compact_free_scanned >>>> 755491 32% -81.6% 138856 28% proc-vmstat.compact_isolated >>>> 1026719 30% -81.2% 193485 30% proc-vmstat.compact_migrate_scanned >>>> 137.75 34% +2.8e+06% 3801062 2% proc-vmstat.kswapd_inodesteal >>>> 6749 20% -53.6% 3131 12% proc-vmstat.kswapd_low_wmark_hit_quickly >>>> 844991 +11.2% 939487 proc-vmstat.nr_active_file >>>> 3900576 -10.5% 3490567 proc-vmstat.nr_dirtied >>>> 459789 -17.8% 377930 proc-vmstat.nr_inactive_file >>>> 137947 -10.3% 123720 proc-vmstat.nr_slab_reclaimable >>>> 49165 -12.6% 42989 proc-vmstat.nr_slab_unreclaimable >>>> 1382 11% -26.2% 1020 20% proc-vmstat.nr_writeback >>>> 3809266 -10.7% 3403350 proc-vmstat.nr_written >>>> 844489 +11.2% 938974 proc-vmstat.nr_zone_active_file >>>> 459855 -17.8% 378121 proc-vmstat.nr_zone_inactive_file >>>> 7055 18% -52.0% 3389 11% proc-vmstat.pageoutrun >>>> 33764911 2% +21.3% 40946445 proc-vmstat.pgactivate >>>> 42044161 2% +12.1% 47139065 proc-vmstat.pgdeactivate >>>> 92153 20% -69.1% 28514 24% proc-vmstat.pgmigrate_success >>>> 15212270 -10.7% 13591573 proc-vmstat.pgpgout >>>> 42053817 2% +12.1% 47151755 proc-vmstat.pgrefill >>>> 11297 107% +1025.4% 127138 21% proc-vmstat.pgscan_direct >>>> 19930162 -24.0% 15141439 proc-vmstat.pgscan_kswapd >>>> 19423629 -24.0% 14758807 proc-vmstat.pgsteal_kswapd >>>> 10868768 +184.8% 30950752 proc-vmstat.slabs_scanned >>> >>> The slab scan number increased a lot. >>> >>>> 3361780 3% -22.9% 2593327 3% proc-vmstat.workingset_activate >>>> 4994722 2% -43.2% 2835020 2% proc-vmstat.workingset_refault >>>> 316427 -9.3% 286844 slabinfo.Acpi-Namespace.active_objs >>>> 3123 -9.4% 2829 slabinfo.Acpi-Namespace.active_slabs >>>> 318605 -9.4% 288623 slabinfo.Acpi-Namespace.num_objs >>>> 3123 -9.4% 2829 slabinfo.Acpi-Namespace.num_slabs >>>> 220514 -40.7% 130747 slabinfo.btrfs_delayed_node.active_objs >>>> 9751 -25.3% 7283 slabinfo.btrfs_delayed_node.active_slabs >>>> 263293 -25.3% 196669 slabinfo.btrfs_delayed_node.num_objs >>>> 9751 -25.3% 7283 slabinfo.btrfs_delayed_node.num_slabs >>>> 6383 8% -12.0% 5615 2% slabinfo.btrfs_delayed_ref_head.num_objs >>>> 9496 +15.5% 10969 slabinfo.btrfs_extent_buffer.active_objs >>>> 9980 +20.5% 12022 slabinfo.btrfs_extent_buffer.num_objs >>>> 260933 -10.7% 233136 slabinfo.btrfs_extent_map.active_objs >>>> 9392 -10.6% 8396 slabinfo.btrfs_extent_map.active_slabs >>>> 263009 -10.6% 235107 slabinfo.btrfs_extent_map.num_objs >>>> 9392 -10.6% 8396 slabinfo.btrfs_extent_map.num_slabs >>>> 271938 -10.3% 243802 slabinfo.btrfs_inode.active_objs >>>> 9804 -10.6% 8768 slabinfo.btrfs_inode.active_slabs >>>> 273856 -10.4% 245359 slabinfo.btrfs_inode.num_objs >>>> 9804 -10.6% 8768 slabinfo.btrfs_inode.num_slabs >>>> 7085 5% -5.5% 6692 2% slabinfo.btrfs_path.num_objs >>>> 311936 -16.4% 260797 slabinfo.dentry.active_objs >>>> 7803 -9.6% 7058 slabinfo.dentry.active_slabs >>>> 327759 -9.6% 296439 slabinfo.dentry.num_objs >>>> 7803 -9.6% 7058 slabinfo.dentry.num_slabs >>>> 2289 -23.3% 1755 6% slabinfo.proc_inode_cache.active_objs >>>> 2292 -19.0% 1856 6% slabinfo.proc_inode_cache.num_objs >>>> 261546 -12.3% 229485 slabinfo.radix_tree_node.active_objs >>>> 9404 -11.9% 8288 slabinfo.radix_tree_node.active_slabs >>>> 263347 -11.9% 232089 slabinfo.radix_tree_node.num_objs >>>> 9404 -11.9% 8288 slabinfo.radix_tree_node.num_slabs >>> >>> The slab size decreased with the new commit. >>> >>> From perf-profile result, >>> >>> 26.81 ± 2% -6.5 20.35 ± 2% perf-profile.calltrace.cycles-pp.secondary_startup_64 >>> 24.48 ± 2% -5.8 18.73 perf-profile.calltrace.cycles-pp.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64 >>> 24.48 ± 2% -5.8 18.73 perf-profile.calltrace.cycles-pp.cpu_startup_entry.start_secondary.secondary_startup_64 >>> 24.48 ± 2% -5.8 18.73 perf-profile.calltrace.cycles-pp.start_secondary.secondary_startup_64 >>> 22.80 ± 2% -5.5 17.30 ± 2% perf-profile.calltrace.cycles-pp.cpuidle_enter_state.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64 >>> 20.20 ± 2% -4.3 15.85 perf-profile.calltrace.cycles-pp.intel_idle.cpuidle_enter_state.do_idle.cpu_startup_entry.start_secondary >>> 23.03 ± 2% -2.6 20.42 perf-profile.calltrace.cycles-pp.sys_rename.entry_SYSCALL_64_fastpath >>> 17.02 ± 2% -1.8 15.17 ± 2% perf-profile.calltrace.cycles-pp.btrfs_rename.vfs_rename.sys_rename.entry_SYSCALL_64_fastpath >>> 17.03 ± 2% -1.8 15.19 ± 2% perf-profile.calltrace.cycles-pp.vfs_rename.sys_rename.entry_SYSCALL_64_fastpath >>> 13.70 ± 2% -1.2 12.47 ± 3% perf-profile.calltrace.cycles-pp.__btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename.entry_SYSCALL_64_fastpath >>> 5.06 ± 3% -1.1 3.96 ± 2% perf-profile.calltrace.cycles-pp.btrfs_async_run_delayed_root.normal_work_helper.process_one_work.worker_thread.kthread >>> 7.37 ± 4% -0.9 6.49 ± 2% perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.finish_wait.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot >>> 7.32 ± 4% -0.9 6.45 ± 2% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.finish_wait.btrfs_tree_lock.btrfs_lock_root_node >>> 1.54 ± 4% -0.7 0.81 ± 7% perf-profile.calltrace.cycles-pp.poll_idle.cpuidle_enter_state.do_idle.cpu_startup_entry.start_secondary >>> 2.33 ± 2% -0.7 1.62 ± 4% perf-profile.calltrace.cycles-pp.do_idle.cpu_startup_entry.start_kernel.secondary_startup_64 >>> 2.33 ± 2% -0.7 1.62 ± 4% perf-profile.calltrace.cycles-pp.cpu_startup_entry.start_kernel.secondary_startup_64 >>> 2.33 ± 2% -0.7 1.62 ± 4% perf-profile.calltrace.cycles-pp.start_kernel.secondary_startup_64 >>> 2.23 ± 3% -0.7 1.53 ± 4% perf-profile.calltrace.cycles-pp.cpuidle_enter_state.do_idle.cpu_startup_entry.start_kernel.secondary_startup_64 >>> 2.13 ± 2% -0.7 1.46 ± 4% perf-profile.calltrace.cycles-pp.intel_idle.cpuidle_enter_state.do_idle.cpu_startup_entry.start_kernel >>> 5.59 -0.7 4.94 ± 3% perf-profile.calltrace.cycles-pp.__dentry_kill.dput.sys_rename.entry_SYSCALL_64_fastpath >>> 5.60 -0.7 4.94 ± 3% perf-profile.calltrace.cycles-pp.dput.sys_rename.entry_SYSCALL_64_fastpath >>> 6.96 -0.7 6.31 ± 3% perf-profile.calltrace.cycles-pp.btrfs_del_inode_ref.__btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename >>> 5.59 -0.7 4.94 ± 3% perf-profile.calltrace.cycles-pp.evict.__dentry_kill.dput.sys_rename.entry_SYSCALL_64_fastpath >>> 5.58 -0.6 4.94 ± 3% perf-profile.calltrace.cycles-pp.btrfs_evict_inode.evict.__dentry_kill.dput.sys_rename >>> 6.94 ± 2% -0.6 6.30 ± 3% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_del_inode_ref.__btrfs_unlink_inode.btrfs_rename.vfs_rename >>> 6.66 ± 4% -0.6 6.10 ± 3% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_rename.vfs_rename >>> 6.66 ± 4% -0.6 6.10 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename >>> 3.38 ± 3% -0.5 2.84 ± 3% >> perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_delete_delayed_items.btrfs_async_run_delayed_root.normal_work_helper.process_one_work >>> 3.40 ± 3% -0.5 2.86 ± 3% perf-profile.calltrace.cycles-pp.btrfs_delete_delayed_items.btrfs_async_run_delayed_root.normal_work_helper.process_one_work.worker_thread >>> 7.30 ± 4% -0.4 6.86 ± 2% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item.__btrfs_unlink_inode >>> 1.18 ± 4% -0.4 0.76 ± 2% perf-profile.calltrace.cycles-pp.__btrfs_update_delayed_inode.btrfs_async_run_delayed_root.normal_work_helper.process_one_work.worker_thread >>> 1.16 ± 4% -0.4 0.74 ± 2% >> perf-profile.calltrace.cycles-pp.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_async_run_delayed_root.normal_work_helper.process_one_work >>> 5.96 ± 2% -0.4 5.54 ± 3% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_inode_ref.__btrfs_unlink_inode >>> 5.96 ± 2% -0.4 5.54 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_inode_ref.__btrfs_unlink_inode.btrfs_rename >>> 1.16 ± 4% -0.4 0.74 ± 2% >> perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_async_run_delayed_root.normal_work_helper >>> 2.95 ± 3% -0.4 2.54 ± 3% >> perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items.btrfs_async_run_delayed_root >>> 2.95 ± 3% -0.4 2.54 ± 3% >> perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items.btrfs_async_run_delayed_root.normal_work_helper >>> 0.94 ± 2% -0.3 0.59 ± 4% perf-profile.calltrace.cycles-pp.shrink_inactive_list.shrink_node_memcg.shrink_node.kswapd.kthread >>> 1.25 -0.3 0.91 ± 2% perf-profile.calltrace.cycles-pp.shrink_node_memcg.shrink_node.kswapd.kthread.ret_from_fork >>> 0.84 ± 2% -0.3 0.52 ± 3% perf-profile.calltrace.cycles-pp.shrink_page_list.shrink_inactive_list.shrink_node_memcg.shrink_node.kswapd >>> 1.01 -0.3 0.71 ± 3% perf-profile.calltrace.cycles-pp.btrfs_create.path_openat.do_filp_open.do_sys_open.entry_SYSCALL_64_fastpath >>> 0.58 ± 4% -0.2 0.34 ± 70% > perf-profile.calltrace.cycles-pp.queued_write_lock_slowpath.