Re: [linux-next:master] [fork] 6e553c6bcb: will-it-scale.per_process_ops 94.7% improvement

[Liam R. Howlett]

* Peng Zhang <zhangpeng.00@xxxxxxxxxxxxx> [231204 02:15]:
> 
> 
> 在 2023/11/28 21:55, kernel test robot 写道:
> > 
> > 
> > Hello,
> > 
> > kernel test robot noticed a 94.7% improvement of will-it-scale.per_process_ops on:
> > 
> > 
> > commit: 6e553c6bcb7746abad29ce63e0cb7a18348e88fb ("fork: use __mt_dup() to duplicate maple tree in dup_mmap()")
> > https://git.kernel.org/cgit/linux/kernel/git/next/linux-next.git master
> > 
> > testcase: will-it-scale
> > test machine: 104 threads 2 sockets (Skylake) with 192G memory
> > parameters:
> > 
> > 	nr_task: 100%
> > 	mode: process
> > 	test: brk2
> > 	cpufreq_governor: performance
> > 
> > 
> > 
> > 
> > 
> > 
> > Details are as below:
> > -------------------------------------------------------------------------------------------------->
> > 
> > 
> > The kernel config and materials to reproduce are available at:
> > https://download.01.org/0day-ci/archive/20231128/202311282145.ff13737b-oliver.sang@xxxxxxxxx
> > 
> > =========================================================================================
> > compiler/cpufreq_governor/kconfig/mode/nr_task/rootfs/tbox_group/test/testcase:
> >    gcc-12/performance/x86_64-rhel-8.3/process/100%/debian-11.1-x86_64-20220510.cgz/lkp-skl-fpga01/brk2/will-it-scale
> > 
> > commit:
> >    ec81deb6b7 ("maple_tree: preserve the tree attributes when destroying maple tree")
> >    6e553c6bcb ("fork: use __mt_dup() to duplicate maple tree in dup_mmap()")
> > 
> > ec81deb6b769dc1b 6e553c6bcb7746abad29ce63e0c
> > ---------------- ---------------------------
> >           %stddev     %change         %stddev
> >               \          |                \
> >       24.47            -7.2       17.29        mpstat.cpu.all.soft%
> >        8.76            +7.6       16.36        mpstat.cpu.all.usr%
> >        8.80           +85.5%      16.33        vmstat.cpu.us
> >        7732           -18.3%       6318        vmstat.system.cs
> >    13467005           +94.7%   26223847        will-it-scale.104.processes
> >      129489           +94.7%     252151        will-it-scale.per_process_ops
> >    13467005           +94.7%   26223847        will-it-scale.workload
> >   1.792e+08           -13.6%  1.549e+08        numa-numastat.node0.local_node
> >   1.793e+08           -13.6%  1.549e+08        numa-numastat.node0.numa_hit
> >   1.867e+08           -12.6%  1.631e+08        numa-numastat.node1.local_node
> >   1.867e+08           -12.6%  1.632e+08        numa-numastat.node1.numa_hit
> >   1.793e+08           -13.6%  1.549e+08        numa-vmstat.node0.numa_hit
> >   1.792e+08           -13.6%  1.549e+08        numa-vmstat.node0.numa_local
> >   1.867e+08           -12.6%  1.631e+08        numa-vmstat.node1.numa_hit
> >   1.867e+08           -12.6%  1.631e+08        numa-vmstat.node1.numa_local
> >      122.70           +43.4      166.10        turbostat.PKG_%
> >       56.67 ±  2%      -6.8%      52.83 ±  3%  turbostat.PkgTmp
> >      376.36            +1.3%     381.14        turbostat.PkgWatt
> >       45.03            -5.8%      42.40        turbostat.RAMWatt
> >      121103            -2.7%     117803        proc-vmstat.nr_slab_unreclaimable
> >       68338 ±  7%     +24.8%      85281 ±  6%  proc-vmstat.numa_hint_faults
> >    3.66e+08           -13.1%  3.181e+08        proc-vmstat.numa_hit
> >   3.659e+08           -13.1%   3.18e+08        proc-vmstat.numa_local
> >   7.155e+08           -15.6%  6.038e+08        proc-vmstat.pgalloc_normal
> >     1025781            +2.6%    1052377        proc-vmstat.pgfault
> >   7.131e+08           -15.5%  6.028e+08        proc-vmstat.pgfree
> >    12305161           +10.3%   13578139        sched_debug.cfs_rq:/.avg_vruntime.avg
> >    12419603           +10.6%   13739004        sched_debug.cfs_rq:/.avg_vruntime.max
> >      122981 ±  5%     +41.3%     173809 ± 12%  sched_debug.cfs_rq:/.avg_vruntime.stddev
> >      376.33 ± 43%     -53.5%     174.89 ± 34%  sched_debug.cfs_rq:/.load_avg.avg
> >       85.75 ±  9%     -59.1%      35.08 ± 16%  sched_debug.cfs_rq:/.load_avg.min
> >    12305176           +10.3%   13578151        sched_debug.cfs_rq:/.min_vruntime.avg
> >    12419621           +10.6%   13739007        sched_debug.cfs_rq:/.min_vruntime.max
> >      122979 ±  5%     +41.3%     173810 ± 12%  sched_debug.cfs_rq:/.min_vruntime.stddev
> >       13056           -15.6%      11025        sched_debug.cpu.nr_switches.avg
> >        9228           -22.3%       7174        sched_debug.cpu.nr_switches.min
> >        0.28 ±  7%     -63.7%       0.10 ± 10%  perf-sched.sch_delay.avg.ms.schedule_timeout.rcu_gp_fqs_loop.rcu_gp_kthread.kthread
> >        0.08 ±  5%     -27.0%       0.06 ±  8%  perf-sched.sch_delay.avg.ms.smpboot_thread_fn.kthread.ret_from_fork.ret_from_fork_asm
> >        3.56 ± 45%     -79.6%       0.73 ±199%  perf-sched.sch_delay.max.ms.__cond_resched.smpboot_thread_fn.kthread.ret_from_fork.ret_from_fork_asm
> >       10.02 ± 62%     -70.0%       3.01 ±  2%  perf-sched.sch_delay.max.ms.schedule_timeout.rcu_gp_fqs_loop.rcu_gp_kthread.kthread
> >       11.16 ± 29%     -48.2%       5.78 ± 45%  perf-sched.sch_delay.max.ms.wait_for_partner.fifo_open.do_dentry_open.do_open
> >       41.80           +25.8%      52.58        perf-sched.total_wait_and_delay.average.ms
> >       37576 ±  2%     -21.0%      29680        perf-sched.total_wait_and_delay.count.ms
> >       41.73           +25.8%      52.50        perf-sched.total_wait_time.average.ms
> >        4.50 ±  2%     -16.1%       3.78 ±  6%  perf-sched.wait_and_delay.avg.ms.__cond_resched.down_write.__split_vma.do_vmi_align_munmap.__do_sys_brk
> >        4.61 ±  6%     -18.0%       3.78 ±  2%  perf-sched.wait_and_delay.avg.ms.__cond_resched.down_write.do_brk_flags.__do_sys_brk.do_syscall_64
> >        4.26 ±  7%     -11.7%       3.77        perf-sched.wait_and_delay.avg.ms.__cond_resched.kmem_cache_alloc.mas_alloc_nodes.mas_preallocate.__split_vma
> >       32.16           +53.0%      49.20        perf-sched.wait_and_delay.avg.ms.__cond_resched.smpboot_thread_fn.kthread.ret_from_fork.ret_from_fork_asm
> >       63.20           +35.4%      85.56 ±  2%  perf-sched.wait_and_delay.avg.ms.smpboot_thread_fn.kthread.ret_from_fork.ret_from_fork_asm
> >      735.00 ±  2%     -68.0%     234.83 ±  4%  perf-sched.wait_and_delay.count.__cond_resched.down_write.__split_vma.do_vmi_align_munmap.__do_sys_brk
> >        3090 ±  6%     -93.8%     190.50 ±  4%  perf-sched.wait_and_delay.count.__cond_resched.down_write.do_brk_flags.__do_sys_brk.do_syscall_64
> >      975.67 ±  3%     +61.5%       1576        perf-sched.wait_and_delay.count.__cond_resched.down_write_killable.__do_sys_brk.do_syscall_64.entry_SYSCALL_64_after_hwframe
> >      298.17 ±  4%     +54.6%     461.00 ±  4%  perf-sched.wait_and_delay.count.__cond_resched.kmem_cache_alloc.mas_alloc_nodes.mas_preallocate.__split_vma
> >      524.33 ±  4%     +56.2%     819.17 ±  2%  perf-sched.wait_and_delay.count.__cond_resched.kmem_cache_alloc.mas_alloc_nodes.mas_preallocate.do_brk_flags
> >      402.83 ±  2%     +54.2%     621.17 ±  3%  perf-sched.wait_and_delay.count.__cond_resched.kmem_cache_alloc.vm_area_dup.__split_vma.do_vmi_align_munmap
> >      265.50 ±  4%     +28.7%     341.67 ±  2%  perf-sched.wait_and_delay.count.__cond_resched.remove_vma.do_vmi_align_munmap.__do_sys_brk.do_syscall_64
> >      661.33 ±  7%     -27.6%     478.83 ± 10%  perf-sched.wait_and_delay.count.__cond_resched.smpboot_thread_fn.kthread.ret_from_fork.ret_from_fork_asm
> >       66.33 ±141%    +310.3%     272.17 ±  7%  perf-sched.wait_and_delay.count.__cond_resched.zap_pmd_range.isra.0.unmap_page_range
> >        1517 ±  4%     +72.0%       2609        perf-sched.wait_and_delay.count.exit_to_user_mode_loop.exit_to_user_mode_prepare.irqentry_exit_to_user_mode.asm_sysvec_apic_timer_interrupt
> >        6322 ±  3%     -41.1%       3725 ±  4%  perf-sched.wait_and_delay.count.exit_to_user_mode_loop.exit_to_user_mode_prepare.syscall_exit_to_user_mode.do_syscall_64
> >       14299 ±  2%     -25.3%      10677        perf-sched.wait_and_delay.count.smpboot_thread_fn.kthread.ret_from_fork.ret_from_fork_asm
> >       19.65 ± 25%     -38.8%      12.03 ±  7%  perf-sched.wait_and_delay.max.ms.__cond_resched.down_write.__split_vma.do_vmi_align_munmap.__do_sys_brk
> >       20.07 ± 27%     -42.3%      11.57 ±  8%  perf-sched.wait_and_delay.max.ms.__cond_resched.down_write.do_brk_flags.__do_sys_brk.do_syscall_64
> >       18.62 ± 28%     -32.5%      12.56 ± 12%  perf-sched.wait_and_delay.max.ms.__cond_resched.kmem_cache_alloc.vm_area_dup.__split_vma.do_vmi_align_munmap
