Hello,
kernel test robot noticed a 11.4% improvement of will-it-scale.per_thread_ops on:
commit: 7c33b8c4229af19797c78de48827ca70228c1f47 ("mm: remove use of folio list from folios_put()")
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master
testcase: will-it-scale
test machine: 104 threads 2 sockets (Skylake) with 192G memory
parameters:
nr_task: 16
mode: thread
test: page_fault2
cpufreq_governor: performance
Details are as below:
-------------------------------------------------------------------------------------------------->
The kernel config and materials to reproduce are available at:
https://download.01.org/0day-ci/archive/20240325/[email protected]
=========================================================================================
compiler/cpufreq_governor/kconfig/mode/nr_task/rootfs/tbox_group/test/testcase:
gcc-12/performance/x86_64-rhel-8.3/thread/16/debian-12-x86_64-20240206.cgz/lkp-skl-fpga01/page_fault2/will-it-scale
commit:
4882c80975 ("memcg: add mem_cgroup_uncharge_folios()")
7c33b8c422 ("mm: remove use of folio list from folios_put()")
4882c80975e2bf72 7c33b8c4229af19797c78de4882
---------------- ---------------------------
%stddev %change %stddev
\ | \
9.19 +1.1 10.32 mpstat.cpu.all.sys%
2.00 +0.2 2.23 mpstat.cpu.all.usr%
1.171e+09 +11.0% 1.3e+09 numa-numastat.node0.local_node
1.173e+09 +10.9% 1.301e+09 numa-numastat.node0.numa_hit
12.80 +10.6% 14.16 vmstat.procs.r
2179 +2.7% 2239 vmstat.system.cs
28134 +10.1% 30977 vmstat.system.in
3858794 +11.4% 4296995 will-it-scale.16.threads
88.59 -1.6% 87.20 will-it-scale.16.threads_idle
241174 +11.4% 268562 will-it-scale.per_thread_ops
3858794 +11.4% 4296995 will-it-scale.workload
4597109 ? 16% -30.5% 3195509 ? 34% numa-meminfo.node0.FilePages
75985 ? 12% -35.5% 48981 ? 34% numa-meminfo.node0.KReclaimable
75985 ? 12% -35.5% 48981 ? 34% numa-meminfo.node0.SReclaimable
2516178 ? 30% -55.8% 1113291 ? 99% numa-meminfo.node0.Unevictable
6475 ?125% +199.9% 19417 ? 59% numa-meminfo.node1.Mapped
1149260 ? 16% -30.6% 797998 ? 34% numa-vmstat.node0.nr_file_pages
18996 ? 12% -35.5% 12243 ? 34% numa-vmstat.node0.nr_slab_reclaimable
629044 ? 30% -55.8% 278322 ? 99% numa-vmstat.node0.nr_unevictable
629044 ? 30% -55.8% 278322 ? 99% numa-vmstat.node0.nr_zone_unevictable
1.173e+09 +10.9% 1.301e+09 numa-vmstat.node0.numa_hit
1.171e+09 +11.1% 1.3e+09 numa-vmstat.node0.numa_local
1665 ?124% +199.4% 4986 ? 59% numa-vmstat.node1.nr_mapped
2254868 ? 8% +15.4% 2602210 ? 11% sched_debug.cfs_rq:/.avg_vruntime.max
671734 ? 10% +19.6% 803578 ? 11% sched_debug.cfs_rq:/.avg_vruntime.stddev
2254868 ? 8% +15.4% 2602210 ? 11% sched_debug.cfs_rq:/.min_vruntime.max
671734 ? 10% +19.6% 803578 ? 11% sched_debug.cfs_rq:/.min_vruntime.stddev
95.28 ? 3% +12.6% 107.30 ? 6% sched_debug.cfs_rq:/.util_est.avg
530.36 ? 4% +12.6% 597.03 ? 4% sched_debug.cpu.curr->pid.avg
0.13 ? 5% +12.9% 0.15 ? 4% sched_debug.cpu.nr_running.avg