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items >>> 1.73 ± 2% -0.2 1.50 ± 2% perf-profile.calltrace.cycles-pp.finish_wait.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_inode_ref >>> 1.67 ± 4% -0.2 1.44 ± 3% perf-profile.calltrace.cycles-pp.finish_wait.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item >>> 1.89 -0.2 1.67 ± 2% perf-profile.calltrace.cycles-pp.btrfs_commit_inode_delayed_inode.btrfs_evict_inode.evict.__dentry_kill.dput >>> 1.88 -0.2 1.66 ± 2% perf-profile.calltrace.cycles-pp.__btrfs_update_delayed_inode.btrfs_commit_inode_delayed_inode.btrfs_evict_inode.evict.__dentry_kill >>> 1.90 -0.2 1.68 ± 2% perf-profile.calltrace.cycles-pp.btrfs_truncate_inode_items.btrfs_evict_inode.evict.__dentry_kill.dput >>> 1.78 -0.2 1.57 ± 5% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_del_orphan_item.btrfs_orphan_del.btrfs_evict_inode.evict >>> 1.88 -0.2 1.67 ± 2% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_truncate_inode_items.btrfs_evict_inode.evict.__dentry_kill >>> 1.78 -0.2 1.57 ± 5% perf-profile.calltrace.cycles-pp.btrfs_orphan_del.btrfs_evict_inode.evict.__dentry_kill.dput >>> 1.78 -0.2 1.57 ± 5% perf-profile.calltrace.cycles-pp.btrfs_del_orphan_item.btrfs_orphan_del.btrfs_evict_inode.evict.__dentry_kill >>> 1.87 -0.2 1.66 ± 2% >> perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_commit_inode_delayed_inode.btrfs_evict_inode >>> 1.87 -0.2 1.66 ± 2% perf-profile.calltrace.cycles-pp.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_commit_inode_delayed_inode.btrfs_evict_inode.evict >>> 0.75 ± 11% -0.2 0.57 ± 7% >> perf-profile.calltrace.cycles-pp.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent.__btrfs_drop_extents >>> 0.75 ± 11% -0.2 0.57 ± 7% >> perf-profile.calltrace.cycles-pp.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent.__btrfs_drop_extents.insert_reserved_file_extent >>> 1.39 ± 3% -0.2 1.23 ± 4% perf-profile.calltrace.cycles-pp.prepare_to_wait_event.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items >>> 1.84 -0.2 1.68 ± 3% perf-profile.calltrace.cycles-pp.__btrfs_unlink_inode.btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename >>> 1.84 -0.2 1.68 ± 3% perf-profile.calltrace.cycles-pp.btrfs_unlink_inode.btrfs_rename.vfs_rename.sys_rename.entry_SYSCALL_64_fastpath >>> 1.62 -0.2 1.46 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_truncate_inode_items.btrfs_evict_inode.evict >>> 0.87 ± 5% -0.2 0.72 ± 5% perf-profile.calltrace.cycles-pp.finish_wait.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_delete_delayed_items >>> 1.81 -0.2 1.66 ± 3% perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_unlink_inode.btrfs_rename >>> 1.81 -0.2 1.66 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_unlink_inode.btrfs_rename.vfs_rename >>> 1.62 -0.2 1.46 ± 3% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_truncate_inode_items.btrfs_evict_inode >>> 1.69 ± 2% -0.1 1.55 ± 2% perf-profile.calltrace.cycles-pp.end_bio_extent_readpage.normal_work_helper.process_one_work.worker_thread.kthread >>> 1.58 -0.1 1.44 ± 2% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_inode.__btrfs_update_delayed_inode >>> 1.58 -0.1 1.44 ± 2% >> perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_inode.__btrfs_update_delayed_inode.btrfs_commit_inode_delayed_inode >>> 2.89 ± 3% -0.1 2.77 ± 2% >> perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath.queued_write_lock_slowpath.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot >>> 1.51 -0.1 1.39 ± 5% perf-profile.calltrace.cycles-pp.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_orphan_item.btrfs_orphan_del >>> 1.51 -0.1 1.39 ± 5% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_del_orphan_item.btrfs_orphan_del.btrfs_evict_inode >>> 0.94 -0.1 0.82 perf-profile.calltrace.cycles-pp.schedule_idle.do_idle.cpu_startup_entry.start_secondary.secondary_startup_64 >>> 0.93 -0.1 0.81 perf-profile.calltrace.cycles-pp.__schedule.schedule_idle.do_idle.cpu_startup_entry.start_secondary >>> 1.52 -0.1 1.44 ± 3% perf-profile.calltrace.cycles-pp.btrfs_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item.__btrfs_unlink_inode.btrfs_unlink_inode >>> 0.71 ± 3% -0.1 0.66 ± 2% perf-profile.calltrace.cycles-pp.prepare_to_wait_event.btrfs_tree_lock.btrfs_lock_root_node.btrfs_search_slot.btrfs_truncate_inode_items >>> 0.60 ± 2% -0.0 0.56 ± 2% perf-profile.calltrace.cycles-pp.ttwu_do_activate.try_to_wake_up.autoremove_wake_function.__wake_up_common.wake_up_page_bit >>> 1.61 ± 2% +0.1 1.67 ± 2% perf-profile.calltrace.cycles-pp.kswapd.kthread.ret_from_fork >>> 1.61 ± 2% +0.1 1.67 ± 2% perf-profile.calltrace.cycles-pp.shrink_node.kswapd.kthread.ret_from_fork >>> 0.55 +0.1 0.68 ± 4% perf-profile.calltrace.cycles-pp.find_get_entry.pagecache_get_page.generic_file_read_iter.__vfs_read.vfs_read >>> 0.57 +0.1 0.70 ± 4% perf-profile.calltrace.cycles-pp.pagecache_get_page.generic_file_read_iter.__vfs_read.vfs_read.sys_read >>> 0.59 ± 3% +0.3 0.87 perf-profile.calltrace.cycles-pp.__account_scheduler_latency.enqueue_entity.enqueue_task_fair.ttwu_do_activate.try_to_wake_up >>> 1.29 ± 3% +0.3 1.58 perf-profile.calltrace.cycles-pp.enqueue_task_fair.ttwu_do_activate.try_to_wake_up.autoremove_wake_function.__wake_up_common >>> 0.54 +0.3 0.83 ± 2% >> perf-profile.calltrace.cycles-pp.queued_read_lock_slowpath.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent >>> 0.84 +0.3 1.18 ± 3% >> perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.prepare_to_wait_event.btrfs_tree_read_lock.btrfs_read_lock_root_node >>> 0.71 ± 4% +0.3 1.05 perf-profile.calltrace.cycles-pp.ttwu_do_activate.try_to_wake_up.autoremove_wake_function.__wake_up_common.__wake_up_common_lock >>> 0.86 +0.4 1.21 ± 3% >> perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.prepare_to_wait_event.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot >>> 0.90 +0.4 1.28 ± 3% > perf-profile.calltrace.cycles-pp.prepare_to_wait_event.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item >>> 0.35 ± 71% +0.4 0.79 ± 3% >> perf-profile.calltrace.cycles-pp.queued_read_lock_slowpath.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_inode >>> 0.98 ± 4% +0.4 1.41 ± 2% perf-profile.calltrace.cycles-pp.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget >>> 0.98 ± 5% +0.4 1.43 ± 2% perf-profile.calltrace.cycles-pp.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget.btrfs_lookup_dentry >>> 0.94 ± 3% +0.4 1.39 perf-profile.calltrace.cycles-pp.try_to_wake_up.autoremove_wake_function.__wake_up_common.__wake_up_common_lock.btrfs_clear_path_blocking >>> 0.96 ± 3% +0.5 1.42 perf-profile.calltrace.cycles-pp.autoremove_wake_function.__wake_up_common.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot >>> 1.05 ± 2% +0.5 1.52 ± 4% perf-profile.calltrace.cycles-pp.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent >>> 1.06 ± 3% +0.5 1.53 ± 4% perf-profile.calltrace.cycles-pp.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent.__do_readpage >>> 0.97 ± 3% +0.5 1.45 perf-profile.calltrace.cycles-pp.__wake_up_common.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item >>> 0.62 ± 4% +0.5 1.13 >> perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath.queued_read_lock_slowpath.btrfs_clear_lock_blocking_rw.btrfs_clear_path_blocking.btrfs_search_slot >>> 0.65 ± 4% +0.5 1.19 >> perf-profile.calltrace.cycles-pp.queued_read_lock_slowpath.btrfs_clear_lock_blocking_rw.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item >>> 0.17 ±141% +0.6 0.73 perf-profile.calltrace.cycles-pp.save_stack_trace_tsk.__account_scheduler_latency.enqueue_entity.enqueue_task_fair.ttwu_do_activate >>> 0.00 +0.6 0.57 ± 2% perf-profile.calltrace.cycles-pp.task_work_run.exit_to_usermode_loop.syscall_return_slowpath.entry_SYSCALL_64_fastpath >>> 0.00 +0.6 0.59 ± 2% perf-profile.calltrace.cycles-pp.exit_to_usermode_loop.syscall_return_slowpath.entry_SYSCALL_64_fastpath >>> 0.00 +0.6 0.60 ± 2% perf-profile.calltrace.cycles-pp.syscall_return_slowpath.entry_SYSCALL_64_fastpath >>> 0.56 ± 4% +0.6 1.17 >> perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_file_extent >>> 1.17 +0.6 1.78 ± 2% perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.finish_wait.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot >>> 1.16 +0.6 1.77 ± 2% >> perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.finish_wait.btrfs_tree_read_lock.btrfs_read_lock_root_node >>> 1.18 +0.6 1.80 ± 2% perf-profile.calltrace.cycles-pp.finish_wait.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item >>> 0.57 ± 7% +0.6 1.21 ± 2% perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_inode >>> 0.75 ± 4% +0.6 1.38 >> perf-profile.calltrace.cycles-pp.btrfs_clear_lock_blocking_rw.