> >       71.03 ± 10%     +44.2%     102.39 ± 36%  perf-sched.wait_and_delay.max.ms.__cond_resched.smpboot_thread_fn.kthread.ret_from_fork.ret_from_fork_asm
> >       23.72 ± 26%     -33.5%      15.79 ± 12%  perf-sched.wait_and_delay.max.ms.__cond_resched.unmap_vmas.unmap_region.constprop.0
> >       14.70 ± 21%     -33.3%       9.80 ±  2%  perf-sched.wait_and_delay.max.ms.schedule_timeout.rcu_gp_fqs_loop.rcu_gp_kthread.kthread
> >        4.50 ±  2%     -16.1%       3.78 ±  6%  perf-sched.wait_time.avg.ms.__cond_resched.down_write.__split_vma.do_vmi_align_munmap.__do_sys_brk
> >        4.61 ±  6%     -18.0%       3.78 ±  2%  perf-sched.wait_time.avg.ms.__cond_resched.down_write.do_brk_flags.__do_sys_brk.do_syscall_64
> >        4.26 ±  7%     -11.7%       3.77        perf-sched.wait_time.avg.ms.__cond_resched.kmem_cache_alloc.mas_alloc_nodes.mas_preallocate.__split_vma
> >        4.26 ±  9%    -100.0%       0.00        perf-sched.wait_time.avg.ms.__cond_resched.kmem_cache_alloc_bulk.mas_alloc_nodes.mas_preallocate.__split_vma
> >        4.61 ±  3%    -100.0%       0.00        perf-sched.wait_time.avg.ms.__cond_resched.kmem_cache_alloc_bulk.mas_alloc_nodes.mas_preallocate.do_brk_flags
> >       32.13           +53.1%      49.19        perf-sched.wait_time.avg.ms.__cond_resched.smpboot_thread_fn.kthread.ret_from_fork.ret_from_fork_asm
> >       63.12           +35.5%      85.50 ±  2%  perf-sched.wait_time.avg.ms.smpboot_thread_fn.kthread.ret_from_fork.ret_from_fork_asm
> >       19.65 ± 25%     -38.8%      12.03 ±  7%  perf-sched.wait_time.max.ms.__cond_resched.down_write.__split_vma.do_vmi_align_munmap.__do_sys_brk
> >       20.07 ± 27%     -42.3%      11.57 ±  8%  perf-sched.wait_time.max.ms.__cond_resched.down_write.do_brk_flags.__do_sys_brk.do_syscall_64
> >       18.62 ± 28%     -32.5%      12.56 ± 12%  perf-sched.wait_time.max.ms.__cond_resched.kmem_cache_alloc.vm_area_dup.__split_vma.do_vmi_align_munmap
> >       16.69 ± 36%    -100.0%       0.00        perf-sched.wait_time.max.ms.__cond_resched.kmem_cache_alloc_bulk.mas_alloc_nodes.mas_preallocate.__split_vma
> >       16.03 ± 17%    -100.0%       0.00        perf-sched.wait_time.max.ms.__cond_resched.kmem_cache_alloc_bulk.mas_alloc_nodes.mas_preallocate.do_brk_flags
> >       70.43 ± 10%     +45.4%     102.38 ± 36%  perf-sched.wait_time.max.ms.__cond_resched.smpboot_thread_fn.kthread.ret_from_fork.ret_from_fork_asm
> >       23.72 ± 26%     -33.5%      15.79 ± 12%  perf-sched.wait_time.max.ms.__cond_resched.unmap_vmas.unmap_region.constprop.0
> >       17.57 ± 34%     -29.3%      12.43 ±  7%  perf-sched.wait_time.max.ms.__cond_resched.zap_pmd_range.isra.0.unmap_page_range
> >        8.69 ±  9%     -20.8%       6.88        perf-sched.wait_time.max.ms.schedule_timeout.rcu_gp_fqs_loop.rcu_gp_kthread.kthread
> >        2.62           -12.2%       2.30        perf-stat.i.MPKI
> >   2.471e+10            +5.8%  2.614e+10        perf-stat.i.branch-instructions
> >        0.82            +0.1        0.90 ±  3%  perf-stat.i.branch-miss-rate%
> >   2.032e+08           +14.1%  2.319e+08        perf-stat.i.branch-misses
> >       61.39            +1.3       62.74        perf-stat.i.cache-miss-rate%
> >   3.567e+08           -14.2%  3.061e+08        perf-stat.i.cache-misses
> >    5.81e+08           -16.2%  4.871e+08        perf-stat.i.cache-references
> >        7695           -18.5%       6271        perf-stat.i.context-switches
> >        2.04            +2.6%       2.09        perf-stat.i.cpi
> >      249.60           -26.8%     182.76        perf-stat.i.cpu-migrations
> >      780.53           +16.7%     911.09        perf-stat.i.cycles-between-cache-misses
> >        0.05            +0.0        0.08 ±  4%  perf-stat.i.dTLB-load-miss-rate%
> >    15775916           +75.7%   27724344        perf-stat.i.dTLB-load-misses
> >    3.43e+10            +5.8%   3.63e+10        perf-stat.i.dTLB-loads
> >      459336 ±  4%     -18.1%     376380 ± 10%  perf-stat.i.dTLB-store-misses
> >   2.189e+10            -2.7%   2.13e+10        perf-stat.i.dTLB-stores
> >    31159069           +37.1%   42716400        perf-stat.i.iTLB-load-misses
> >      456253 ±  4%     +83.6%     837671 ± 31%  perf-stat.i.iTLB-loads
> >   1.364e+11            -1.9%  1.337e+11        perf-stat.i.instructions
> >        4379           -28.5%       3131        perf-stat.i.instructions-per-iTLB-miss
> >        0.49            -2.1%       0.48        perf-stat.i.ipc
> >        1327           -23.4%       1017        perf-stat.i.metric.K/sec
> >      783.30            +3.4%     809.81        perf-stat.i.metric.M/sec
> >        2733            +2.7%       2807        perf-stat.i.minor-faults
> >     4668970           -23.7%    3562283 ± 12%  perf-stat.i.node-load-misses
> >    44341556           -21.0%   35024109 ±  2%  perf-stat.i.node-loads
> >        2.44            -0.5        1.91 ±  9%  perf-stat.i.node-store-miss-rate%
> >     2146790           -44.9%    1183052 ±  2%  perf-stat.i.node-store-misses
> >    85964398           -24.9%   64569512 ±  2%  perf-stat.i.node-stores
> >        2733            +2.7%       2807        perf-stat.i.page-faults
> >        2.62           -12.5%       2.29        perf-stat.overall.MPKI
> >        0.82            +0.1        0.89        perf-stat.overall.branch-miss-rate%
> >       61.39            +1.4       62.83        perf-stat.overall.cache-miss-rate%
> >        2.04            +2.1%       2.08        perf-stat.overall.cpi
> >      780.25           +16.6%     910.13        perf-stat.overall.cycles-between-cache-misses
> >        0.05            +0.0        0.08        perf-stat.overall.dTLB-load-miss-rate%
> >        0.00 ±  4%      -0.0        0.00 ± 10%  perf-stat.overall.dTLB-store-miss-rate%
> >        4376           -28.5%       3130        perf-stat.overall.instructions-per-iTLB-miss
> >        0.49            -2.0%       0.48        perf-stat.overall.ipc
> >        2.44            -0.6        1.80 ±  2%  perf-stat.overall.node-store-miss-rate%
> >     3051945           -49.6%    1539638        perf-stat.overall.path-length
> >   2.462e+10            +5.8%  2.605e+10        perf-stat.ps.branch-instructions
> >   2.025e+08           +14.1%  2.311e+08        perf-stat.ps.branch-misses
> >   3.555e+08           -14.2%  3.051e+08        perf-stat.ps.cache-misses
> >   5.791e+08           -16.2%  4.855e+08        perf-stat.ps.cache-references
> >        7668           -18.5%       6250        perf-stat.ps.context-switches
> >      248.68           -26.7%     182.19        perf-stat.ps.cpu-migrations
> >    15739140           +75.6%   27640979        perf-stat.ps.dTLB-load-misses
> >   3.418e+10            +5.8%  3.618e+10        perf-stat.ps.dTLB-loads
> >      458389 ±  4%     -18.1%     375627 ± 10%  perf-stat.ps.dTLB-store-misses
> >   2.181e+10            -2.7%  2.123e+10        perf-stat.ps.dTLB-stores
> >    31053977           +37.1%   42572874        perf-stat.ps.iTLB-load-misses
> >      454414 ±  4%     +83.8%     835372 ± 31%  perf-stat.ps.iTLB-loads
> >   1.359e+11            -1.9%  1.333e+11        perf-stat.ps.instructions
> >        2730            +2.6%       2802        perf-stat.ps.minor-faults
> >     4653258           -23.7%    3549888 ± 12%  perf-stat.ps.node-load-misses
> >    44203348           -21.0%   34917073 ±  2%  perf-stat.ps.node-loads
> >     2139791           -44.9%    1179122 ±  2%  perf-stat.ps.node-store-misses
> >    85677729           -24.9%   64357015 ±  2%  perf-stat.ps.node-stores
> >        2731            +2.6%       2803        perf-stat.ps.page-faults
> >    4.11e+13            -1.8%  4.038e+13        perf-stat.total.instructions
> >       33.52           -19.0       14.50        perf-profile.calltrace.cycles-pp.do_brk_flags.__do_sys_brk.do_syscall_64.entry_SYSCALL_64_after_hwframe.brk
> >       17.05           -14.1        2.93        perf-profile.calltrace.cycles-pp.mas_store_prealloc.do_brk_flags.__do_sys_brk.do_syscall_64.entry_SYSCALL_64_after_hwframe
> Hi, Liam
> 
> From the performance data shown above, we can see that there is a
> significant variation in the overhead of mas_store_prealloc() within
> the do_brk_flags() function. I speculate that the maple tree duplicated
> through __mt_dup() has a different structure compared to the maple tree
> duplicated through bulk store. In this test case, the maple tree duplicated
> through __mt_dup() triggers a fast path for maple tree store operations,
> thus avoiding memory allocation. It also avoids the slow path for maple
> tree store operations.