0.03 ? 14% -39.9% 0.02 ? 29% sched_debug.cpu.nr_uninterruptible.avg
25765 +2.9% 26515 proc-vmstat.nr_active_anon
463751 -4.3% 443812 proc-vmstat.nr_anon_pages
988326 -2.0% 968357 proc-vmstat.nr_inactive_anon
25765 +2.9% 26515 proc-vmstat.nr_zone_active_anon
988327 -2.0% 968357 proc-vmstat.nr_zone_inactive_anon
1.173e+09 +10.9% 1.302e+09 proc-vmstat.numa_hit
1.171e+09 +11.1% 1.301e+09 proc-vmstat.numa_local
21662 +5.2% 22790 proc-vmstat.pgactivate
1.168e+09 +11.2% 1.299e+09 proc-vmstat.pgalloc_normal
1.164e+09 +11.2% 1.295e+09 proc-vmstat.pgfault
1.168e+09 +11.2% 1.298e+09 proc-vmstat.pgfree
3.587e+09 +10.5% 3.964e+09 perf-stat.i.branch-instructions
23777908 +7.3% 25506059 perf-stat.i.branch-misses
3.201e+08 +11.4% 3.567e+08 perf-stat.i.cache-misses
4.1e+08 +13.2% 4.641e+08 perf-stat.i.cache-references
2139 +3.3% 2210 perf-stat.i.context-switches
3.885e+10 +10.8% 4.305e+10 perf-stat.i.cpu-cycles
1.844e+10 +10.1% 2.031e+10 perf-stat.i.instructions
74.11 +11.2% 82.38 perf-stat.i.metric.K/sec
3853671 +11.2% 4283769 perf-stat.i.minor-faults
3853671 +11.2% 4283769 perf-stat.i.page-faults
0.66 -0.1 0.54 ? 44% perf-stat.overall.branch-miss-rate%
1442267 -17.6% 1188877 ? 44% perf-stat.overall.path-length
2.16 -0.6 1.52 perf-profile.calltrace.cycles-pp.zap_present_ptes.zap_pte_range.zap_pmd_range.unmap_page_range.unmap_vmas
5.70 -0.5 5.18 ? 2% perf-profile.calltrace.cycles-pp.__munmap
5.70 -0.5 5.18 ? 2% perf-profile.calltrace.cycles-pp.__vm_munmap.__x64_sys_munmap.do_syscall_64.entry_SYSCALL_64_after_hwframe.__munmap
5.70 -0.5 5.18 ? 2% perf-profile.calltrace.cycles-pp.__x64_sys_munmap.do_syscall_64.entry_SYSCALL_64_after_hwframe.__munmap
5.70 -0.5 5.18 ? 2% perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.__munmap
5.70 -0.5 5.18 ? 2% perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_after_hwframe.__munmap
5.69 -0.5 5.17 ? 2% perf-profile.calltrace.cycles-pp.unmap_region.do_vmi_align_munmap.do_vmi_munmap.__vm_munmap.__x64_sys_munmap
5.69 -0.5 5.17 ? 2% perf-profile.calltrace.cycles-pp.do_vmi_align_munmap.do_vmi_munmap.__vm_munmap.__x64_sys_munmap.do_syscall_64
5.69 -0.5 5.17 ? 2% perf-profile.calltrace.cycles-pp.do_vmi_munmap.__vm_munmap.__x64_sys_munmap.do_syscall_64.entry_SYSCALL_64_after_hwframe
5.27 -0.5 4.75 ? 2% perf-profile.calltrace.cycles-pp.zap_pte_range.zap_pmd_range.unmap_page_range.unmap_vmas.unmap_region
5.28 -0.5 4.76 ? 2% perf-profile.calltrace.cycles-pp.unmap_page_range.unmap_vmas.unmap_region.do_vmi_align_munmap.do_vmi_munmap
5.28 -0.5 4.76 ? 2% perf-profile.calltrace.cycles-pp.unmap_vmas.unmap_region.do_vmi_align_munmap.do_vmi_munmap.__vm_munmap
5.28 -0.5 4.76 ? 2% perf-profile.calltrace.cycles-pp.zap_pmd_range.unmap_page_range.unmap_vmas.unmap_region.do_vmi_align_munmap
4.92 -0.4 4.54 perf-profile.calltrace.cycles-pp.folio_prealloc.do_fault.__handle_mm_fault.handle_mm_fault.do_user_addr_fault
31.58 -0.3 31.25 perf-profile.calltrace.cycles-pp.secondary_startup_64_no_verify