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry >>> 1.95 +0.7 2.61 perf-profile.calltrace.cycles-pp.copy_user_enhanced_fast_string.copyout.copy_page_to_iter.generic_file_read_iter.__vfs_read >>> 1.96 +0.7 2.62 perf-profile.calltrace.cycles-pp.copyout.copy_page_to_iter.generic_file_read_iter.__vfs_read.vfs_read >>> 0.00 +0.7 0.70 perf-profile.calltrace.cycles-pp.__save_stack_trace.save_stack_trace_tsk.__account_scheduler_latency.enqueue_entity.enqueue_task_fair >>> 2.10 +0.7 2.80 perf-profile.calltrace.cycles-pp.copy_page_to_iter.generic_file_read_iter.__vfs_read.vfs_read.sys_read >>> 0.85 ± 5% +0.7 1.58 perf-profile.calltrace.cycles-pp.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent >>> 0.84 ± 6% +0.7 1.57 perf-profile.calltrace.cycles-pp.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget >>> 0.00 +0.8 0.75 ± 4% perf-profile.calltrace.cycles-pp.super_cache_scan.shrink_slab.shrink_node.kswapd.kthread >>> 0.00 +0.8 0.75 ± 5% perf-profile.calltrace.cycles-pp.shrink_slab.shrink_node.kswapd.kthread.ret_from_fork >>> 1.07 ± 5% +0.9 1.98 ± 2% perf-profile.calltrace.cycles-pp.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent.__do_readpage >>> 1.09 ± 7% +0.9 2.01 ± 2% perf-profile.calltrace.cycles-pp.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget.btrfs_lookup_dentry >>> 4.90 ± 2% +1.3 6.19 perf-profile.calltrace.cycles-pp.__do_page_cache_readahead.ondemand_readahead.generic_file_read_iter.__vfs_read.vfs_read >>> 4.90 ± 2% +1.3 6.20 perf-profile.calltrace.cycles-pp.ondemand_readahead.generic_file_read_iter.__vfs_read.vfs_read.sys_read >>> 4.44 ± 2% +1.3 5.74 perf-profile.calltrace.cycles-pp.extent_readpages.__do_page_cache_readahead.ondemand_readahead.generic_file_read_iter.__vfs_read >>> 2.87 ± 3% +1.4 4.29 perf-profile.calltrace.cycles-pp.__extent_readpages.extent_readpages.__do_page_cache_readahead.ondemand_readahead.generic_file_read_iter >>> 2.22 ± 4% +1.4 3.65 perf-profile.calltrace.cycles-pp.btrfs_lookup_file_extent.btrfs_get_extent.__do_readpage.__extent_readpages.extent_readpages >>> 2.22 ± 4% +1.4 3.65 perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_file_extent.btrfs_get_extent.__do_readpage.__extent_readpages >>> 2.27 ± 6% +1.4 3.72 perf-profile.calltrace.cycles-pp.btrfs_lookup_inode.btrfs_iget.btrfs_lookup_dentry.btrfs_lookup.path_openat >>> 2.27 ± 6% +1.4 3.72 perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_inode.btrfs_iget.btrfs_lookup_dentry.btrfs_lookup >>> 2.72 ± 3% +1.4 4.17 perf-profile.calltrace.cycles-pp.__do_readpage.__extent_readpages.extent_readpages.__do_page_cache_readahead.ondemand_readahead >>> 2.33 ± 4% +1.5 3.79 perf-profile.calltrace.cycles-pp.btrfs_get_extent.__do_readpage.__extent_readpages.extent_readpages.__do_page_cache_readahead >>> 2.38 +1.6 3.95 >> perf-profile.calltrace.cycles-pp.queued_read_lock_slowpath.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item >>> 2.63 ± 6% +1.7 4.33 perf-profile.calltrace.cycles-pp.btrfs_iget.btrfs_lookup_dentry.btrfs_lookup.path_openat.do_filp_open >>> 1.68 ± 4% +2.1 3.79 >> perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item >>> 8.66 +2.2 10.90 perf-profile.calltrace.cycles-pp.generic_file_read_iter.__vfs_read.vfs_read.sys_read.entry_SYSCALL_64_fastpath >>> 8.71 +2.3 10.97 perf-profile.calltrace.cycles-pp.__vfs_read.vfs_read.sys_read.entry_SYSCALL_64_fastpath >>> 8.85 +2.3 11.15 perf-profile.calltrace.cycles-pp.vfs_read.sys_read.entry_SYSCALL_64_fastpath >>> 9.02 +2.3 11.33 perf-profile.calltrace.cycles-pp.sys_read.entry_SYSCALL_64_fastpath >>> 2.71 ± 3% +2.6 5.35 perf-profile.calltrace.cycles-pp.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry >>> 4.63 +2.7 7.29 perf-profile.calltrace.cycles-pp.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry >>> 4.65 +2.7 7.32 perf-profile.calltrace.cycles-pp.btrfs_read_lock_root_node.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry.btrfs_lookup >>> 1.94 ± 2% +2.8 4.77 >> perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath.queued_read_lock_slowpath.btrfs_tree_read_lock.btrfs_read_lock_root_node.btrfs_search_slot >>> 2.81 ± 4% +3.3 6.15 >> perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.__wake_up_common_lock.btrfs_clear_path_blocking.btrfs_search_slot >>> 3.58 ± 2% +3.4 6.98 perf-profile.calltrace.cycles-pp.btrfs_clear_path_blocking.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry.btrfs_lookup >>> 8.69 +6.3 15.03 perf-profile.calltrace.cycles-pp.btrfs_search_slot.btrfs_lookup_dir_item.btrfs_lookup_dentry.btrfs_lookup.path_openat >>> 8.75 +6.4 15.12 perf-profile.calltrace.cycles-pp.btrfs_lookup_dir_item.btrfs_lookup_dentry.btrfs_lookup.path_openat.do_filp_open >>> 47.34 +8.1 55.46 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_fastpath >>> 11.47 +8.1 19.60 perf-profile.calltrace.cycles-pp.btrfs_lookup.path_openat.do_filp_open.do_sys_open.entry_SYSCALL_64_fastpath >>> 11.46 +8.1 19.59 perf-profile.calltrace.cycles-pp.btrfs_lookup_dentry.btrfs_lookup.path_openat.do_filp_open.do_sys_open >>> 13.90 +8.3 22.20 perf-profile.calltrace.cycles-pp.path_openat.do_filp_open.do_sys_open.entry_SYSCALL_64_fastpath >>> 13.91 +8.3 22.22 perf-profile.calltrace.cycles-pp.do_filp_open.do_sys_open.entry_SYSCALL_64_fastpath >>> 14.13 +8.4 22.52 perf-profile.calltrace.cycles-pp.do_sys_open.entry_SYSCALL_64_fastpath >>> >>> The cycles for dentry lookup increased much too. Is this the reason why >>> write score decreased? >>> >>> If you need more information, please let me know. >>> >>> Best Regards, >>> Huang, Ying >>> >>>> 1140424 12% +40.2% 1598980 14% sched_debug.cfs_rq:/.MIN_vruntime.max >>>> 790.55 +13.0% 893.20 3% sched_debug.cfs_rq:/.exec_clock.stddev >>>> 1140425 12% +40.2% 1598982 14% sched_debug.cfs_rq:/.max_vruntime.max >>>> 0.83 10% +21.5% 1.00 8% sched_debug.cfs_rq:/.nr_running.avg >>>> 3.30 99% +266.3% 12.09 13% sched_debug.cfs_rq:/.removed.load_avg.avg >>>> 153.02 97% +266.6% 560.96 13% sched_debug.cfs_rq:/.removed.runnable_sum.avg >>>> 569.93 102% +173.2% 1556 14% sched_debug.cfs_rq:/.removed.runnable_sum.stddev >>>> 1.42 60% +501.5% 8.52 34% sched_debug.cfs_rq:/.removed.util_avg.avg >>>> 19.88 59% +288.9% 77.29 16% sched_debug.cfs_rq:/.removed.util_avg.max >>>> 5.05 58% +342.3% 22.32 22% sched_debug.cfs_rq:/.removed.util_avg.stddev >>>> 791.44 3% +47.7% 1168 8% sched_debug.cfs_rq:/.util_avg.avg >>>> 1305 6% +33.2% 1738 5% sched_debug.cfs_rq:/.util_avg.max >>>> 450.25 11% +66.2% 748.17 14% sched_debug.cfs_rq:/.util_avg.min >>>> 220.82 8% +21.1% 267.46 5% sched_debug.cfs_rq:/.util_avg.stddev >>>> 363118 11% -23.8% 276520 11% sched_debug.cpu.avg_idle.avg >>>> 726003 8% -30.8% 502313 4% sched_debug.cpu.avg_idle.max >>>> 202629 3% -32.2% 137429 18% sched_debug.cpu.avg_idle.stddev >>>> 31.96 28% +54.6% 49.42 14% sched_debug.cpu.cpu_load[3].min >>>> 36.21 25% +64.0% 59.38 6% sched_debug.cpu.cpu_load[4].min >>>> 1007 5% +20.7% 1216 7% sched_debug.cpu.curr->pid.avg >>>> 4.50 5% +14.8% 5.17 5% sched_debug.cpu.nr_running.max >>>> 2476195 -11.8% 2185022 sched_debug.cpu.nr_switches.max >>>> 212888 -26.6% 156172 3% sched_debug.cpu.nr_switches.stddev >>>> 3570 2% -58.7% 1474 2% sched_debug.cpu.nr_uninterruptible.max >>>> -803.67 -28.7% -573.38 sched_debug.cpu.nr_uninterruptible.min >>>> 1004 2% -50.4% 498.55 3% sched_debug.cpu.nr_uninterruptible.stddev >>>> 2478809 -11.7% 2189310 sched_debug.cpu.sched_count.max >>>> 214130 -26.5% 157298 3% sched_debug.cpu.sched_count.stddev >>>> 489430 2% -16.6% 408309 2% sched_debug.cpu.sched_goidle.avg >>>> 724333 2% -28.2% 520263 2% sched_debug.cpu.sched_goidle.max >>>> 457611 -18.1% 374746 3% sched_debug.cpu.sched_goidle.min >>>> 62957 2% -47.4% 33138 3% sched_debug.cpu.sched_goidle.stddev >>>> 676053 2% -15.4% 571816 2% sched_debug.cpu.ttwu_local.max >>>> 42669 3% +22.3% 52198 sched_debug.cpu.ttwu_local.min >>>> 151873 2% -18.3% 124118 2% sched_debug.cpu.ttwu_local.stddev >>>> >>>> >>>> >>>> blogbench.write_score >>>> >>>> 3300 +-+------------------------------------------------------------------+ >>>> 3250 +-+ +. .+ +. .+ : : : +. .+ .+.+.+. .| >>>> |: +. .+ +.+.+.+ + + + : +. : : +. + +.+ + + | >>>> 3200 +-+ + +.+ + : + + : + + | >>>> 3150 +-+.+ ++ +.