> 

Hi Peng,

Thanks for the analysis.  I would expect such a difference to be
equalised the first loop - the maple tree would be expanded to make
space and then contracted, but each new loop should cause a node store.

If your speculation is correct, then we were causing a contraction
followed by an expansion; effectively 'jittering' after a fork.  I will
add this to the work items to investigate and see if we are not hitting
this in other scenarios.

This speculation looks correct considering we see a reduction of
spanning stores and associated allocations/cleanup.  I'm guessing we
filled the nodes too much and the brk VMA was the last entry in a node,
and so we end up rebalancing to relocate that entry and constantly hit
our worst case.  It's probably never fixed to avoid causing a larger
rebalance event.  Hopefully this only shows up in the old fork method.

There is also a lot of down_write and __split_vma delta, which is odd.
Perhaps the __split_vma could be explained by the jittering during a
split.  Maybe the down_write is due to the excessive time in allocations
and thus potentially triggering a timeout that ends up causing a
rescheduling.

Either way, this is a welcomed result.  Thanks for having a look at
these numbers, and for the patches that caused them.

Apologies for leaving all the benchmarking data in the reply, but
perhaps there is more to add that I've missed in this data.

Regards,
Liam


> >       11.58           -11.6        0.00        perf-profile.calltrace.cycles-pp.___slab_alloc.__kmem_cache_alloc_bulk.kmem_cache_alloc_bulk.mas_alloc_nodes.mas_preallocate
> >       10.92           -10.9        0.00        perf-profile.calltrace.cycles-pp.kmem_cache_alloc_bulk.mas_alloc_nodes.mas_preallocate.do_brk_flags.__do_sys_brk
> >       10.60           -10.6        0.00        perf-profile.calltrace.cycles-pp.__kmem_cache_alloc_bulk.kmem_cache_alloc_bulk.mas_alloc_nodes.mas_preallocate.do_brk_flags
> >       10.54           -10.5        0.00        perf-profile.calltrace.cycles-pp.mas_wr_bnode.mas_store_prealloc.vma_complete.__split_vma.do_vmi_align_munmap
> >       10.26           -10.3        0.00        perf-profile.calltrace.cycles-pp.mas_wr_spanning_store.mas_store_prealloc.do_brk_flags.__do_sys_brk.do_syscall_64
> >       11.70            -9.9        1.78        perf-profile.calltrace.cycles-pp.mas_alloc_nodes.mas_preallocate.do_brk_flags.__do_sys_brk.do_syscall_64
> >       12.19            -9.2        2.97        perf-profile.calltrace.cycles-pp.mas_preallocate.do_brk_flags.__do_sys_brk.do_syscall_64.entry_SYSCALL_64_after_hwframe
> >       69.63            -9.1       60.49        perf-profile.calltrace.cycles-pp.__do_sys_brk.do_syscall_64.entry_SYSCALL_64_after_hwframe.brk
> >        8.93            -8.9        0.00        perf-profile.calltrace.cycles-pp.mas_split.mas_wr_bnode.mas_store_prealloc.vma_complete.__split_vma
> >        7.42            -7.4        0.00        perf-profile.calltrace.cycles-pp.get_partial_node.___slab_alloc.__kmem_cache_alloc_bulk.kmem_cache_alloc_bulk.mas_alloc_nodes
> >        7.32            -7.3        0.00        perf-profile.calltrace.cycles-pp.mas_spanning_rebalance.mas_wr_spanning_store.mas_store_prealloc.do_brk_flags.__do_sys_brk
> >       12.16            -6.0        6.13        perf-profile.calltrace.cycles-pp.mas_store_prealloc.vma_complete.__split_vma.do_vmi_align_munmap.__do_sys_brk
> >       12.28            -5.9        6.43        perf-profile.calltrace.cycles-pp.vma_complete.__split_vma.do_vmi_align_munmap.__do_sys_brk.do_syscall_64
> >        6.38            -5.6        0.73 ±  2%  perf-profile.calltrace.cycles-pp.mas_destroy.mas_store_prealloc.do_brk_flags.__do_sys_brk.do_syscall_64
> >       15.46            -5.0       10.49 ±  4%  perf-profile.calltrace.cycles-pp.smpboot_thread_fn.kthread.ret_from_fork.ret_from_fork_asm
> >       15.44            -5.0       10.48 ±  4%  perf-profile.calltrace.cycles-pp.__do_softirq.run_ksoftirqd.smpboot_thread_fn.kthread.ret_from_fork
> >       15.44            -5.0       10.48 ±  4%  perf-profile.calltrace.cycles-pp.run_ksoftirqd.smpboot_thread_fn.kthread.ret_from_fork.ret_from_fork_asm
> >       15.47            -5.0       10.50 ±  4%  perf-profile.calltrace.cycles-pp.ret_from_fork_asm
> >       15.47            -5.0       10.50 ±  4%  perf-profile.calltrace.cycles-pp.ret_from_fork.ret_from_fork_asm
> >       15.47            -5.0       10.50 ±  4%  perf-profile.calltrace.cycles-pp.kthread.ret_from_fork.ret_from_fork_asm
> >       15.44            -5.0       10.48 ±  4%  perf-profile.calltrace.cycles-pp.rcu_core.__do_softirq.run_ksoftirqd.smpboot_thread_fn.kthread
> >       15.42            -5.0       10.46 ±  4%  perf-profile.calltrace.cycles-pp.rcu_do_batch.rcu_core.__do_softirq.run_ksoftirqd.smpboot_thread_fn
> >       74.20            -4.5       69.66        perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.brk
> >       20.44            -3.4       17.09        perf-profile.calltrace.cycles-pp.__split_vma.do_vmi_align_munmap.__do_sys_brk.do_syscall_64.entry_SYSCALL_64_after_hwframe
> >        3.84            -2.5        1.34 ±  4%  perf-profile.calltrace.cycles-pp.__slab_free.rcu_do_batch.rcu_core.__do_softirq.run_ksoftirqd
> >       76.40            -2.4       74.03        perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_after_hwframe.brk
> >        4.90 ±  2%      -2.0        2.86 ±  4%  perf-profile.calltrace.cycles-pp.rcu_core.__do_softirq.irq_exit_rcu.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt
> >        4.89 ±  2%      -2.0        2.84 ±  4%  perf-profile.calltrace.cycles-pp.rcu_do_batch.rcu_core.__do_softirq.irq_exit_rcu.sysvec_apic_timer_interrupt
> >        2.88            -1.9        1.03        perf-profile.calltrace.cycles-pp.mas_alloc_nodes.mas_preallocate.__split_vma.do_vmi_align_munmap.__do_sys_brk
> >        3.53            -1.3        2.22        perf-profile.calltrace.cycles-pp.mas_preallocate.__split_vma.do_vmi_align_munmap.__do_sys_brk.do_syscall_64
> >        0.95 ±  3%      -0.2        0.71 ±  4%  perf-profile.calltrace.cycles-pp.rcu_cblist_dequeue.rcu_do_batch.rcu_core.__do_softirq.run_ksoftirqd
> >        0.71            -0.1        0.62 ±  2%  perf-profile.calltrace.cycles-pp.kmem_cache_free.rcu_do_batch.rcu_core.__do_softirq.run_ksoftirqd
> >        0.74 ±  3%      +0.2        0.94 ±  4%  perf-profile.calltrace.cycles-pp.__unfreeze_partials.rcu_do_batch.rcu_core.__do_softirq.irq_exit_rcu
> >        0.67            +0.3        0.97 ±  3%  perf-profile.calltrace.cycles-pp.vm_area_free_rcu_cb.rcu_do_batch.rcu_core.__do_softirq.run_ksoftirqd
> >        0.56 ±  2%      +0.4        0.97 ±  5%  perf-profile.calltrace.cycles-pp.irq_exit_rcu.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt.syscall_enter_from_user_mode.do_syscall_64
> >        0.56 ±  2%      +0.4        0.97 ±  5%  perf-profile.calltrace.cycles-pp.__do_softirq.irq_exit_rcu.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt.syscall_enter_from_user_mode
> >        0.58 ±  2%      +0.4        1.01 ±  5%  perf-profile.calltrace.cycles-pp.asm_sysvec_apic_timer_interrupt.syscall_enter_from_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe.brk
> >        0.58 ±  2%      +0.4        1.01 ±  5%  perf-profile.calltrace.cycles-pp.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt.syscall_enter_from_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe
> >        0.00            +0.5        0.51        perf-profile.calltrace.cycles-pp.mas_pop_node.mas_wr_node_store.mas_wr_store_entry.mas_store_gfp.do_vmi_align_munmap