1.80 ? 2% -0.2 1.65 ? 4% perf-profile.calltrace.cycles-pp.asm_sysvec_apic_timer_interrupt.cpuidle_enter_state.cpuidle_enter.cpuidle_idle_call.do_idle
1.66 ? 2% -0.1 1.52 ? 4% perf-profile.calltrace.cycles-pp.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt.cpuidle_enter_state.cpuidle_enter.cpuidle_idle_call
1.12 ? 2% -0.1 1.03 ? 3% perf-profile.calltrace.cycles-pp.__sysvec_apic_timer_interrupt.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt.cpuidle_enter_state.cpuidle_enter
1.10 ? 2% -0.1 1.01 ? 3% perf-profile.calltrace.cycles-pp.hrtimer_interrupt.__sysvec_apic_timer_interrupt.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt.cpuidle_enter_state
0.86 ? 3% -0.1 0.80 ? 2% perf-profile.calltrace.cycles-pp.__hrtimer_run_queues.hrtimer_interrupt.__sysvec_apic_timer_interrupt.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt
2.13 +0.1 2.20 perf-profile.calltrace.cycles-pp.shmem_fault.__do_fault.do_fault.__handle_mm_fault.handle_mm_fault
0.60 ? 2% +0.1 0.69 ? 2% perf-profile.calltrace.cycles-pp.folio_remove_rmap_ptes.zap_present_ptes.zap_pte_range.zap_pmd_range.unmap_page_range
2.26 +0.1 2.35 perf-profile.calltrace.cycles-pp.__do_fault.do_fault.__handle_mm_fault.handle_mm_fault.do_user_addr_fault
0.59 ? 4% +0.1 0.72 ? 13% perf-profile.calltrace.cycles-pp.folio_lruvec_lock_irqsave.folios_put_refs.release_pages.__tlb_batch_free_encoded_pages.tlb_flush_mmu
0.58 ? 4% +0.1 0.72 ? 13% perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.folio_lruvec_lock_irqsave.folios_put_refs.release_pages.__tlb_batch_free_encoded_pages
0.81 ? 4% +0.2 0.96 ? 10% perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.folio_lruvec_lock_irqsave.folio_batch_move_lru.folio_add_lru_vma.set_pte_range
0.82 ? 4% +0.2 0.98 ? 10% perf-profile.calltrace.cycles-pp.folio_lruvec_lock_irqsave.folio_batch_move_lru.folio_add_lru_vma.set_pte_range.finish_fault
0.71 ? 4% +0.2 0.87 ? 11% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.folio_lruvec_lock_irqsave.folio_batch_move_lru.folio_add_lru_vma
2.85 +0.2 3.05 ? 5% perf-profile.calltrace.cycles-pp.folio_batch_move_lru.folio_add_lru_vma.set_pte_range.finish_fault.do_fault
3.06 +0.2 3.27 ? 4% perf-profile.calltrace.cycles-pp.folio_add_lru_vma.set_pte_range.finish_fault.do_fault.__handle_mm_fault
0.45 ? 44% +0.2 0.67 ? 14% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irqsave.folio_lruvec_lock_irqsave.folios_put_refs.release_pages
9.85 +0.3 10.17 perf-profile.calltrace.cycles-pp.finish_fault.do_fault.__handle_mm_fault.handle_mm_fault.do_user_addr_fault
14.80 +0.3 15.15 perf-profile.calltrace.cycles-pp.copy_page.do_fault.__handle_mm_fault.handle_mm_fault.do_user_addr_fault
32.43 +0.4 32.82 perf-profile.calltrace.cycles-pp.do_fault.__handle_mm_fault.handle_mm_fault.do_user_addr_fault.exc_page_fault
33.75 +0.4 34.18 perf-profile.calltrace.cycles-pp.__handle_mm_fault.handle_mm_fault.do_user_addr_fault.exc_page_fault.asm_exc_page_fault