+ | >>>> 3100 +-+ | >>>> 3050 +-+ | >>>> | | >>>> 3000 +-+ | >>>> 2950 +-+ O O | >>>> 2900 +-O O O O | >>>> 2850 +-+ O O O O O O O OO O O O | >>>> | O O O O | >>>> 2800 O-+ O O | >>>> 2750 +-+------------------------------------------------------------------+ >>>> >>>> >>>> [*] 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. >>>> >>>> >>>> Thanks, >>>> Xiaolong >>> _______________________________________________ >>> LKP mailing list >>> LKP@lists.01.org >>> https://lists.01.org/mailman/listinfo/lkp > _______________________________________________ > LKP mailing list > LKP@lists.01.org > https://lists.01.org/mailman/listinfo/lkp ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [LKP] [lkp-robot] [mm] 9092c71bb7: blogbench.write_score -12.3% regression 2018-06-20 3:51 ` Huang, Ying @ 2018-06-20 12:38 ` Chris Mason 2018-06-21 0:38 ` Huang, Ying 2018-07-13 1:55 ` Huang, Ying 0 siblings, 2 replies; 11+ messages in thread From: Chris Mason @ 2018-06-20 12:38 UTC (permalink / raw) To: Huang, Ying; +Cc: Josef Bacik, David Sterba, lkp, linux-btrfs On 19 Jun 2018, at 23:51, Huang, Ying wrote: >>> "Huang, Ying" <ying.huang@intel.com> writes: >>> >>>> Hi, Josef, >>>> >>>> Do you have time to take a look at the regression? >>>> >>>> kernel test robot <xiaolong.ye@intel.com> writes: >>>> >>>>> Greeting, >>>>> >>>>> FYI, we noticed a -12.3% regression of blogbench.write_score and a >>>>> +9.6% improvement >>>>> of blogbench.read_score due to commit: >>>>> >>>>> >>>>> commit: 9092c71bb724dba2ecba849eae69e5c9d39bd3d2 ("mm: use >>>>> sc->priority for slab shrink targets") >>>>> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git >>>>> master >>>>> >>>>> in testcase: blogbench >>>>> on test machine: 16 threads Intel(R) Xeon(R) CPU D-1541 @ 2.10GHz >>>>> with 8G memory >>>>> with following parameters: >>>>> >>>>> disk: 1SSD >>>>> fs: btrfs >>>>> cpufreq_governor: performance >>>>> >>>>> test-description: Blogbench is a portable filesystem benchmark >>>>> that tries to reproduce the load of a real-world busy file server. >>>>> test-url: I'm surprised, this patch is a big win in production here at FB. I'll have to reproduce these results to better understand what is going on. My first guess is that since we have fewer inodes in slab, we're reading more inodes from disk in order to do the writes. But that should also make our read scores lower. -chris ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [LKP] [lkp-robot] [mm] 9092c71bb7: blogbench.write_score -12.3% regression 2018-06-20 12:38 ` Chris Mason @ 2018-06-21 0:38 ` Huang, Ying 2018-07-13 1:55 ` Huang, Ying 1 sibling, 0 replies; 11+ messages in thread From: Huang, Ying @ 2018-06-21 0:38 UTC (permalink / raw) To: Chris Mason; +Cc: Josef Bacik, David Sterba, lkp, linux-btrfs Chris Mason <clm@fb.com> writes: > On 19 Jun 2018, at 23:51, Huang, Ying wrote: >>>> "Huang, Ying" <ying.huang@intel.com> writes: >>>> >>>>> Hi, Josef, >>>>> >>>>> Do you have time to take a look at the regression? >>>>> >>>>> kernel test robot <xiaolong.ye@intel.com> writes: >>>>> >>>>>> Greeting, >>>>>> >>>>>> FYI, we noticed a -12.3% regression of blogbench.write_score and >>>>>> a +9.6% improvement >>>>>> of blogbench.read_score due to commit: >>>>>> >>>>>> >>>>>> commit: 9092c71bb724dba2ecba849eae69e5c9d39bd3d2 ("mm: use >>>>>> sc->priority for slab shrink targets") >>>>>> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git >>>>>> master >>>>>> >>>>>> in testcase: blogbench >>>>>> on test machine: 16 threads Intel(R) Xeon(R) CPU D-1541 @ >>>>>> 2.10GHz with 8G memory >>>>>> with following parameters: >>>>>> >>>>>> disk: 1SSD >>>>>> fs: btrfs >>>>>> cpufreq_governor: performance >>>>>> >>>>>> test-description: Blogbench is a portable filesystem benchmark >>>>>> that tries to reproduce the load of a real-world busy file >>>>>> server. >>>>>> test-url: > > I'm surprised, this patch is a big win in production here at FB. I'll > have to reproduce these results to better understand what is going on. > My first guess is that since we have fewer inodes in slab, we're > reading more inodes from disk in order to do the writes. > > But that should also make our read scores lower. Thanks for looking at this. If you need more information, please let me know. Best Regards, Huang, Ying ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [LKP] [lkp-robot] [mm] 9092c71bb7: blogbench.write_score -12.3% regression 2018-06-20 12:38 ` Chris Mason 2018-06-21 0:38 ` Huang, Ying @ 2018-07-13 1:55 ` Huang, Ying 2018-08-02 5:55 ` Huang, Ying 1 sibling, 1 reply; 11+ messages in thread From: Huang, Ying @ 2018-07-13 1:55 UTC (permalink / raw) To: Chris Mason; +Cc: Josef Bacik, David Sterba, lkp, linux-btrfs Hi, Chris, Chris Mason <clm@fb.com> writes: > On 19 Jun 2018, at 23:51, Huang, Ying wrote: >>>> "Huang, Ying" <ying.huang@intel.com> writes: >>>> >>>>> Hi, Josef, >>>>> >>>>> Do you have time to take a look at the regression? >>>>> >>>>> kernel test robot <xiaolong.ye@intel.com> writes: >>>>> >>>>>> Greeting, >>>>>> >>>>>> FYI, we noticed a -12.3% regression of blogbench.write_score and >>>>>> a +9.6% improvement >>>>>> of blogbench.read_score due to commit: >>>>>> >>>>>> >>>>>> commit: 9092c71bb724dba2ecba849eae69e5c9d39bd3d2 ("mm: use >>>>>> sc->priority for slab shrink targets") >>>>>> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git >>>>>> master >>>>>> >>>>>> in testcase: blogbench >>>>>> on test machine: 16 threads Intel(R) Xeon(R) CPU D-1541 @ >>>>>> 2.10GHz with 8G memory >>>>>> with following parameters: >>>>>> >>>>>> disk: 1SSD >>>>>> fs: btrfs >>>>>> cpufreq_governor: performance >>>>>> >>>>>> test-description: Blogbench is a portable filesystem benchmark >>>>>> that tries to reproduce the load of a real-world busy file >>>>>> server. >>>>>> test-url: > > I'm surprised, this patch is a big win in production here at FB. I'll > have to reproduce these results to better understand what is going on. > My first guess is that since we have fewer inodes in slab, we're > reading more inodes from disk in order to do the writes. > > But that should also make our read scores lower. Any update on this? Best Regards, Huang, Ying ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [LKP] [lkp-robot] [mm] 9092c71bb7: blogbench.write_score -12.3% regression 2018-07-13 1:55 ` Huang, Ying @ 2018-08-02 5:55 ` Huang, Ying 2018-08-02 16:23 ` Josef Bacik 0 siblings, 1 reply; 11+ messages in thread From: Huang, Ying @ 2018-08-02 5:55 UTC (permalink / raw) To: Chris Mason; +Cc: Josef Bacik, David Sterba, lkp, linux-btrfs "Huang, Ying" <ying.huang@intel.com> writes: > Hi, Chris, > > Chris Mason <clm@fb.com> writes: > >> On 19 Jun 2018, at 23:51, Huang, Ying wrote: >>>>> "Huang, Ying" <ying.huang@intel.com> writes: >>>>> >>>>>> Hi, Josef, >>>>>> >>>>>> Do you have time to take a look at the regression? >>>>>> >>>>>> kernel test robot <xiaolong.ye@intel.com> writes: >>>>>> >>>>>>> Greeting, >>>>>>> >>>>>>> FYI, we noticed a -12.3% regression of blogbench.write_score and >>>>>>> a +9.6% improvement >>>>>>> of blogbench.read_score due to commit: >>>>>>> >>>>>>> >>>>>>> commit: 9092c71bb724dba2ecba849eae69e5c9d39bd3d2 ("mm: use >>>>>>> sc->priority for slab shrink targets") >>>>>>> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git >>>>>>> master >>>>>>> >>>>>>> in testcase: blogbench >>>>>>> on test machine: 16 threads Intel(R) Xeon(R) CPU D-1541 @ >>>>>>> 2.10GHz with 8G memory >>>>>>> with following parameters: >>>>>>> >>>>>>> disk: 1SSD >>>>>>> fs: btrfs >>>>>>> cpufreq_governor: performance >>>>>>> >>>>>>> test-description: Blogbench is a portable filesystem benchmark >>>>>>> that tries to reproduce the load of a real-world busy file >>>>>>> server. >>>>>>> test-url: >> >> I'm surprised, this patch is a big win in production here at FB. I'll >> have to reproduce these results to better understand what is going on. >> My first guess is that since we have fewer inodes in slab, we're >> reading more inodes from disk in order to do the writes. >> >> But that should also make our read scores lower. > > Any update on this? Ping. Best Regards, Huang, Ying ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [LKP] [lkp-robot] [mm] 9092c71bb7: blogbench.write_score -12.3% regression 2018-08-02 5:55 ` Huang, Ying @ 2018-08-02 16:23 ` Josef Bacik 2018-08-03 8:22 ` Huang, Ying 0 siblings, 1 reply; 11+ messages in thread From: Josef Bacik @ 2018-08-02 16:23 UTC (permalink / raw) To: Huang, Ying; +Cc: Chris Mason, Josef Bacik, David Sterba, lkp, linux-btrfs On Thu, Aug 02, 2018 at 01:55:23PM +0800, Huang, Ying wrote: > "Huang, Ying" <ying.huang@intel.com> writes: > > > Hi, Chris, > > > > Chris Mason <clm@fb.com> writes: > > > >> On 19 Jun 2018, at 23:51, Huang, Ying wrote: > >>>>> "Huang, Ying" <ying.huang@intel.com> writes: > >>>>> > >>>>>> Hi, Josef, > >>>>>> > >>>>>> Do you have time to take a look at the regression? > >>>>>> > >>>>>> kernel test robot <xiaolong.ye@intel.com> writes: > >>>>>> > >>>>>>> Greeting, > >>>>>>> > >>>>>>> FYI, we noticed a -12.3% regression