> >        0.00            +0.5        0.52 ±  2%  perf-profile.calltrace.cycles-pp.__pte_offset_map_lock.zap_pte_range.zap_pmd_range.unmap_page_range.unmap_vmas
> >        0.00            +0.5        0.53 ±  2%  perf-profile.calltrace.cycles-pp.mas_pop_node.mas_wr_node_store.mas_wr_store_entry.mas_store_prealloc.vma_complete
> >        0.00            +0.5        0.54        perf-profile.calltrace.cycles-pp.mas_update_gap.mas_wr_node_store.mas_wr_store_entry.mas_store_prealloc.vma_complete
> >        0.43 ± 44%      +0.5        0.97        perf-profile.calltrace.cycles-pp.zap_pte_range.zap_pmd_range.unmap_page_range.unmap_vmas.unmap_region
> >        0.00            +0.5        0.55        perf-profile.calltrace.cycles-pp.arch_get_unmapped_area_topdown.get_unmapped_area.check_brk_limits.__do_sys_brk.do_syscall_64
> >        0.63 ±  2%      +0.6        1.18        perf-profile.calltrace.cycles-pp.get_unmapped_area.check_brk_limits.__do_sys_brk.do_syscall_64.entry_SYSCALL_64_after_hwframe
> >        0.00            +0.6        0.56 ±  3%  perf-profile.calltrace.cycles-pp.kmem_cache_free.mas_destroy.mas_store_prealloc.do_brk_flags.__do_sys_brk
> >        6.03 ±  2%      +0.6        6.62 ±  4%  perf-profile.calltrace.cycles-pp.__unfreeze_partials.rcu_do_batch.rcu_core.__do_softirq.run_ksoftirqd
> >        0.71 ±  2%      +0.6        1.31 ±  4%  perf-profile.calltrace.cycles-pp.syscall_enter_from_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe.brk
> >        0.00            +0.6        0.63 ±  2%  perf-profile.calltrace.cycles-pp.mas_update_gap.mas_wr_slot_store.mas_wr_store_entry.mas_store_prealloc.do_brk_flags
> >        0.73 ±  2%      +0.6        1.36        perf-profile.calltrace.cycles-pp.check_brk_limits.__do_sys_brk.do_syscall_64.entry_SYSCALL_64_after_hwframe.brk
> >        0.00            +0.7        0.68        perf-profile.calltrace.cycles-pp.obj_cgroup_charge.kmem_cache_alloc.vm_area_dup.__split_vma.do_vmi_align_munmap
> >        0.00            +0.7        0.69        perf-profile.calltrace.cycles-pp.___slab_alloc.kmem_cache_alloc.mas_alloc_nodes.mas_preallocate.do_brk_flags
> >        0.74 ±  2%      +0.7        1.46        perf-profile.calltrace.cycles-pp.zap_pmd_range.unmap_page_range.unmap_vmas.unmap_region.do_vmi_align_munmap
> >        0.00            +0.7        0.72        perf-profile.calltrace.cycles-pp.___slab_alloc.kmem_cache_alloc.mas_alloc_nodes.mas_wr_node_store.mas_wr_store_entry
> >        0.00            +0.7        0.72        perf-profile.calltrace.cycles-pp.kmem_cache_alloc.mas_alloc_nodes.mas_preallocate.__split_vma.do_vmi_align_munmap
> >        0.00            +0.7        0.74        perf-profile.calltrace.cycles-pp.allocate_slab.___slab_alloc.kmem_cache_alloc.vm_area_dup.__split_vma
> >        0.72            +0.7        1.46        perf-profile.calltrace.cycles-pp.memcg_slab_post_alloc_hook.kmem_cache_alloc.vm_area_dup.__split_vma.do_vmi_align_munmap
> >        0.00            +0.8        0.76        perf-profile.calltrace.cycles-pp.userfaultfd_unmap_complete.__do_sys_brk.do_syscall_64.entry_SYSCALL_64_after_hwframe.brk
> >        0.00            +0.8        0.78        perf-profile.calltrace.cycles-pp.down_write_killable.__do_sys_brk.do_syscall_64.entry_SYSCALL_64_after_hwframe.brk
> >        0.00            +0.8        0.80 ±  2%  perf-profile.calltrace.cycles-pp.free_pgtables.unmap_region.do_vmi_align_munmap.__do_sys_brk.do_syscall_64
> >        1.00 ±  5%      +0.9        1.89 ±  4%  perf-profile.calltrace.cycles-pp.__do_softirq.irq_exit_rcu.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt.brk
> >        1.00 ±  5%      +0.9        1.89 ±  4%  perf-profile.calltrace.cycles-pp.irq_exit_rcu.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt.brk
> >        0.00            +0.9        0.89        perf-profile.calltrace.cycles-pp.mas_wr_slot_store.mas_wr_store_entry.mas_store_prealloc.do_brk_flags.__do_sys_brk
> >        1.03 ±  5%      +0.9        1.94 ±  4%  perf-profile.calltrace.cycles-pp.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt.brk
> >        0.56            +0.9        1.48        perf-profile.calltrace.cycles-pp.kmem_cache_alloc.mas_alloc_nodes.mas_preallocate.do_brk_flags.__do_sys_brk
> >        1.04 ±  4%      +0.9        1.96 ±  4%  perf-profile.calltrace.cycles-pp.asm_sysvec_apic_timer_interrupt.brk
> >        0.00            +1.0        0.98        perf-profile.calltrace.cycles-pp.___slab_alloc.kmem_cache_alloc.vm_area_dup.__split_vma.do_vmi_align_munmap
> >        0.25 ±100%      +1.0        1.24 ±  2%  perf-profile.calltrace.cycles-pp.mtree_range_walk.mas_walk.mas_find.__do_sys_brk.do_syscall_64
> >        0.75            +1.0        1.74        perf-profile.calltrace.cycles-pp.mas_walk.mas_find.__do_sys_brk.do_syscall_64.entry_SYSCALL_64_after_hwframe
> >        0.55            +1.0        1.57 ±  2%  perf-profile.calltrace.cycles-pp.__call_rcu_common.do_vmi_align_munmap.__do_sys_brk.do_syscall_64.entry_SYSCALL_64_after_hwframe
> >        1.13 ±  2%      +1.1        2.18        perf-profile.calltrace.cycles-pp.perf_event_mmap_output.perf_iterate_sb.perf_event_mmap_event.perf_event_mmap.do_brk_flags
> >        0.95            +1.1        2.03        perf-profile.calltrace.cycles-pp.mas_find.__do_sys_brk.do_syscall_64.entry_SYSCALL_64_after_hwframe.brk
> >        1.30 ±  2%      +1.2        2.49        perf-profile.calltrace.cycles-pp.unmap_page_range.unmap_vmas.unmap_region.do_vmi_align_munmap.__do_sys_brk
> >        0.00            +1.3        1.28        perf-profile.calltrace.cycles-pp.kmem_cache_alloc.mas_alloc_nodes.mas_wr_node_store.mas_wr_store_entry.mas_store_gfp
> >        1.50            +1.4        2.90        perf-profile.calltrace.cycles-pp.__entry_text_start.brk
> >        1.62            +1.5        3.08        perf-profile.calltrace.cycles-pp.perf_iterate_sb.perf_event_mmap_event.perf_event_mmap.do_brk_flags.__do_sys_brk
> >        1.78            +1.5        3.28        perf-profile.calltrace.cycles-pp.unmap_vmas.unmap_region.do_vmi_align_munmap.__do_sys_brk.do_syscall_64
> >        0.00            +1.5        1.52 ±  2%  perf-profile.calltrace.cycles-pp.__call_rcu_common.mas_wr_node_store.mas_wr_store_entry.mas_store_gfp.do_vmi_align_munmap
> >        0.00            +1.5        1.52        perf-profile.calltrace.cycles-pp.mas_alloc_nodes.mas_wr_node_store.mas_wr_store_entry.mas_store_gfp.do_vmi_align_munmap
> >        0.00            +1.5        1.53        perf-profile.calltrace.cycles-pp.__call_rcu_common.mas_wr_node_store.mas_wr_store_entry.mas_store_prealloc.vma_complete
> >        0.00            +1.7        1.74        perf-profile.calltrace.cycles-pp.mas_wr_store_entry.mas_store_prealloc.do_brk_flags.__do_sys_brk.do_syscall_64
> >        0.00            +1.9        1.90        perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_safe_stack.brk
> >        2.35            +2.2        4.51        perf-profile.calltrace.cycles-pp.perf_event_mmap_event.perf_event_mmap.do_brk_flags.__do_sys_brk.do_syscall_64
> >        2.30            +2.4        4.65        perf-profile.calltrace.cycles-pp.kmem_cache_alloc.vm_area_dup.__split_vma.do_vmi_align_munmap.__do_sys_brk
> >        2.62            +2.5        5.10        perf-profile.calltrace.cycles-pp.perf_event_mmap.do_brk_flags.__do_sys_brk.do_syscall_64.entry_SYSCALL_64_after_hwframe
> >        3.12            +2.6        5.70        perf-profile.calltrace.cycles-pp.unmap_region.do_vmi_align_munmap.__do_sys_brk.do_syscall_64.entry_SYSCALL_64_after_hwframe