35.10 +0.5 35.56 perf-profile.calltrace.cycles-pp.handle_mm_fault.do_user_addr_fault.exc_page_fault.asm_exc_page_fault.testcase
38.62 +0.6 39.21 perf-profile.calltrace.cycles-pp.do_user_addr_fault.exc_page_fault.asm_exc_page_fault.testcase
38.89 +0.6 39.48 perf-profile.calltrace.cycles-pp.exc_page_fault.asm_exc_page_fault.testcase
0.00 +0.7 0.68 perf-profile.calltrace.cycles-pp.free_unref_folios.folios_put_refs.release_pages.__tlb_batch_free_encoded_pages.tlb_flush_mmu
54.53 +0.8 55.31 perf-profile.calltrace.cycles-pp.asm_exc_page_fault.testcase
60.07 +0.9 60.95 perf-profile.calltrace.cycles-pp.testcase
1.06 ? 2% -0.8 0.22 ? 7% perf-profile.children.cycles-pp._compound_head
2.20 -0.6 1.57 perf-profile.children.cycles-pp.zap_present_ptes
5.85 -0.5 5.32 ? 2% perf-profile.children.cycles-pp.entry_SYSCALL_64_after_hwframe
5.70 -0.5 5.18 ? 2% perf-profile.children.cycles-pp.__munmap
5.70 -0.5 5.18 ? 2% perf-profile.children.cycles-pp.do_vmi_munmap
5.84 -0.5 5.32 ? 2% perf-profile.children.cycles-pp.do_syscall_64
5.69 -0.5 5.17 ? 2% perf-profile.children.cycles-pp.unmap_region
5.70 -0.5 5.18 ? 2% perf-profile.children.cycles-pp.__vm_munmap
5.70 -0.5 5.18 ? 2% perf-profile.children.cycles-pp.__x64_sys_munmap
5.70 -0.5 5.18 ? 2% perf-profile.children.cycles-pp.do_vmi_align_munmap
5.28 -0.5 4.77 ? 2% perf-profile.children.cycles-pp.unmap_vmas
5.28 -0.5 4.76 ? 2% perf-profile.children.cycles-pp.zap_pte_range
5.28 -0.5 4.77 ? 2% perf-profile.children.cycles-pp.unmap_page_range
5.28 -0.5 4.77 ? 2% perf-profile.children.cycles-pp.zap_pmd_range
0.85 ? 6% -0.4 0.46 ? 5% perf-profile.children.cycles-pp.__folio_throttle_swaprate
4.98 -0.4 4.60 perf-profile.children.cycles-pp.folio_prealloc
31.58 -0.3 31.25 perf-profile.children.cycles-pp.cpu_startup_entry
31.50 -0.3 31.18 perf-profile.children.cycles-pp.cpuidle_idle_call
31.58 -0.3 31.25 perf-profile.children.cycles-pp.secondary_startup_64_no_verify
31.58 -0.3 31.25 perf-profile.children.cycles-pp.do_idle
31.04 -0.3 30.75 perf-profile.children.cycles-pp.cpuidle_enter_state
31.06 -0.3 30.77 perf-profile.children.cycles-pp.cpuidle_enter
2.26 -0.1 2.14 ? 3% perf-profile.children.cycles-pp.asm_sysvec_apic_timer_interrupt
28.90 -0.1 28.79 perf-profile.children.cycles-pp.intel_idle_ibrs
2.06 -0.1 1.95 ? 3% perf-profile.children.cycles-pp.sysvec_apic_timer_interrupt
1.45 -0.0 1.40 ? 2% perf-profile.children.cycles-pp.hrtimer_interrupt
0.20 ? 3% -0.0 0.17 ? 4% perf-profile.children.cycles-pp.__rmqueue_pcplist
0.33 ? 2% -0.0 0.30 ? 4% perf-profile.children.cycles-pp.menu_select
0.18 ? 4% -0.0 0.16 ? 3% perf-profile.children.cycles-pp.free_swap_cache
0.20 ? 2% -0.0 0.18 perf-profile.children.cycles-pp.free_pages_and_swap_cache
0.22 ? 4% +0.0 0.24 ? 5% perf-profile.children.cycles-pp.cgroup_rstat_updated
0.06 ? 7% +0.0 0.09 ? 6% perf-profile.children.cycles-pp.__tlb_remove_folio_pages_size
0.49 ? 2% +0.0 0.52 ? 2% perf-profile.children.cycles-pp.__mod_node_page_state