of blogbench.write_score and > >>>>>>> a +9.6% improvement > >>>>>>> of blogbench.read_score due to commit: > >>>>>>> > >>>>>>> > >>>>>>> commit: 9092c71bb724dba2ecba849eae69e5c9d39bd3d2 ("mm: use > >>>>>>> sc->priority for slab shrink targets") > >>>>>>> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git > >>>>>>> master > >>>>>>> > >>>>>>> in testcase: blogbench > >>>>>>> on test machine: 16 threads Intel(R) Xeon(R) CPU D-1541 @ > >>>>>>> 2.10GHz with 8G memory > >>>>>>> with following parameters: > >>>>>>> > >>>>>>> disk: 1SSD > >>>>>>> fs: btrfs > >>>>>>> cpufreq_governor: performance > >>>>>>> > >>>>>>> test-description: Blogbench is a portable filesystem benchmark > >>>>>>> that tries to reproduce the load of a real-world busy file > >>>>>>> server. > >>>>>>> test-url: > >> > >> I'm surprised, this patch is a big win in production here at FB. I'll > >> have to reproduce these results to better understand what is going on. > >> My first guess is that since we have fewer inodes in slab, we're > >> reading more inodes from disk in order to do the writes. > >> > >> But that should also make our read scores lower. > > > > Any update on this? > > Ping. > I can't reproduce this, and what's more it appears that blogbench doesn't use much memory at all. I have the slab shrinking tracepoints on and we never go into this code at all, so I'm pretty sure these results are bogus. How are you running blogbench? I'm doing blogbench -d /whatever, if I need to be doing something else let me know. But from what I can tell this thing uses less than 100m of memory, and on an 8gig of ram box we're never going to trip over this code. Thanks, Josef ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [LKP] [lkp-robot] [mm] 9092c71bb7: blogbench.write_score -12.3% regression 2018-08-02 16:23 ` Josef Bacik @ 2018-08-03 8:22 ` Huang, Ying 2018-08-29 6:55 ` Huang, Ying 0 siblings, 1 reply; 11+ messages in thread From: Huang, Ying @ 2018-08-03 8:22 UTC (permalink / raw) To: Josef Bacik; +Cc: Chris Mason, Josef Bacik, David Sterba, lkp, linux-btrfs [-- Attachment #1: Type: text/plain, Size: 2502 bytes --] Josef Bacik <josef@toxicpanda.com> writes: > On Thu, Aug 02, 2018 at 01:55:23PM +0800, Huang, Ying wrote: >> "Huang, Ying" <ying.huang@intel.com> writes: >> >> > Hi, Chris, >> > >> > Chris Mason <clm@fb.com> writes: >> > >> >> On 19 Jun 2018, at 23:51, Huang, Ying wrote: >> >>>>> "Huang, Ying" <ying.huang@intel.com> writes: >> >>>>> >> >>>>>> Hi, Josef, >> >>>>>> >> >>>>>> Do you have time to take a look at the regression? >> >>>>>> >> >>>>>> kernel test robot <xiaolong.ye@intel.com> writes: >> >>>>>> >> >>>>>>> Greeting, >> >>>>>>> >> >>>>>>> FYI, we noticed a -12.3% regression of blogbench.write_score and >> >>>>>>> a +9.6% improvement >> >>>>>>> of blogbench.read_score due to commit: >> >>>>>>> >> >>>>>>> >> >>>>>>> commit: 9092c71bb724dba2ecba849eae69e5c9d39bd3d2 ("mm: use >> >>>>>>> sc->priority for slab shrink targets") >> >>>>>>> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git >> >>>>>>> master >> >>>>>>> >> >>>>>>> in testcase: blogbench >> >>>>>>> on test machine: 16 threads Intel(R) Xeon(R) CPU D-1541 @ >> >>>>>>> 2.10GHz with 8G memory >> >>>>>>> with following parameters: >> >>>>>>> >> >>>>>>> disk: 1SSD >> >>>>>>> fs: btrfs >> >>>>>>> cpufreq_governor: performance >> >>>>>>> >> >>>>>>> test-description: Blogbench is a portable filesystem benchmark >> >>>>>>> that tries to reproduce the load of a real-world busy file >> >>>>>>> server. >> >>>>>>> test-url: >> >> >> >> I'm surprised, this patch is a big win in production here at FB. I'll >> >> have to reproduce these results to better understand what is going on. >> >> My first guess is that since we have fewer inodes in slab, we're >> >> reading more inodes from disk in order to do the writes. >> >> >> >> But that should also make our read scores lower. >> > >> > Any update on this? >> >> Ping. >> > > I can't reproduce this, and what's more it appears that blogbench doesn't use > much memory at all. I have the slab shrinking tracepoints on and we never go > into this code at all, so I'm pretty sure these results are bogus. How are you > running blogbench? I'm doing blogbench -d /whatever, if I need to be doing > something else let me know. But from what I can tell this thing uses less than > 100m of memory, and on an 8gig of ram box we're never going to trip over this > code. Thanks, Thanks for looking at this. In my testing, blogbench will eat up system memory. Please check the vmstat result attached. The SSD disk size is about 745GB. Best Regards, Huang, Ying [-- Attachment #2: vmstat --] [-- Type: application/octet-stream, Size: 32309 bytes --] procs -----------memory---------- ---swap-- -----io---- -system-- ------cpu----- -----timestamp----- r b swpd free buff cache si so bi bo in cs us sy id wa st CST 14 0 0 6642188 1060 1248084 0 0 3 4 73 45 0 1 95 4 0 2018-05-25 10:51:34 3 0 0 6620208 1060 1250940 0 0 0 0 19914 6209 1 5 94 0 0 2018-05-25 10:51:35 7 0 0 6497528 1060 1360696 0 0 0 8148 19013 384269 0 29 71 0 0 2018-05-25 10:51:36 7 0 0 6415000 1060 1440780 0 0 0 8156 26960 278704 3 60 36 0 0 2018-05-25 10:51:37 8 0 0 6341700 1060 1512136 0 0 0 11604 37912 201107 6 81 13 0 0 2018-05-25 10:51:38 24 0 0 6263708 1060 1582232 0 0 0 15412 44290 172224 7 86 7 0 0 2018-05-25 10:51:39 32 0 0 6187240 1060 1656268 0 0 0 14324 47792 120067 8 91 2 0 0 2018-05-25 10:51:40 46 0 0 6127800 1060 1716044 0 0 0 12248 49691 105476 7 92 1 0 0 2018-05-25 10:51:41 48 0 0 6058420 1060 1784328 0 0 0 12276 44119 87456 8 92 0 0 0 2018-05-25 10:51:42 52 0 0 5986160 1060 1853400 0 0 0 14268 42946 85125 8 92 0 0 0 2018-05-25 10:51:43 55 0 0 5899824 1060 1938660 0 0 0 16380 39720 72212 8 92 0 0 0 2018-05-25 10:51:44 69 0 0 5815876 1060 2019760 0 0 0 13180 42299 78941 8 92 0 0 0 2018-05-25 10:51:45 71 0 0 5714780 1060 2119380 0 0 0 17436 35191 109966 8 92 0 0 0 2018-05-25 10:51:46 86 0 0 5666764 1060 2165416 0 0 0 76944 68235 134569 7 93 0 0 0 2018-05-25 10:51:47 76 0 0 5599836 1060 2230540 0 0 0 94060 49109 89942 7 93 0 0 0 2018-05-25 10:51:48 74 0 0 5530456 1060 2295772 0 0 0 13180 51140 102879 7 93 0 0 0 2018-05-25 10:51:49 79 0 0 5459936 1060 2365036 0 0 0 120556 67851 131159 7 93 0 0 0 2018-05-25 10:51:50 77 0 0 5363912 1060 2454576 0 0 0 97520 41808 75273 8 92 0 0 0 2018-05-25 10:51:51 84 0 0 5355112 1060 2462996 0 0 0 378852 120202 238064 6 94 0 0 0 2018-05-25 10:51:52 87 0 0 5348348 1060 2472388 0 0 0 1540 108886 213065 6 94 0 0 0 2018-05-25 10:51:53 84 0 0 5292048 1060 2526180 0 0 0 10792 66284 122672 7 93 0 0 0 2018-05-25 10:51:54 83 0 0 5212544 1060 2604840 0 0 0 13556 30783 49602 8 92 0 0 0 2018-05-25 10:51:55 87 0 0 5123440 1060 2692760 0 0 0 13528 33120 55286 8 92 0 0 0 2018-05-25 10:51:56 76 0 0 5038640 1060 2776444 0 0 0 15644 35084 58778 8 92 0 0 0 2018-05-25 10:51:57 79 0 0 4951064 1060 2860372 0 0 0 13388 36160 122190 8 92 0 0 0 2018-05-25 10:51:58 87 0 0 4876520 1060 2935112 0 0 0 13232 38157 66745 8 92 0 0 0 2018-05-25 10:51:59 97 0 0 4812444 1060 2997960 0 0 0 132996 81242 146545 7 93 0 0 0 2018-05-25 10:52:00 81 0 0 4739468 1060 3071280 0 0 0 13612 39189 69867 8 92 0 0 0 2018-05-25 10:52:01 95 0 0 4680284 1060 3125984 0 0 0 98752 68985 131210 7 93 0 0 0 2018-05-25 10:52:02 98 0 0 4607340 1060 3199116 0 0 0 13912 33563 52438 8 92 0 0 0 2018-05-25 10:52:03 101 0 0 4512852 1060 3228036 0 0 0 528404 93205 192932 6 94 0 0 0 2018-05-25 10:52:04 109 0 0 4560816 1060 3238416 0 0 0 736 92678 204068 6 94 0 0 0 2018-05-25 10:52:05 91 0 0 4513972 1060 3285744 0 0 0 10408 70951 146450 7 93 0 0 0 2018-05-25 10:52:06 85 0 0 4439888 1060 3359920 0 0 0 13944 37519 67192 8 92 0 0 0 2018-05-25 10:52:07 89 0 0 4349672 1060 3447380 0 0 0 16992 30923 46997 8 92 0 0 0 2018-05-25 10:52:08 92 0 0 4257240 1060 3537772 0 0 0 19664 37809 64057 8 92 0 0 0 2018-05-25 10:52:09 96 0 0 4158816 1060 3635796 0 0 0 76188 32904 114926 8 92 0 0 0 2018-05-25 10:52:10 96 0 0 4076380 1060 3717872 0 0 0 15652 35763 61020 8 92 0 0 0 2018-05-25 10:52:11 79 0 0 3976928 1060 3814760 0 0 0 18572 35678 61598 8 92 0 0 0 2018-05-25 10:52:12 102 0 0 3897084 1060 3894444 0 0 0 146180 64104 113584 7 93 0 0 0 2018-05-25 10:52:13 97 0 0 3813860 1060 3974568 0 0 0 11916 49469 84198 8 92 0 0 0 2018-05-25 10:52:14 107 1 0 3699484 1060 4030616 0 0 0 245356 82101 152495 7 93 0 0 0 2018-05-25 10:52:15 106 0 0 3736120 1060 4046108 0 0 0 362108 89591 210520 6 94 0 0 0 2018-05-25 10:52:16 98 0 0 3690320 1060 4092808 0 0 0 7120 54281 137038 7 93 0 0 0 2018-05-25 10:52:17 84 0 0 3606060 1060 4176408 0 0 0 17768 35664 63500 7 92 0 0 0 2018-05-25 10:52:18 89 0 0 3520716 1060 4257204 0 0 0 17776 34047 55694 8 92 0 0 0 2018-05-25 10:52:19 92 0 0 3450244 1060 4327504 0 0 0 14548 35126 59935 8 92 0 0 0 2018-05-25 10:52:20 94 0 0 3372952 1060 4406208 0 0 0 13840 34021 60502 8 92 0 0 0 2018-05-25 10:52:21 103 0 0 3318184 1060 4458128 0 0 0 11604 35715 120931 8 92 0 0 0 2018-05-25 10:52:22 91 0 0 3227348 1060 4547492 0 0 0 18524 34857 66821 8 92 0 0 0 2018-05-25 10:52:23 93 0 0 3144992 1060 4629212 0 0 0 22160 33336 54199 8 92 0 0 0 2018-05-25 10:52:24 94 0 0 3069600 1060 4700180 0 0 0 157816 52298 97336 7 93 0 0 0 2018-05-25 10:52:25 93 0 0 2997840 1060 4771928 0 0 0 11808 59970 107866 7 93 0 0 0 2018-05-25 10:52:26 104 0 0 2901676 1060 4845120 0 0 0 157320 50768 95525 8 92 0 0 0 2018-05-25 10:52:27 108 0 0 2894308 1060 4855224 0 0 0 425912 114900 241862 6 94 0 0 0 2018-05-25 10:52:28 99 0 0 2895808 1060 4867292 0 0 0 1452 91706 205371 6 94 0 0 0 2018-05-25 10:52:29 103 0 0 2836108 1060 4928676 0 0 0 12428 52910 116465 8 92 0 0 0 2018-05-25 10:52:30 100 0 0 2740412 1060 5021104 0 0 0 13784 31440 49351 8 92 0 0 0 2018-05-25 10:52:31 87 0 0 2666732 1060 5094060 0 0 0 17024 34167 57728 8 92 0 0 0 2018-05-25 10:52:32 100 0 0 2585492 1060 5174044 0 0 0 18140 36647 60620 8 92 0 0 0 2018-05-25 10:52:33 105 0 0 2520824 1060 5237736 0 0 0 14128 34403 108787 8 92 0 0 0 2018-05-25 10:52:34 103 0 0 2429924 1060 5326468 0 0 0 14644 28682 57748 8 92 0 0 0 2018-05-25 10:52:35 101 0 0 2357940 1060 5397672 0 0 0 15880 34275 55063 8 92 0 0 0 2018-05-25 10:52:36 107 0 0 2286048 1060 5462464 0 0 0 103896 43750 79358 8 92 0 0 0 2018-05-25 10:52:37 98 0 0 2235468 1060 5516996 0 0 0 12056 59000 110222 8 92 0 0 0 2018-05-25 10:52:38 108 0 0 2158392 1060 5560656 0 0 0 289600 57669 126563 7 93 0 0 0 2018-05-25 10:52:39 112 1 0 2172848 1060 5572548 0 0 0 365588 80805 212245 7 93 0 0 0 2018-05-25 10:52:40 92 0 0 2110028 1060 5635868 0 0 0 11564 61873 128218 7 93 0 0 0 2018-05-25 10:52:41 94 0 0 1981496 1060 5761380 0 0 0 21112 30777 46142 8 92 0 0 0 2018-05-25 10:52:42 111 0 0 1850540 1060 5887612 0 0 0 21044 28639 41959 8 92 0 0 0 2018-05-25 10:52:43 101 0 0 1750420 1060 5986276 0 0 0 19548 32024 49157 8 92 0 0 0 2018-05-25 10:52:44 103 0 0 1646920 1060 6088200 0 0 0 22748 37583 61471 8 92 0 0 0 2018-05-25 10:52:45 98 0 0 1546664 1060 6183960 0 0 0 21372 34586 108530 8 92 0 0 0 2018-05-25 10:52:46 114 0 0 1471108 1060 6258208 0 0 0 77808 40236 77738 8 93 0 0 0 2018-05-25 10:52:47 111 0 0 1401164 1060 6327888 0 0 0 17436 47599 81622 8 92 0 0 0 2018-05-25 10:52:48 95 0 0 1327344 1060 6401852 0 0 0 142768 50838 91746 7 93 0 0 0 2018-05-25 10:52:49 106 1 0 1243300 1060 6476620 0 0 0 80512 48373 84152 8 92 0 0 0 2018-05-25 10:52:50 100 0 0 1223256 1060 6496788 0 0 0 435432 87737 203993 6 94 0 0 0 2018-05-25 10:52:51 98 0 0 1187592 1060 6536360 0 0 0 7124 55860 145391 7 93 0 0 0 2018-05-25 10:52:52 98 0 0 1122896 1060 6600272 0 0 0 14640 37966 65520 8 92 0 0 0 2018-05-25 10:52:53 89 0 0 1043428 1060 6677904 0 0 0 15224 37454 64691 8 92 0 0 0 2018-05-25 10:52:54 104 0 0 965376 1060 6754212 0 0 0 13160 32815 52791 8 92 0 0 0 2018-05-25 10:52:55 106 0 0 863536 1060 6854472 0 0 0 18260 33885 54347 8 92 0 0 0 2018-05-25 10:52:56 104 0 0 768128 1060 6947632 0 0 0 18604 30365 48403 8 92 0 0 0 2018-05-25 10:52:57 101 0 0 699724 1060 7016864 0 0 0 13996 37649 108244 8 92 0 0 0 2018-05-25 10:52:58 109 0 0 627508 1060 7077380 0 0 0 128264 53258 119239 8 92 0 0 0 2018-05-25 10:52:59 103 0 0 560176 1060 7153132 0 0 0 29436 34049 65098 8 92 0 0 0 2018-05-25 10:53:00 95 0 0 471204 1060 7241576 0 0 0 163508 42062 74730 7 92 0 0 0 2018-05-25 10:53:01 100 0 0 370152 1060 7322976 0 0 0 142236 45227 88302 8 92 0 0 0 2018-05-25 10:53:02 107 0 0 367424 1060 7338656 0 0 0 291676 74375 205658 7 93 0 0 0 2018-05-25 10:53:03 106 0 0 281288 1060 7423196 0 0 0 16436 39134 71585 8 92 0 0 0 2018-05-25 10:53:04 99 0 0 191476 1060 7513072 0 0 0 16984 31075 50493 8 92 0 0 0 2018-05-25 10:53:05 101 0 0 218328 1052 7476176 0 0 28 13204 31880 47376 7 93 0 0 0 2018-05-25 10:53:06 95 0 0 310764 1052 7395748 0 0 339632 7300 71844 161470 6 92 0 2 0 2018-05-25 10:53:07 104 0 0 253652 1052 7448908 0 0 8392 82784 53482 108682 7 93 0 0 0 2018-05-25 10:53:08 105 0 0 184260 1052 7516612 0 0 0 84728 49453 101418 7 93 0 0 0 2018-05-25 10:53:09 120 0 0 175316 1052 7521444 0 0 36 214628 33057 95648 8 92 0 0 0 2018-05-25 10:53:10 108 0 0 181940 1052 7515136 0 0 0 62644 28320 69892 8 92 0 0 0 2018-05-25 10:53:11 98 1 0 167556 1052 7524272 0 0 0 44440 34556 55278 8 92 0 0 0 2018-05-25 10:53:12 108 0 0 181500 1052 7515284 0 0 0 223656 79509 155366 7 93 0 0 0 2018-05-25 10:53:13 105 0 0 147276 1052 7547720 0 0 0 7416 63985 129401 7 93 0 0 0 2018-05-25 10:53:14 108 0 0 228660 1052 7464332 0 0 0 17208 48161 47322 8 92 0 0 0 2018-05-25 10:53:15 102 0 0 156392 1052 7536668 0 0 88 72316 72729 92978 8 92 0 0 0 2018-05-25 10:53:16 111 1 0 129904 1052 7558844 0 0 32 70436 62794 72422 8 92 0 0 0 2018-05-25 10:53:17 105 0 0 128576 1052 7568420 0 0 16 66228 79585 110304 8 92 0 0 0 2018-05-25 10:53:18 110 0 0 141064 1052 7536828 0 0 32 323988 102469 160187 6 94 0 0 0 2018-05-25 10:53:19 99 0 0 142156 1052 7555972 0 0 0 1848 112728 200118 6 94 0 0 0 2018-05-25 10:53:20 107 0 0 134116 1052 7560900 0 0 252 17600 71661 86141 8 92 0 0 0 2018-05-25 10:53:21 105 0 0 167772 12 7523648 0 0 144 17464 55513 75736 8 92 0 0 0 2018-05-25 10:53:22 110 0 0 140464 12 7548004 0 0 128 19924 50522 87637 8 92 0 0 0 2018-05-25 10:53:23 103 0 0 150604 12 7535444 0 0 16 88248 81610 123959 7 93 0 0 0 2018-05-25 10:53:24 98 0 0 200720 12 7502044 0 0 144 14672 41683 56971 9 91 0 0 0 2018-05-25 10:53:25 104 0 0 132944 12 7567844 0 0 128 199056 51161 101316 8 92 0 0 0 2018-05-25 10:53:26 103 0 0 128284 12 7570640 0 0 444 20180 32621 48968 8 92 0 0 0 2018-05-25 10:53:27 95 0 0 131396 12 7563404 0 0 276 87352 57011 115486 8 92 0 0 0 2018-05-25 10:53:28 102 0 0 129416 12 7578116 0 0 824 123800 34406 50301 7 93 0 0 0 2018-05-25 10:53:29 102 1 0 141736 12 7560268 0 0 96 356652 96246 177356 7 93 0 0 0 2018-05-25 10:53:30 96 0 0 129572 12 7573044 0 0 128 148688 114213 221997 6 94 0 0 0 2018-05-25 10:53:31 115 0 0 132316 12 7572556 0 0 208 1016 86217 176902 6 94 0 0 0 2018-05-25 10:53:32 95 0 0 130788 12 7573912 0 0 504 10920 48019 98535 7 93 0 0 0 2018-05-25 10:53:33 97 0 0 131148 12 7572972 0 0 1368 17332 32562 60243 8 92 0 0 0 2018-05-25 10:53:34 96 0 0 140520 12 7562444 0 0 2060 16632 33005 113232 8 92 0 0 0 2018-05-25 10:53:35 92 0 0 132088 12 7570052 0 0 2660 16004 32178 52290 8 92 0 0 0 2018-05-25 10:53:36 100 0 0 126372 12 7554188 0 0 1264 406692 59921 113499 7 93 0 0 0 2018-05-25 10:53:37 103 0 0 140372 12 7564444 0 0 304 35148 82014 186294 6 94 0 0 0 2018-05-25 10:53:38 105 0 0 136652 12 7565084 0 0 3316 11156 36601 66834 8 92 0 0 0 2018-05-25 10:53:39 87 2 0 143720 12 7539180 0 0 4732 201872 42553 80674 7 93 0 0 0 2018-05-25 10:53:40 94 1 0 129224 12 7570652 0 0 5700 8668 54472 117802 7 93 0 0 0 2018-05-25 10:53:41 101 8 0 140404 12 7556392 0 0 28656 123792 42629 71016 7 93 0 0 0 2018-05-25 10:53:42 110 4 0 145168 12 7554888 0 0 32168 45032 76235 152858 7 93 0 0 0 2018-05-25 10:53:43 61 36 0 147204 12 7541052 0 0 98020 100148 66951 124749 7 93 0 0 0 2018-05-25 10:53:44 106 0 0 131320 12 7570176 0 0 121044 6112 72974 128334 7 93 0 0 0 2018-05-25 10:53:45 102 4 0 127904 12 7574252 0 0 90608 45948 79039 150799 7 93 0 0 0 2018-05-25 10:53:46 93 7 0 152744 12 7548532 0 0 143928 37200 79768 176287 7 93 0 0 0 2018-05-25 10:53:47 108 0 0 150416 12 7547188 0 0 177768 73588 80760 176343 7 93 0 0 0 2018-05-25 10:53:48 101 0 0 127640 12 7573060 0 0 194340 33004 80137 143974 7 93 0 0 0 2018-05-25 10:53:49 73 5 0 144912 12 7552208 0 0 300124 36644 92662 163519 6 94 0 0 0 2018-05-25 10:53:50 102 6 0 136644 12 7565768 0 0 321000 83732 84754 164890 6 86 0 9 0 2018-05-25 10:53:51 83 23 0 130664 12 7572048 0 0 363420 30532 104885 191123 6 94 0 0 0 2018-05-25 10:53:52 25 43 0 159320 12 7540440 0 0 408548 35384 88153 145478 6 93 0 1 0 2018-05-25 10:53:53 44 66 0 127708 12 7577436 0 0 435460 6244 106368 206175 6 91 0 3 0 2018-05-25 10:53:54 106 4 0 129720 12 7577548 0 0 350288 93284 72255 208024 4 75 0 21 0 2018-05-25 10:53:55 32 77 0 162036 12 7542324 0 0 429888 11776 111876 243905 5 93 0 2 0 2018-05-25 10:53:56 102 6 0 144868 12 7559344 0 0 383992 52936 89887 294473 4 86 0 10 0 2018-05-25 10:53:57 19 82 0 132616 12 7571356 0 0 442692 6016 103999 247342 5 91 0 4 0 2018-05-25 10:53:58 0 88 0 128952 12 7571932 0 0 365476 91032 72113 267656 3 71 1 25 0 2018-05-25 10:53:59 68 47 0 125092 12 7579940 0 0 431052 18168 61535 201748 3 56 2 39 0 2018-05-25 10:54:00 16 88 0 128728 12 7577600 0 0 453848 4052 67658 264867 4 71 1 25 0 2018-05-25 10:54:01 25 83 0 152468 12 7552488 0 0 458924 3584 67690 351836 3 80 0 17 0 2018-05-25 10:54:02 14 88 0 133048 12 7572460 0 0 458028 3736 69708 347342 3 80 0 17 0 2018-05-25 10:54:03 10 74 0 128524 12 7576532 0 0 424732 25028 46989 157869 2 35 5 58 0 2018-05-25 10:54:04 5 85 0 154148 12 7549364 0 0 321940 109128 53520 220444 2 52 2 44 0 2018-05-25 10:54:05 65 35 0 136996 12 7568392 0 0 451332 100 53220 164556 2 36 4 58 0 2018-05-25 10:54:06 3 83 0 139560 12 7566068 0 0 438968 14044 88680 281950 3 79 1 18 0 2018-05-25 10:54:07 10 69 0 133368 12 7570540 0 0 457908 44 45224 165594 3 33 3 61 0 2018-05-25 10:54:08 25 64 0 130200 12 7574940 0 0 406796 45576 68021 229525 2 58 2 38 0 2018-05-25 10:54:09 25 83 0 143440 12 7562040 0 0 459564 4468 68324 374065 3 79 0 18 0 2018-05-25 10:54:10 4 84 0 125984 12 7578864 0 0 411348 38868 69510 291737 2 67 2 30 0 2018-05-25 10:54:11 5 57 0 143336 12 7561324 0 0 461212 24 44460 142072 2 26 6 65 0 2018-05-25 10:54:12 19 90 0 141468 12 7564396 0 0 458660 3312 74142 338372 2 78 0 19 0 2018-05-25 10:54:13 16 86 0 137788 12 7567612 0 0 454988 3548 65651 362384 2 79 0 19 0 2018-05-25 10:54:14 2 77 0 129456 12 7574680 0 0 346692 98392 46570 271064 2 50 2 46 0 2018-05-25 10:54:15 63 50 0 131188 12 7575232 0 0 452980 220 56708 178662 2 39 4 55 0 2018-05-25 10:54:16 44 66 0 140248 12 7566636 0 0 406440 43736 83251 294529 2 76 1 21 0 2018-05-25 10:54:17 7 42 0 135992 12 7567860 0 0 461236 1320 72214 282195 2 64 3 31 0 2018-05-25 10:54:18 10 95 0 125820 12 7581184 0 0 460676 3356 63433 350467 2 73 4 21 0 2018-05-25 10:54:19 28 88 0 127420 12 7577872 0 0 367444 83684 67967 334559 2 77 0 21 0 2018-05-25 10:54:20 30 80 0 145856 12 7558352 0 0 461340 2616 68477 368746 2 81 0 17 0 2018-05-25 10:54:21 22 92 0 124208 12 7579904 0 0 460572 3072 59933 384423 2 78 0 20 0 2018-05-25 10:54:22 13 93 0 134512 12 7571524 0 0 462384 3468 63635 385048 2 79 0 19 0 2018-05-25 10:54:23 15 96 0 141560 12 7566412 0 0 460620 2952 60522 389345 2 79 0 19 0 2018-05-25 10:54:24 34 80 0 151648 12 7553248 0 0 459008 3244 60553 387047 2 77 0 21 0 2018-05-25 10:54:25 3 96 0 128308 12 7577012 0 0 355600 81428 56175 320752 2 68 1 30 0 2018-05-25 10:54:26 4 49 0 160776 12 7542132 0 0 466224 1444 42163 128578 1 17 7 74 0 2018-05-25 10:54:27 4 55 0 134692 12 7571388 0 0 450740 8176 42377 127593 1 20 8 70 0 2018-05-25 10:54:28 21 77 0 141912 12 7564732 0 0 434572 25716 80099 326804 2 84 0 14 0 2018-05-25 10:54:29 7 79 0 127244 12 7581448 0 0 461624 504 85322 315433 2 81 1 17 0 2018-05-25 10:54:30 12 97 0 128012 12 7579492 0 0 365956 80164 62667 249926 2 57 1 40 0 2018-05-25 10:54:31 18 91 0 123156 12 7583512 0 0 457096 3480 57608 404524 2 75 0 23 0 2018-05-25 10:54:32 13 90 0 131860 12 7576668 0 0 464084 976 46273 197727 2 33 5 60 0 2018-05-25 10:54:33 19 83 0 147740 12 7559076 0 0 461612 2984 60151 393209 2 76 0 22 0 2018-05-25 10:54:34 15 90 0 125680 12 7579880 0 0 461400 3172 55877 392767 2 75 0 23 0 2018-05-25 10:54:35 32 80 0 131600 12 7576556 0 0 455980 3708 59214 393334 2 77 0 21 0 2018-05-25 10:54:36 1 83 0 155320 12 7550880 0 0 285704 148548 46771 248893 1 49 3 48 0 2018-05-25 10:54:37 3 68 0 136292 12 7569292 0 0 464536 28 43051 129006 1 18 6 75 0 2018-05-25 10:54:38 21 69 0 148468 12 7558352 0 0 464088 164 46794 151665 1 26 5 67 0 2018-05-25 10:54:39 29 65 0 128772 12 7579208 0 0 459908 828 89147 336066 2 88 0 10 0 2018-05-25 10:54:40 20 74 0 136820 12 7572748 0 0 455876 692 77579 270405 2 68 1 30 0 2018-05-25 10:54:41 21 88 0 138180 12 7570556 0 0 404400 55656 68642 373084 1 80 0 18 0 2018-05-25 10:54:42 18 91 0 125984 12 7584264 0 0 462500 3436 55759 400911 1 74 0 25 0 2018-05-25 10:54:43 39 72 0 123708 12 7585192 0 0 388932 60192 56723 390531 2 77 0 22 0 2018-05-25 10:54:44 6 88 0 135136 12 7576400 0 0 452568 6896 70968 292847 1 68 2 29 0 2018-05-25 10:54:45 26 84 0 128992 12 7582760 0 0 459840 2136 67795 345036 2 71 1 26 0 2018-05-25 10:54:46 13 88 0 148648 12 7559860 0 0 464244 3128 57510 401102 1 75 0 23 0 2018-05-25 10:54:47 17 91 0 142924 12 7567792 0 0 463624 3868 59825 404793 1 77 0 21 0 2018-05-25 10:54:48 13 88 0 153400 12 7555804 0 0 465716 1152 47891 215800 1 37 7 55 0 2018-05-25 10:54:49 29 84 0 137264 12 7574336 0 0 463112 3212 55033 404905 1 73 0 26 0 2018-05-25 10:54:50 8 92 0 125828 12 7587032 0 0 286440 169260 45394 373984 1 69 0 30 0 2018-05-25 10:54:51 39 77 0 142928 12 7568352 0 0 467568 416 64610 229365 1 52 3 45 0 2018-05-25 10:54:52 22 86 0 135752 12 7575000 0 0 462232 1152 83941 342770 1 87 0 11 0 2018-05-25 10:54:53 22 89 0 144524 12 7567788 0 0 447620 17752 72915 363659 1 83 0 16 0 2018-05-25 10:54:54 18 88 0 127536 12 7585088 0 0 454676 3016 55525 409097 2 74 0 24 0 2018-05-25 10:54:55 8 98 0 123476 12 7588136 0 0 465372 2872 56825 399782 1 75 0 23 0 2018-05-25 10:54:56 10 97 0 139580 12 7572000 0 0 463156 3564 58186 397539 1 76 0 22 0 2018-05-25 10:54:57 28 86 0 132888 12 7578408 0 0 382600 72872 70557 357498 1 84 0 15 0 2018-05-25 10:54:58 18 91 0 133104 12 7580756 0 0 462588 2024 75006 363324 1 85 0 14 0 2018-05-25 10:54:59 0 80 0 167068 12 7542132 0 0 424600 48140 44702 215490 1 37 4 58 0 2018-05-25 10:55:00 46 71 0 129464 12 7586356 0 0 409808 33528 52014 207451 1 41 6 52 0 2018-05-25 10:55:01 13 97 0 127056 12 7588128 0 0 464984 2548 58859 397456 1 76 0 22 0 2018-05-25 10:55:02 25 86 0 123300 12 7592004 0 0 467380 2716 53126 399670 1 72 0 26 0 2018-05-25 10:55:03 19 94 0 138784 12 7567800 0 0 430000 23964 57737 395437 1 77 0 22 0 2018-05-25 10:55:04 4 66 0 138620 12 7575472 0 0 399024 42648 43191 153707 1 26 6 67 0 2018-05-25 10:55:05 2 42 0 126884 12 7588952 0 0 465508 0 42710 122554 1 15 10 73 0 2018-05-25 10:55:06 32 78 0 139268 12 7578668 0 0 464460 996 78089 321909 1 82 1 16 0 2018-05-25 10:55:07 4 87 0 131232 12 7585844 0 0 468296 1036 58482 248948 1 50 3 46 0 2018-05-25 10:55:08 8 65 0 129248 12 7589928 0 0 466604 36 42296 121164 1 16 7 77 0 2018-05-25 10:55:09 11 90 0 163560 12 7556352 0 0 464428 3236 55759 380750 1 72 1 27 0 2018-05-25 10:55:10 16 90 0 131032 12 7589088 0 0 463724 3368 55930 405018 1 75 0 23 0 2018-05-25 10:55:11 12 92 0 140964 12 7577896 0 0 466052 3572 51224 411231 1 71 0 28 0 2018-05-25 10:55:12 18 92 0 144412 12 7579740 0 0 469800 3060 53935 399507 1 75 0 24 0 2018-05-25 10:55:13 2 88 0 133092 12 7586020 0 0 271948 165852 39032 280783 1 49 2 48 0 2018-05-25 10:55:14 1 72 0 148872 12 7570604 0 0 461588 3744 41329 114644 1 14 8 78 0 2018-05-25 10:55:15 29 71 0 133436 12 7589816 0 0 472036 444 57861 213671 1 45 5 49 0 2018-05-25 10:55:16 28 84 0 123660 12 7601992 0 0 397452 58904 75625 344724 1 86 0 13 0 2018-05-25 10:55:17 14 93 0 142324 12 7581284 0 0 474728 612 77816 323599 1 80 1 18 0 2018-05-25 10:55:18 18 93 0 130272 12 7596256 0 0 463936 3156 53329 404617 1 73 0 26 0 2018-05-25 10:55:19 21 88 0 145188 12 7582092 0 0 468088 3040 53783 408047 1 74 0 24 0 2018-05-25 10:55:20 19 91 0 132212 12 7595184 0 0 466688 2624 52277 405081 1 73 0 26 0 2018-05-25 10:55:21 14 93 0 145984 12 7582668 0 0 464268 2808 56063 409300 1 75 0 24 0 2018-05-25 10:55:22 17 88 0 130964 12 7600100 0 0 464272 2844 53509 407264 1 74 0 25 0 2018-05-25 10:55:23 18 87 0 135292 12 7595264 0 0 468236 2948 53109 409275 1 74 0 25 0 2018-05-25 10:55:24 15 98 0 131916 12 7583360 0 0 329600 146720 44764 348545 1 64 2 33 0 2018-05-25 10:55:25 24 88 0 140700 12 7588104 0 0 415752 28432 74700 364957 1 84 0 15 0 2018-05-25 10:55:26 3 70 0 136820 12 7592552 0 0 473268 120 43017 141574 1 18 7 74 0 2018-05-25 10:55:27 36 63 0 144036 12 7588420 0 0 454468 12092 47240 153448 1 27 7 65 0 2018-05-25 10:55:28 38 73 0 140424 12 7595680 0 0 464664 852 86175 354927 1 88 0 11 0 2018-05-25 10:55:29 14 85 0 127796 12 7608428 0 0 468556 2184 69737 382042 1 81 0 18 0 2018-05-25 10:55:30 2 99 0 142224 12 7593560 0 0 464772 2752 52617 401038 1 71 0 28 0 2018-05-25 10:55:31 11 98 0 151792 12 7584680 0 0 406996 48140 44996 276431 1 46 3 50 0 2018-05-25 10:55:32 25 91 0 134344 12 7602504 0 0 466012 3084 53925 398930 1 74 0 25 0 2018-05-25 10:55:33 2 92 0 150372 12 7584904 0 0 403116 51480 49011 300757 1 54 2 44 0 2018-05-25 10:55:34 3 56 0 130824 12 7607476 0 0 464472 0 41033 114947 1 13 8 78 0 2018-05-25 10:55:35 1 83 0 149132 12 7589948 0 0 436024 27688 42445 149882 1 22 9 68 0 2018-05-25 10:55:36 28 81 0 134752 12 7605860 0 0 470064 852 76205 326665 1 80 1 18 0 2018-05-25 10:55:37 1 91 0 133192 12 7606820 0 0 470940 516 62245 247725 1 55 3 41 0 2018-05-25 10:55:38 3 74 0 123832 12 7616164 0 0 467060 16 42534 120591 1 15 7 77 0 2018-05-25 10:55:39 12 95 0 146936 12 7596260 0 0 467948 2180 56478 379203 1 70 1 28 0 2018-05-25 10:55:40 3 99 0 124704 12 7617116 0 0 459916 3000 53344 396771 1 74 0 25 0 2018-05-25 10:55:41 15 91 0 123440 12 7622804 0 0 466148 2696 50764 404367 1 72 0 26 0 2018-05-25 10:55:42 16 87 0 127932 12 7617208 0 0 472684 2676 48534 420990 1 70 0 29 0 2018-05-25 10:55:43 4 95 0 138632 12 7599360 0 0 349856 101288 48070 373995 1 68 0 31 0 2018-05-25 10:55:44 2 72 0 129728 12 7616144 0 0 448288 16164 40698 118181 1 14 6 80 0 2018-05-25 10:55:45 4 60 0 125208 12 7619848 0 0 456904 7008 40438 109829 1 13 8 78 0 2018-05-25 10:55:46 3 95 0 141124 12 7599884 0 0 405484 63184 62867 289148 1 66 2 31 0 2018-05-25 10:55:47 28 88 0 145484 12 7601928 0 0 433120 18520 64135 271809 1 63 2 35 0 2018-05-25 10:55:48 15 92 0 147148 12 7601408 0 0 462568 1300 80034 354519 1 86 0 13 0 2018-05-25 10:55:49 8 98 0 125792 12 7623252 0 0 470768 3700 53216 411514 1 74 0 25 0 2018-05-25 10:55:50 6 95 0 129764 12 7620244 0 0 470108 3288 53767 413409 1 74 0 24 0 2018-05-25 10:55:51 5 84 0 127140 12 7623932 0 0 465492 2332 47989 284335 1 48 3 48 0 2018-05-25 10:55:52 40 78 0 138880 12 7612204 0 0 390188 68852 52495 221759 1 47 4 48 0 2018-05-25 10:55:53 20 85 0 123188 12 7630764 0 0 466672 736 81016 351115 1 87 0 11 0 2018-05-25 10:55:54 7 97 0 125984 12 7626708 0 0 469696 3300 51302 409227 1 73 0 26 0 2018-05-25 10:55:55 11 95 0 141688 12 7610416 0 0 384532 74896 64132 365018 1 80 0 19 0 2018-05-25 10:55:56 21 87 0 123200 12 7629232 0 0 464460 2868 50946 419402 1 72 0 27 0 2018-05-25 10:55:57 3 94 0 131540 12 7624856 0 0 471660 3340 51116 408406 1 72 0 27 0 2018-05-25 10:55:58 12 95 0 130336 12 7625108 0 0 473176 2748 50958 411308 1 73 0 26 0 2018-05-25 10:55:59 5 95 0 127392 12 7627764 0 0 473440 2928 49458 422309 1 72 0 27 0 2018-05-25 10:56:00 17 89 0 127028 12 7632428 0 0 462496 2940 52613 421843 1 73 0 26 0 2018-05-25 10:56:01 18 92 0 123284 12 7633444 0 0 471500 3084 52465 411338 1 73 0 26 0 2018-05-25 10:56:02 13 96 0 139752 12 7617432 0 0 455804 2672 49023 407753 1 71 0 28 0 2018-05-25 10:56:03 29 84 0 127968 12 7628216 0 0 211704 232320 37133 263874 0 48 2 50 0 2018-05-25 10:56:04 27 84 0 143956 12 7613016 0 0 468584 984 79061 350542 1 87 0 12 0 2018-05-25 10:56:05 27 85 0 150236 12 7607000 0 0 465788 612 73258 308956 1 74 1 24 0 2018-05-25 10:56:06 26 88 0 137516 12 7620828 0 0 433840 21796 73747 349209 1 86 0 13 0 2018-05-25 10:56:07 10 95 0 145120 12 7611632 0 0 471400 1308 71191 364605 1 84 0 15 0 2018-05-25 10:56:08 25 85 0 146404 12 7612680 0 0 474812 2672 49291 412657 1 73 0 26 0 2018-05-25 10:56:09 3 80 0 124672 12 7633560 0 0 468196 1976 47474 271141 1 44 3 53 0 2018-05-25 10:56:10 1 44 0 147636 12 7611600 0 0 471100 0 40983 112988 1 12 10 77 0 2018-05-25 10:56:11 18 85 0 130284 12 7629452 0 0 468816 832 46500 208843 1 34 10 55 0 2018-05-25 10:56:12 17 88 0 124224 12 7636852 0 0 463196 3112 53463 406813 1 73 0 26 0 2018-05-25 10:56:13 8 97 0 125404 12 7636864 0 0 473332 3192 51280 411285 1 73 0 25 0 2018-05-25 10:56:14 10 101 0 144440 12 7596420 0 0 362004 99696 48381 387828 1 70 0 29 0 2018-05-25 10:56:15 5 65 0 143772 12 7616072 0 0 408988 40920 38223 117197 1 12 5 82 0 2018-05-25 10:56:16 1 35 0 148172 12 7612884 0 0 472928 0 40599 108029 1 11 11 77 0 2018-05-25 10:56:17 18 88 0 126088 12 7638176 0 0 468828 904 76352 332253 1 81 2 17 0 2018-05-25 10:56:18 16 92 0 139132 12 7622432 0 0 473760 1140 73054 363085 1 84 0 15 0 2018-05-25 10:56:19 13 93 0 134692 12 7626848 0 0 470556 2848 49072 422913 1 70 0 28 0 2018-05-25 10:56:20 25 85 0 131520 12 7632208 0 0 464632 3356 50943 416423 1 73 0 26 0 2018-05-25 10:56:21 18 90 0 137408 12 7626916 0 0 470592 3176 51964 419548 1 73 0 26 0 2018-05-25 10:56:22 13 92 0 124492 12 7639472 0 0 468912 3076 51905 420745 1 73 0 26 0 2018-05-25 10:56:23 14 92 0 126016 12 7639384 0 0 468336 3320 51071 416386 1 72 0 27 0 2018-05-25 10:56:24 21 87 0 142652 12 7619092 0 0 469728 3272 52672 414698 1 73 0 26 0 2018-05-25 10:56:25 1 85 0 152400 12 7609020 0 0 471232 1648 40773 117302 1 13 7 78 0 2018-05-25 10:56:26 4 68 0 147480 12 7616508 0 0 469280 3300 39542 102849 1 10 8 81 0 2018-05-25 10:56:27 20 99 0 190992 12 7565312 0 0 274324 196264 32609 187996 0 29 5 65 0 2018-05-25 10:56:28 20 83 0 149260 12 7612240 0 0 423732 25656 71971 355225 1 85 0 14 0 2018-05-25 10:56:29 31 76 0 132484 12 7630224 0 0 465440 812 78426 358513 1 87 0 12 0 2018-05-25 10:56:30 25 89 0 128768 12 7637680 0 0 468748 704 78327 350697 1 87 0 12 0 2018-05-25 10:56:31 18 93 0 154140 12 7610140 0 0 468364 1952 62361 380200 1 79 0 20 0 2018-05-25 10:56:32 23 93 0 147660 12 7616904 0 0 422292 49648 51943 389850 1 74 0 25 0 2018-05-25 10:56:33 21 87 0 125596 12 7640908 0 0 468064 1100 76707 354450 1 86 0 13 0 2018-05-25 10:56:34 0 64 0 149168 12 7617212 0 0 472076 3404 48712 382455 1 65 1 33 0 2018-05-25 10:56:35 0 0 0 156236 12 7620200 0 0 76004 43548 19972 15272 0 2 88 10 0 2018-05-25 10:56:36 ^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [LKP] [lkp-robot] [mm] 9092c71bb7: blogbench.write_score -12.3% regression 2018-08-03 8:22 ` Huang, Ying @ 2018-08-29 6:55 ` Huang, Ying 0 siblings, 0 replies; 11+ messages in thread From: Huang, Ying @ 2018-08-29 6:55 UTC (permalink / raw) To: Josef Bacik; +Cc: Josef Bacik, Chris Mason, David Sterba, lkp, linux-btrfs "Huang, Ying" <ying.huang@intel.com> writes: > Josef Bacik <josef@toxicpanda.com> writes: > >> On Thu, Aug 02, 2018 at 01:55:23PM +0800, Huang, Ying wrote: >>> "Huang, Ying" <ying.huang@intel.com> writes: >>> >>> > Hi, Chris, >>> > >>> > Chris Mason <clm@fb.com> writes: >>> > >>> >> On 19 Jun 2018, at 23:51, Huang, Ying wrote: >>> >>>>> "Huang, Ying" <ying.huang@intel.com> writes: >>> >>>>> >>> >>>>>> Hi, Josef, >>> >>>>>> >>> >>>>>> Do you have time to take a look at the regression? >>> >>>>>> >>> >>>>>> kernel test robot <xiaolong.ye@intel.com> writes: >>> >>>>>> >>> >>>>>>> Greeting, >>> >>>>>>> >>> >>>>>>> FYI, we noticed a -12.3% regression of blogbench.write_score and >>> >>>>>>> a +9.6% improvement >>> >>>>>>> of blogbench.read_score due to commit: >>> >>>>>>> >>> >>>>>>> >>> >>>>>>> commit: 9092c71bb724dba2ecba849eae69e5c9d39bd3d2 ("mm: use >>> >>>>>>> sc->priority for slab shrink targets") >>> >>>>>>> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git >>> >>>>>>> master >>> >>>>>>> >>> >>>>>>> in testcase: blogbench >>> >>>>>>> on test machine: 16 threads Intel(R) Xeon(R) CPU D-1541 @ >>> >>>>>>> 2.10GHz with 8G memory >>> >>>>>>> with following parameters: >>> >>>>>>> >>> >>>>>>> disk: 1SSD >>> >>>>>>> fs: btrfs >>> >>>>>>> cpufreq_governor: performance >>> >>>>>>> >>> >>>>>>> test-description: Blogbench is a portable filesystem benchmark >>> >>>>>>> that tries to reproduce the load of a real-world busy file >>> >>>>>>> server. >>> >>>>>>> test-url: >>> >> >>> >> I'm surprised, this patch is a big win in production here at FB. I'll >>> >> have to reproduce these results to better understand what is going on. >>> >> My first guess is that since we have fewer inodes in slab, we're >>> >> reading more inodes from disk in order to do the writes. >>> >> >>> >> But that should also make our read scores lower. >>> > >>> > Any update on this? >>> >>> Ping. >>> >> >> I can't reproduce this, and what's more it appears that blogbench doesn't use >> much memory at all. I have the slab shrinking tracepoints on and we never go >> into this code at all, so I'm pretty sure these results are bogus. How are you >> running blogbench? I'm doing blogbench -d /whatever, if I need to be doing >> something else let me know. But from what I can tell this thing uses less than >> 100m of memory, and on an 8gig of ram box we're never going to trip over this >> code. Thanks, > > Thanks for looking at this. In my testing, blogbench will eat up system > memory. Please check the vmstat result attached. The SSD disk size is > about 745GB. Hi, Josef, Do you need more information? Best Regards, Huang, Ying > Best Regards, > Huang, Ying ^ permalink raw reply [flat|nested] 11+ messages in thread
end of thread, other threads:[~2018-08-29 10:51 UTC | newest]
Thread overview: 11+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
[not found] <20180408015739.GN3845@yexl-desktop>
2018-05-29 7:30 ` [LKP] [lkp-robot] [mm] 9092c71bb7: blogbench.write_score -12.3% regression Huang, Ying
2018-06-05 4:58 ` Huang, Ying
2018-06-14 1:37 ` Huang, Ying
2018-06-20 3:51 ` Huang, Ying
2018-06-20 12:38 ` Chris Mason
2018-06-21 0:38 ` Huang, Ying
2018-07-13 1:55 ` Huang, Ying
2018-08-02 5:55 ` Huang, Ying
2018-08-02 16:23 ` Josef Bacik
2018-08-03 8:22 ` Huang, Ying
2018-08-29 6:55 ` Huang, Ying
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for NNTP newsgroup(s).