> >        2.74            +2.8        5.53        perf-profile.calltrace.cycles-pp.vm_area_dup.__split_vma.do_vmi_align_munmap.__do_sys_brk.do_syscall_64
> >        4.28 ±  3%      +2.9        7.16 ±  4%  perf-profile.calltrace.cycles-pp._raw_spin_lock_irqsave.free_pcppages_bulk.free_unref_page_commit.free_unref_page.__unfreeze_partials
> >        4.38 ±  3%      +2.9        7.31 ±  4%  perf-profile.calltrace.cycles-pp.free_pcppages_bulk.free_unref_page_commit.free_unref_page.__unfreeze_partials.rcu_do_batch
> >        4.41 ±  3%      +2.9        7.35 ±  4%  perf-profile.calltrace.cycles-pp.free_unref_page_commit.free_unref_page.__unfreeze_partials.rcu_do_batch.rcu_core
> >        4.43 ±  3%      +3.0        7.38 ±  4%  perf-profile.calltrace.cycles-pp.free_unref_page.__unfreeze_partials.rcu_do_batch.rcu_core.__do_softirq
> >        3.44            +3.5        6.99        perf-profile.calltrace.cycles-pp.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe.brk
> >        3.91            +3.8        7.74        perf-profile.calltrace.cycles-pp.syscall_return_via_sysret.brk
> >        4.63            +4.3        8.91        perf-profile.calltrace.cycles-pp.mas_store_gfp.do_vmi_align_munmap.__do_sys_brk.do_syscall_64.entry_SYSCALL_64_after_hwframe
> >       83.68            +4.7       88.36        perf-profile.calltrace.cycles-pp.brk
> >        0.00            +4.8        4.76        perf-profile.calltrace.cycles-pp.mas_wr_node_store.mas_wr_store_entry.mas_store_prealloc.vma_complete.__split_vma
> >        0.00            +5.5        5.54        perf-profile.calltrace.cycles-pp.mas_wr_store_entry.mas_store_prealloc.vma_complete.__split_vma.do_vmi_align_munmap
> >       32.01            +6.3       38.28        perf-profile.calltrace.cycles-pp.do_vmi_align_munmap.__do_sys_brk.do_syscall_64.entry_SYSCALL_64_after_hwframe.brk
> >        0.00            +6.4        6.35        perf-profile.calltrace.cycles-pp.mas_wr_node_store.mas_wr_store_entry.mas_store_gfp.do_vmi_align_munmap.__do_sys_brk
> >        0.82            +6.7        7.53        perf-profile.calltrace.cycles-pp.mas_wr_store_entry.mas_store_gfp.do_vmi_align_munmap.__do_sys_brk.do_syscall_64
> >       29.25           -20.1        9.14        perf-profile.children.cycles-pp.mas_store_prealloc
> >       33.60           -18.9       14.67        perf-profile.children.cycles-pp.do_brk_flags
> >       27.84           -17.3       10.50 ±  2%  perf-profile.children.cycles-pp._raw_spin_lock_irqsave
> >       27.53           -17.1       10.38 ±  2%  perf-profile.children.cycles-pp.native_queued_spin_lock_slowpath
> >       13.75           -13.7        0.00        perf-profile.children.cycles-pp.mas_wr_bnode
> >       13.22           -13.2        0.00        perf-profile.children.cycles-pp.kmem_cache_alloc_bulk
> >       12.52           -12.5        0.00        perf-profile.children.cycles-pp.__kmem_cache_alloc_bulk
> >       15.06           -10.7        4.36        perf-profile.children.cycles-pp.mas_alloc_nodes
> >       15.78           -10.5        5.29        perf-profile.children.cycles-pp.mas_preallocate
> >       10.28           -10.3        0.00        perf-profile.children.cycles-pp.mas_wr_spanning_store
> >       12.34            -9.9        2.40        perf-profile.children.cycles-pp.___slab_alloc
> >        9.30            -9.2        0.09 ±  4%  perf-profile.children.cycles-pp.get_partial_node
> >       69.68            -9.1       60.59        perf-profile.children.cycles-pp.__do_sys_brk
> >        8.99            -9.0        0.00        perf-profile.children.cycles-pp.mas_split
> >       27.56            -8.0       19.53        perf-profile.children.cycles-pp.rcu_do_batch
> >       27.59            -8.0       19.56        perf-profile.children.cycles-pp.__do_softirq
> >       27.58            -8.0       19.55        perf-profile.children.cycles-pp.rcu_core
> >        7.37            -7.4        0.00        perf-profile.children.cycles-pp.mas_spanning_rebalance
> >        7.54            -6.5        1.02        perf-profile.children.cycles-pp.mas_destroy
> >        5.92            -5.9        0.00        perf-profile.children.cycles-pp.mas_topiary_replace
> >       12.39            -5.8        6.61        perf-profile.children.cycles-pp.vma_complete
> >       15.46            -5.0       10.49 ±  4%  perf-profile.children.cycles-pp.smpboot_thread_fn
> >       15.44            -5.0       10.48 ±  4%  perf-profile.children.cycles-pp.run_ksoftirqd
> >       15.47            -5.0       10.50 ±  4%  perf-profile.children.cycles-pp.ret_from_fork_asm
> >       15.47            -5.0       10.50 ±  4%  perf-profile.children.cycles-pp.ret_from_fork
> >       15.47            -5.0       10.50 ±  4%  perf-profile.children.cycles-pp.kthread
> >        8.32            -4.6        3.78 ±  2%  perf-profile.children.cycles-pp.__slab_free
> >       74.27            -4.5       69.76        perf-profile.children.cycles-pp.do_syscall_64
> >       15.32            -4.2       11.08 ±  2%  perf-profile.children.cycles-pp.__unfreeze_partials
> >       20.53            -3.3       17.27        perf-profile.children.cycles-pp.__split_vma
> >       12.15 ±  2%      -3.1        9.09 ±  3%  perf-profile.children.cycles-pp.irq_exit_rcu
> >       12.47 ±  2%      -3.0        9.42 ±  3%  perf-profile.children.cycles-pp.sysvec_apic_timer_interrupt
> >       12.50 ±  2%      -3.0        9.46 ±  3%  perf-profile.children.cycles-pp.asm_sysvec_apic_timer_interrupt
> >       76.50            -2.3       74.18        perf-profile.children.cycles-pp.entry_SYSCALL_64_after_hwframe
> >        1.91 ±  4%      -1.8        0.16 ±  3%  perf-profile.children.cycles-pp.get_any_partial
> >        1.09            -0.9        0.14 ±  2%  perf-profile.children.cycles-pp.memcpy_orig
> >        1.72            -0.7        1.05        perf-profile.children.cycles-pp.mas_pop_node
> >        1.14            -0.2        0.98 ±  2%  perf-profile.children.cycles-pp.mas_prev_slot
> >        1.71 ±  2%      -0.1        1.57 ±  4%  perf-profile.children.cycles-pp.rcu_cblist_dequeue
> >        1.10            -0.1        0.98        perf-profile.children.cycles-pp.mas_next_slot
> >        0.32 ±  2%      -0.1        0.22 ±  3%  perf-profile.children.cycles-pp.inc_slabs_node
> >        0.60 ±  2%      -0.1        0.52        perf-profile.children.cycles-pp.rcu_segcblist_enqueue
> >        0.48 ±  2%      -0.1        0.40 ±  2%  perf-profile.children.cycles-pp.rcu_nocb_try_bypass
> >        0.47 ±  2%      -0.1        0.40        perf-profile.children.cycles-pp.alloc_pages_mpol
> >        0.38 ±  3%      -0.1        0.31 ±  2%  perf-profile.children.cycles-pp.get_page_from_freelist
> >        0.18 ±  2%      -0.1        0.11        perf-profile.children.cycles-pp.__list_del_entry_valid_or_report
> >        0.44 ±  3%      -0.1        0.37        perf-profile.children.cycles-pp.__alloc_pages
> >        0.27 ±  2%      -0.1        0.22 ±  4%  perf-profile.children.cycles-pp.mt_free_rcu
> >        0.25 ±  3%      -0.1        0.20 ±  3%  perf-profile.children.cycles-pp.rmqueue
> >        0.20 ±  6%      -0.0        0.16 ±  3%  perf-profile.children.cycles-pp.__rmqueue_pcplist
> >        0.15 ±  3%      -0.0        0.11 ±  5%  perf-profile.children.cycles-pp.discard_slab
> >        0.16 ±  4%      -0.0        0.12 ±  3%  perf-profile.children.cycles-pp.rmqueue_bulk
> >        0.08 ±  5%      -0.0        0.06        perf-profile.children.cycles-pp.__list_add_valid_or_report
> >        0.24 ±  2%      -0.0        0.22 ±  2%  perf-profile.children.cycles-pp.__free_one_page
> >        0.19 ±  2%      +0.0        0.21 ±  3%  perf-profile.children.cycles-pp.update_process_times
> >        0.44 ±  2%      +0.0        0.46 ±  4%  perf-profile.children.cycles-pp.setup_object
> >        1.00            +0.0        1.03        perf-profile.children.cycles-pp.shuffle_freelist
> >        0.20 ±  3%      +0.0        0.24 ±  4%  perf-profile.children.cycles-pp.should_failslab
> >        0.09 ±  4%      +0.0        0.14 ±  3%  perf-profile.children.cycles-pp.strlen
> >        0.06            +0.0        0.11 ±  4%  perf-profile.children.cycles-pp.__kmalloc_node
> >        0.06 ±  8%      +0.0        0.10 ±  4%  perf-profile.children.cycles-pp.__init_rwsem
> >        0.05            +0.0        0.10 ±  3%  perf-profile.children.cycles-pp.__kmem_cache_alloc_node
> >        0.00            +0.1        0.05        perf-profile.children.cycles-pp.memcg_account_kmem
> >        0.00            +0.1        0.05        perf-profile.children.cycles-pp.testcase
> >        0.06            +0.1        0.11 ±  4%  perf-profile.children.cycles-pp.memcg_alloc_slab_cgroups
> >        0.05 ±  7%      +0.1        0.10 ±  4%  perf-profile.children.cycles-pp.__pte_offset_map
> >        0.00            +0.1        0.05 ±  8%  perf-profile.children.cycles-pp.flush_tlb_batched_pending
> >        0.00            +0.1        0.06 ±  8%  perf-profile.children.cycles-pp.__free_slab
> >        0.02 ±141%      +0.1        0.08 ± 18%  perf-profile.children.cycles-pp.machine__process_mmap2_event
> >        0.07 ± 13%      +0.1        0.13 ±  5%  perf-profile.children.cycles-pp.may_expand_vm
> >        0.05 ± 50%      +0.1        0.11 ± 18%  perf-profile.children.cycles-pp.perf_session__process_user_event
> >        0.05 ± 50%      +0.1        0.11 ± 18%  perf-profile.children.cycles-pp.__ordered_events__flush
> >        0.00            +0.1        0.06 ±  6%  perf-profile.children.cycles-pp.memset_orig
> >        0.12 ±  3%      +0.1        0.18 ±  2%  perf-profile.children.cycles-pp.anon_vma_clone
> >        0.08 ±  4%      +0.1        0.14 ±  3%  perf-profile.children.cycles-pp.vma_dup_policy
> >        0.07 ±  6%      +0.1        0.14 ±  2%  perf-profile.children.cycles-pp.free_pgd_range
> >        0.03 ±100%      +0.1        0.10 ± 20%  perf-profile.children.cycles-pp.perf_session__deliver_event
> >        0.00            +0.1        0.06 ±  7%  perf-profile.children.cycles-pp.mas_prev
> >        0.11 ±  3%      +0.1        0.18 ±  2%  perf-profile.children.cycles-pp.mas_node_count_gfp
> >        0.01 ±223%      +0.1        0.08        perf-profile.children.cycles-pp.tlb_batch_pages_flush
> >        0.11 ±  6%      +0.1        0.18 ±  2%  perf-profile.children.cycles-pp.unmap_single_vma
> >        1.17            +0.1        1.24 ±  2%  perf-profile.children.cycles-pp.mtree_range_walk
> >        0.09 ±  5%      +0.1        0.17 ±  2%  perf-profile.children.cycles-pp.up_read
> >        0.02 ±141%      +0.1        0.10 ±  6%  perf-profile.children.cycles-pp.vma_prepare
> >        0.06 ±  6%      +0.1        0.14 ±  3%  perf-profile.children.cycles-pp.can_vma_merge_after
> >        0.10 ± 10%      +0.1        0.19 ±  6%  perf-profile.children.cycles-pp.kfree
> >        0.08 ±  6%      +0.1        0.16 ±  3%  perf-profile.children.cycles-pp.strnlen
> >        0.10 ±  4%      +0.1        0.20 ±  2%  perf-profile.children.cycles-pp.lru_add_drain_cpu
> >        0.18 ±  2%      +0.1        0.27        perf-profile.children.cycles-pp.tlb_gather_mmu
> >        0.09            +0.1        0.19 ±  2%  perf-profile.children.cycles-pp.refill_obj_stock
> >        0.00            +0.1        0.10 ±  4%  perf-profile.children.cycles-pp.mlock_drain_local
> >        0.10 ±  5%      +0.1        0.20 ±  2%  perf-profile.children.cycles-pp.unlink_anon_vmas
> >        0.11 ±  4%      +0.1        0.21 ±  3%  perf-profile.children.cycles-pp.khugepaged_enter_vma
> >        0.10 ±  5%      +0.1        0.20        perf-profile.children.cycles-pp.current_obj_cgroup
> >        0.10 ±  6%      +0.1        0.20 ±  2%  perf-profile.children.cycles-pp.syscall_exit_to_user_mode_prepare
> >        0.12 ±  5%      +0.1        0.23 ±  3%  perf-profile.children.cycles-pp._raw_spin_lock
> >        0.14 ±  4%      +0.1        0.26        perf-profile.children.cycles-pp.cap_mmap_addr
> >        0.15 ±  6%      +0.1        0.27        perf-profile.children.cycles-pp.exit_to_user_mode_prepare
> >        0.10 ±  3%      +0.1        0.23        perf-profile.children.cycles-pp.downgrade_write
> >        0.14 ±  3%      +0.1        0.28        perf-profile.children.cycles-pp.remove_vma
> >        2.56            +0.1        2.70        perf-profile.children.cycles-pp.kmem_cache_free
> >        0.00            +0.1        0.14 ±  2%  perf-profile.children.cycles-pp.mas_is_err
> >        0.14 ±  3%      +0.1        0.28        perf-profile.children.cycles-pp.lru_add_drain
> >        0.08 ±  5%      +0.1        0.22 ±  2%  perf-profile.children.cycles-pp.__vm_enough_memory
> >        0.15 ±  3%      +0.1        0.30 ±  2%  perf-profile.children.cycles-pp.__x64_sys_brk
> >        0.15 ±  3%      +0.1        0.30 ±  2%  perf-profile.children.cycles-pp.vma_adjust_trans_huge
> >        0.22 ±  2%      +0.2        0.38 ±  3%  perf-profile.children.cycles-pp.rcu_all_qs
> >        0.06            +0.2        0.24 ±  2%  perf-profile.children.cycles-pp.mas_prev_range
> >        0.16 ±  4%      +0.2        0.35        perf-profile.children.cycles-pp.mas_wr_store_setup
> >        0.24 ±  3%      +0.2        0.42        perf-profile.children.cycles-pp.tlb_finish_mmu
> >        0.22 ±  3%      +0.2        0.41        perf-profile.children.cycles-pp.security_mmap_addr
> >        0.20 ±  3%      +0.2        0.39 ±  2%  perf-profile.children.cycles-pp.cap_vm_enough_memory
> >        0.17 ±  5%      +0.2        0.36 ±  2%  perf-profile.children.cycles-pp.percpu_counter_add_batch
> >        0.14 ±  6%      +0.2        0.34 ±  2%  perf-profile.children.cycles-pp.strscpy
> >        0.01 ±223%      +0.2        0.24        perf-profile.children.cycles-pp.mas_wr_append
> >        0.19 ±  4%      +0.2        0.43 ±  2%  perf-profile.children.cycles-pp.mas_nomem
> >        0.25            +0.2        0.50 ±  2%  perf-profile.children.cycles-pp.security_vm_enough_memory_mm
> >        0.26            +0.3        0.51 ±  2%  perf-profile.children.cycles-pp.entry_SYSRETQ_unsafe_stack
> >        0.27 ±  2%      +0.3        0.53 ±  2%  perf-profile.children.cycles-pp.__pte_offset_map_lock
> >        0.27 ±  2%      +0.3        0.56        perf-profile.children.cycles-pp.arch_get_unmapped_area_topdown
> >        0.35 ±  2%      +0.3        0.68        perf-profile.children.cycles-pp.obj_cgroup_charge
> >        0.36 ±  2%      +0.4        0.72 ±  2%  perf-profile.children.cycles-pp.mod_objcg_state
> >        0.39 ±  3%      +0.4        0.78        perf-profile.children.cycles-pp.up_write
> >        0.44 ±  2%      +0.4        0.85 ±  2%  perf-profile.children.cycles-pp.free_pgtables
> >        0.35            +0.4        0.77        perf-profile.children.cycles-pp.userfaultfd_unmap_complete
> >        0.38 ±  2%      +0.4        0.81        perf-profile.children.cycles-pp.down_write_killable
> >        0.56            +0.4        1.01 ±  2%  perf-profile.children.cycles-pp.down_write
> >        0.53 ±  2%      +0.5        1.00        perf-profile.children.cycles-pp.zap_pte_range
> >        0.31 ±  2%      +0.5        0.82        perf-profile.children.cycles-pp.mas_wr_end_piv
> >        0.56            +0.5        1.10        perf-profile.children.cycles-pp.mas_leaf_max_gap
> >        0.89            +0.6        1.45        perf-profile.children.cycles-pp.__cond_resched
> >        0.64 ±  2%      +0.6        1.20        perf-profile.children.cycles-pp.get_unmapped_area
> >        0.72 ±  2%      +0.6        1.32 ±  4%  perf-profile.children.cycles-pp.syscall_enter_from_user_mode
> >        0.48 ±  2%      +0.6        1.08        perf-profile.children.cycles-pp.entry_SYSCALL_64_safe_stack
> >        0.73 ±  2%      +0.6        1.37        perf-profile.children.cycles-pp.check_brk_limits
> >        0.77 ±  2%      +0.7        1.50        perf-profile.children.cycles-pp.zap_pmd_range
> >        0.94            +0.8        1.77        perf-profile.children.cycles-pp.memcg_slab_post_alloc_hook
> >        0.68            +0.8        1.52        perf-profile.children.cycles-pp.mas_wr_walk
> >        0.00            +0.9        0.90        perf-profile.children.cycles-pp.mas_wr_slot_store
> >        1.22            +0.9        2.17        perf-profile.children.cycles-pp.vm_area_free_rcu_cb
> >        1.14 ±  2%      +1.1        2.20        perf-profile.children.cycles-pp.perf_event_mmap_output
> >        0.91            +1.1        2.02        perf-profile.children.cycles-pp.mas_walk
> >        1.32            +1.2        2.54        perf-profile.children.cycles-pp.unmap_page_range
> >        3.44            +1.3        4.69        perf-profile.children.cycles-pp.__call_rcu_common
> >        0.33 ±  4%      +1.4        1.74        perf-profile.children.cycles-pp.mas_update_gap
> >        1.63            +1.5        3.10        perf-profile.children.cycles-pp.perf_iterate_sb
> >        1.79            +1.5        3.30        perf-profile.children.cycles-pp.unmap_vmas
> >        1.60            +1.6        3.19        perf-profile.children.cycles-pp.mas_find
> >        1.85            +1.9        3.73        perf-profile.children.cycles-pp.__entry_text_start
> >        2.39            +2.2        4.60        perf-profile.children.cycles-pp.perf_event_mmap_event
> >        2.62            +2.5        5.12        perf-profile.children.cycles-pp.perf_event_mmap
> >        3.15            +2.6        5.79        perf-profile.children.cycles-pp.unmap_region
> >        2.75            +2.8        5.55        perf-profile.children.cycles-pp.vm_area_dup
> >        3.46            +3.6        7.04        perf-profile.children.cycles-pp.syscall_exit_to_user_mode
> >        3.94            +3.9        7.80        perf-profile.children.cycles-pp.syscall_return_via_sysret
> >        4.68            +4.3        9.03        perf-profile.children.cycles-pp.mas_store_gfp
> >        3.74            +4.7        8.40        perf-profile.children.cycles-pp.kmem_cache_alloc
> >       83.90            +4.9       88.80        perf-profile.children.cycles-pp.brk
> >       32.12            +6.4       38.51        perf-profile.children.cycles-pp.do_vmi_align_munmap
> >        0.24 ±  2%     +11.0       11.20        perf-profile.children.cycles-pp.mas_wr_node_store
> >        1.38           +13.6       14.95        perf-profile.children.cycles-pp.mas_wr_store_entry
> >       27.53           -17.1       10.38 ±  2%  perf-profile.self.cycles-pp.native_queued_spin_lock_slowpath
> >        4.60            -0.8        3.76 ±  2%  perf-profile.self.cycles-pp.__slab_free
> >        0.94            -0.8        0.12 ±  4%  perf-profile.self.cycles-pp.memcpy_orig
> >        1.52            -0.6        0.94        perf-profile.self.cycles-pp.mas_pop_node
> >        0.79            -0.5        0.30        perf-profile.self.cycles-pp.___slab_alloc
> >        0.31            -0.2        0.12 ±  3%  perf-profile.self.cycles-pp._raw_spin_lock_irqsave
> >        0.28 ±  3%      -0.2        0.09 ±  5%  perf-profile.self.cycles-pp.__unfreeze_partials
> >        1.68 ±  2%      -0.1        1.55 ±  5%  perf-profile.self.cycles-pp.rcu_cblist_dequeue
> >        2.14            -0.1        2.02        perf-profile.self.cycles-pp.kmem_cache_free
> >        0.37 ±  2%      -0.1        0.28        perf-profile.self.cycles-pp.rcu_nocb_try_bypass
> >        0.30 ±  2%      -0.1        0.22 ±  3%  perf-profile.self.cycles-pp.inc_slabs_node
> >        0.52 ±  2%      -0.1        0.46 ±  2%  perf-profile.self.cycles-pp.rcu_segcblist_enqueue
> >        0.17 ±  2%      -0.1        0.10 ±  4%  perf-profile.self.cycles-pp.__list_del_entry_valid_or_report
> >        0.26            -0.0        0.21 ±  4%  perf-profile.self.cycles-pp.mt_free_rcu
> >        0.15 ±  3%      -0.0        0.11 ±  3%  perf-profile.self.cycles-pp.discard_slab
> >        0.17 ±  2%      -0.0        0.14 ±  2%  perf-profile.self.cycles-pp.get_any_partial
> >        0.08 ±  5%      -0.0        0.06        perf-profile.self.cycles-pp.__list_add_valid_or_report
> >        0.15 ±  3%      -0.0        0.13 ±  3%  perf-profile.self.cycles-pp.__free_one_page
> >        0.10 ±  3%      +0.0        0.12 ±  4%  perf-profile.self.cycles-pp.rcu_segcblist_pend_cbs
> >        0.08 ± 10%      +0.0        0.10 ±  4%  perf-profile.self.cycles-pp.should_failslab
> >        0.08            +0.0        0.12 ±  3%  perf-profile.self.cycles-pp.strlen
> >        0.06            +0.0        0.10 ±  6%  perf-profile.self.cycles-pp.vma_dup_policy
> >        0.08            +0.0        0.12 ±  3%  perf-profile.self.cycles-pp.anon_vma_clone
> >        0.06 ±  9%      +0.0        0.10 ±  6%  perf-profile.self.cycles-pp.may_expand_vm
> >        0.00            +0.1        0.05        perf-profile.self.cycles-pp.mas_prev
> >        0.03 ± 70%      +0.1        0.08 ±  5%  perf-profile.self.cycles-pp.__init_rwsem
> >        0.09 ±  8%      +0.1        0.14 ±  4%  perf-profile.self.cycles-pp.kfree
> >        0.00            +0.1        0.06 ±  6%  perf-profile.self.cycles-pp.__vm_enough_memory
> >        0.07 ±  6%      +0.1        0.13 ±  2%  perf-profile.self.cycles-pp.mas_node_count_gfp
> >        0.03 ± 70%      +0.1        0.09 ±  7%  perf-profile.self.cycles-pp.security_vm_enough_memory_mm
> >        0.15 ±  3%      +0.1        0.21 ± 11%  perf-profile.self.cycles-pp.setup_object
> >        0.00            +0.1        0.06        perf-profile.self.cycles-pp.memset_orig
> >        0.07 ±  5%      +0.1        0.13 ±  3%  perf-profile.self.cycles-pp.security_mmap_addr
> >        0.00            +0.1        0.06 ±  6%  perf-profile.self.cycles-pp.tlb_batch_pages_flush
> >        0.06 ±  7%      +0.1        0.13 ±  3%  perf-profile.self.cycles-pp.free_pgd_range
> >        0.08            +0.1        0.15 ±  2%  perf-profile.self.cycles-pp.check_brk_limits
> >        0.08 ±  6%      +0.1        0.14 ±  3%  perf-profile.self.cycles-pp.khugepaged_enter_vma
> >        0.09 ±  4%      +0.1        0.17 ±  2%  perf-profile.self.cycles-pp.unmap_single_vma
> >        0.00            +0.1        0.08 ±  6%  perf-profile.self.cycles-pp.vma_prepare
> >        0.07 ±  5%      +0.1        0.15 ±  4%  perf-profile.self.cycles-pp.up_read
> >        0.06 ±  7%      +0.1        0.14 ±  5%  perf-profile.self.cycles-pp.strnlen
> >        0.16 ±  2%      +0.1        0.24        perf-profile.self.cycles-pp.tlb_gather_mmu
> >        0.07            +0.1        0.15 ±  3%  perf-profile.self.cycles-pp.unlink_anon_vmas
> >        0.09 ±  6%      +0.1        0.17        perf-profile.self.cycles-pp.lru_add_drain_cpu
> >        0.00            +0.1        0.08        perf-profile.self.cycles-pp.mlock_drain_local
> >        0.04 ± 44%      +0.1        0.12 ±  3%  perf-profile.self.cycles-pp.can_vma_merge_after
> >        0.14 ±  2%      +0.1        0.22 ±  3%  perf-profile.self.cycles-pp.get_unmapped_area
> >        0.01 ±223%      +0.1        0.09 ±  5%  perf-profile.self.cycles-pp.__pte_offset_map
> >        0.06 ±  7%      +0.1        0.15 ±  3%  perf-profile.self.cycles-pp.remove_vma
> >        0.08 ±  4%      +0.1        0.17 ±  3%  perf-profile.self.cycles-pp.syscall_exit_to_user_mode_prepare
> >        0.09            +0.1        0.18 ±  3%  perf-profile.self.cycles-pp.refill_obj_stock
> >        0.08 ±  4%      +0.1        0.17 ±  3%  perf-profile.self.cycles-pp.current_obj_cgroup
> >        0.09 ±  4%      +0.1        0.18 ±  2%  perf-profile.self.cycles-pp.__pte_offset_map_lock
> >        0.00            +0.1        0.10 ±  4%  perf-profile.self.cycles-pp.mas_is_err
> >        0.09 ±  6%      +0.1        0.19 ±  3%  perf-profile.self.cycles-pp.mas_nomem
> >        0.11 ±  4%      +0.1        0.21 ±  4%  perf-profile.self.cycles-pp._raw_spin_lock
> >        0.24 ±  2%      +0.1        0.35        perf-profile.self.cycles-pp.mas_destroy
> >        0.12 ±  3%      +0.1        0.23 ±  2%  perf-profile.self.cycles-pp.cap_mmap_addr
> >        1.02            +0.1        1.13        perf-profile.self.cycles-pp.mtree_range_walk
> >        0.15 ±  4%      +0.1        0.26 ±  4%  perf-profile.self.cycles-pp.rcu_all_qs
> >        0.12 ±  6%      +0.1        0.22 ±  2%  perf-profile.self.cycles-pp.exit_to_user_mode_prepare
> >        0.09            +0.1        0.20 ±  2%  perf-profile.self.cycles-pp.downgrade_write
> >        0.13            +0.1        0.25 ±  2%  perf-profile.self.cycles-pp.__x64_sys_brk
> >        0.22 ±  2%      +0.1        0.34 ±  2%  perf-profile.self.cycles-pp.unmap_vmas
> >        0.13 ±  3%      +0.1        0.26 ±  2%  perf-profile.self.cycles-pp.vma_adjust_trans_huge
> >        0.12 ±  4%      +0.1        0.25 ±  2%  perf-profile.self.cycles-pp.entry_SYSCALL_64_safe_stack
> >        0.16 ±  3%      +0.1        0.30        perf-profile.self.cycles-pp.tlb_finish_mmu
> >        0.12 ±  4%      +0.1        0.26 ±  2%  perf-profile.self.cycles-pp.mas_wr_store_setup
> >        0.14 ±  4%      +0.2        0.29 ±  2%  perf-profile.self.cycles-pp.syscall_enter_from_user_mode
> >        0.14 ±  2%      +0.2        0.31        perf-profile.self.cycles-pp.free_pgtables
> >        0.12 ±  4%      +0.2        0.28        perf-profile.self.cycles-pp.strscpy
> >        0.00            +0.2        0.17 ±  4%  perf-profile.self.cycles-pp.mas_prev_range
> >        0.14 ±  4%      +0.2        0.31 ±  2%  perf-profile.self.cycles-pp.percpu_counter_add_batch
> >        0.18 ±  2%      +0.2        0.35        perf-profile.self.cycles-pp.cap_vm_enough_memory
> >        0.18 ±  3%      +0.2        0.36 ±  2%  perf-profile.self.cycles-pp.zap_pte_range
> >        0.15 ±  3%      +0.2        0.33 ±  2%  perf-profile.self.cycles-pp.zap_pmd_range
> >        0.27            +0.2        0.47        perf-profile.self.cycles-pp.unmap_region
> >        0.00            +0.2        0.20 ±  2%  perf-profile.self.cycles-pp.mas_wr_append
> >        0.00            +0.2        0.20 ±  3%  perf-profile.self.cycles-pp.mas_wr_slot_store
> >        0.18 ±  2%      +0.2        0.39        perf-profile.self.cycles-pp.vma_complete
> >        0.34 ±  5%      +0.2        0.55 ±  3%  perf-profile.self.cycles-pp.vm_area_free_rcu_cb
> >        0.22 ±  2%      +0.2        0.44 ±  2%  perf-profile.self.cycles-pp.entry_SYSRETQ_unsafe_stack
> >        0.45 ±  2%      +0.2        0.68        perf-profile.self.cycles-pp.mas_alloc_nodes
> >        0.20 ±  2%      +0.2        0.43 ±  2%  perf-profile.self.cycles-pp.down_write_killable
> >        0.20 ±  3%      +0.2        0.44        perf-profile.self.cycles-pp.perf_event_mmap
> >        0.18 ±  3%      +0.2        0.42 ±  2%  perf-profile.self.cycles-pp.do_syscall_64
> >        0.24            +0.3        0.49        perf-profile.self.cycles-pp.arch_get_unmapped_area_topdown
> >        0.27 ±  3%      +0.3        0.54        perf-profile.self.cycles-pp.obj_cgroup_charge
> >        0.36 ±  2%      +0.3        0.66 ±  3%  perf-profile.self.cycles-pp.down_write
> >        0.30 ±  2%      +0.3        0.62        perf-profile.self.cycles-pp.vm_area_dup
> >        0.32 ±  2%      +0.3        0.64 ±  2%  perf-profile.self.cycles-pp.mod_objcg_state
> >        0.35            +0.3        0.68        perf-profile.self.cycles-pp.mas_walk
> >        0.34 ±  2%      +0.3        0.68        perf-profile.self.cycles-pp.perf_event_mmap_event
> >        0.33 ±  2%      +0.4        0.68        perf-profile.self.cycles-pp.up_write
> >        0.44 ±  2%      +0.4        0.80        perf-profile.self.cycles-pp.perf_iterate_sb
> >        0.58            +0.4        0.94 ±  2%  perf-profile.self.cycles-pp.__cond_resched
> >        0.30 ±  2%      +0.4        0.68        perf-profile.self.cycles-pp.userfaultfd_unmap_complete
> >        0.51            +0.4        0.88        perf-profile.self.cycles-pp.mas_next_slot
> >        0.44 ±  2%      +0.4        0.82        perf-profile.self.cycles-pp.unmap_page_range
> >        0.58            +0.4        0.96        perf-profile.self.cycles-pp.mas_find
> >        0.26            +0.4        0.68        perf-profile.self.cycles-pp.mas_wr_end_piv
> >        0.48 ±  2%      +0.4        0.90        perf-profile.self.cycles-pp.brk
> >        0.45            +0.4        0.88        perf-profile.self.cycles-pp.mas_prev_slot
> >        0.49            +0.4        0.94        perf-profile.self.cycles-pp.mas_leaf_max_gap
> >        0.34 ±  2%      +0.4        0.78        perf-profile.self.cycles-pp.mas_store_prealloc
> >        0.20 ±  3%      +0.4        0.65 ±  2%  perf-profile.self.cycles-pp.mas_update_gap
> >        0.41 ±  2%      +0.5        0.92        perf-profile.self.cycles-pp.mas_store_gfp
> >        0.64            +0.6        1.23        perf-profile.self.cycles-pp.memcg_slab_post_alloc_hook
> >        0.59 ±  2%      +0.6        1.22        perf-profile.self.cycles-pp.__split_vma
> >        0.44 ±  2%      +0.7        1.10        perf-profile.self.cycles-pp.mas_wr_store_entry
> >        0.66            +0.7        1.35        perf-profile.self.cycles-pp.mas_preallocate
> >        0.59            +0.7        1.33        perf-profile.self.cycles-pp.do_brk_flags
> >        0.58            +0.7        1.32        perf-profile.self.cycles-pp.mas_wr_walk
> >        1.03            +0.8        1.81        perf-profile.self.cycles-pp.do_vmi_align_munmap
> >        0.78            +0.8        1.57        perf-profile.self.cycles-pp.__do_sys_brk
> >        1.79            +0.9        2.66        perf-profile.self.cycles-pp.__call_rcu_common
> >        1.02            +1.0        1.98        perf-profile.self.cycles-pp.perf_event_mmap_output
> >        1.33            +1.4        2.74        perf-profile.self.cycles-pp.kmem_cache_alloc
> >        1.64            +1.7        3.30        perf-profile.self.cycles-pp.__entry_text_start
> >        2.26            +2.2        4.50        perf-profile.self.cycles-pp.entry_SYSCALL_64_after_hwframe
> >        3.23            +3.3        6.58        perf-profile.self.cycles-pp.syscall_exit_to_user_mode
> >        0.21 ±  2%      +3.6        3.82        perf-profile.self.cycles-pp.mas_wr_node_store
> >        3.93            +3.9        7.78        perf-profile.self.cycles-pp.syscall_return_via_sysret
> > 
> > 
> > 
> > 
> > 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.
> > 
> >