0.14 ? 4% +0.0 0.18 ? 2% perf-profile.children.cycles-pp.get_pfnblock_flags_mask
0.74 +0.1 0.79 perf-profile.children.cycles-pp.free_unref_folios
2.44 +0.1 2.50 perf-profile.children.cycles-pp.native_irq_return_iret
0.28 +0.1 0.35 ? 2% perf-profile.children.cycles-pp.free_unref_page_prepare
2.14 +0.1 2.21 perf-profile.children.cycles-pp.shmem_fault
0.60 ? 2% +0.1 0.69 ? 2% perf-profile.children.cycles-pp.folio_remove_rmap_ptes
2.27 +0.1 2.36 perf-profile.children.cycles-pp.__do_fault
0.80 ? 2% +0.1 0.92 ? 2% perf-profile.children.cycles-pp.__mod_memcg_lruvec_state
1.34 ? 5% +0.1 1.48 ? 4% perf-profile.children.cycles-pp.__lruvec_stat_mod_folio
2.87 ? 2% +0.2 3.06 ? 5% perf-profile.children.cycles-pp.folio_batch_move_lru
3.07 +0.2 3.28 ? 4% perf-profile.children.cycles-pp.folio_add_lru_vma
1.32 ? 4% +0.3 1.62 ? 11% perf-profile.children.cycles-pp.native_queued_spin_lock_slowpath
1.50 ? 4% +0.3 1.80 ? 10% perf-profile.children.cycles-pp._raw_spin_lock_irqsave
1.48 ? 4% +0.3 1.79 ? 10% perf-profile.children.cycles-pp.folio_lruvec_lock_irqsave
9.88 +0.3 10.20 perf-profile.children.cycles-pp.finish_fault
0.00 +0.3 0.34 ? 2% perf-profile.children.cycles-pp.__mem_cgroup_uncharge_folios
14.84 +0.3 15.18 perf-profile.children.cycles-pp.copy_page
32.45 +0.4 32.85 perf-profile.children.cycles-pp.do_fault
33.78 +0.4 34.22 perf-profile.children.cycles-pp.__handle_mm_fault
35.16 +0.4 35.60 perf-profile.children.cycles-pp.handle_mm_fault
38.66 +0.6 39.25 perf-profile.children.cycles-pp.do_user_addr_fault
38.93 +0.6 39.52 perf-profile.children.cycles-pp.exc_page_fault
49.87 +0.7 50.59 perf-profile.children.cycles-pp.asm_exc_page_fault
62.47 +0.9 63.33 perf-profile.children.cycles-pp.testcase
1.04 ? 2% -0.8 0.19 ? 8% perf-profile.self.cycles-pp._compound_head
0.56 ? 7% -0.4 0.18 ? 8% perf-profile.self.cycles-pp.__folio_throttle_swaprate
28.90 -0.1 28.78 perf-profile.self.cycles-pp.intel_idle_ibrs
0.27 ? 2% -0.0 0.24 ? 3% perf-profile.self.cycles-pp.folio_remove_rmap_ptes
0.18 ? 2% -0.0 0.15 ? 2% perf-profile.self.cycles-pp.free_swap_cache
0.20 ? 4% -0.0 0.18 ? 3% perf-profile.self.cycles-pp.cpuidle_enter_state
0.08 ? 5% +0.0 0.10 ? 5% perf-profile.self.cycles-pp.__do_fault
0.06 ? 6% +0.0 0.09 ? 5% perf-profile.self.cycles-pp.__tlb_remove_folio_pages_size
0.14 ? 4% +0.0 0.18 ? 4% perf-profile.self.cycles-pp.get_pfnblock_flags_mask
2.44 +0.1 2.49 perf-profile.self.cycles-pp.native_irq_return_iret
0.30 ? 4% +0.1 0.36 ? 4% perf-profile.self.cycles-pp.shmem_fault
0.15 ? 3% +0.1 0.21 ? 2% perf-profile.self.cycles-pp.zap_pte_range
0.66 ? 2% +0.1 0.76 ? 3% perf-profile.self.cycles-pp.__mod_memcg_lruvec_state
0.45 ? 3% +0.1 0.56 ? 3% perf-profile.self.cycles-pp.zap_present_ptes
1.32 ? 4% +0.3 1.62 ? 11% perf-profile.self.cycles-pp.native_queued_spin_lock_slowpath
14.74 +0.3 15.08 perf-profile.self.cycles-pp.copy_page
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.
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki