Re: [linux-next:master] [xfs] 3062a738d7: filebench.sum_operations/s -85.0% regression

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



This is a known problem.

I have a fix that moves the floor handing into the timekeeper, but
Thomas said he had a better way to do this, so I haven't resent them
yet.

The patches in Christian's tree are out of date, so it may be best to
just drop them for now until I have the newer set ready.

Thanks,
Jeff

On Sun, 2024-09-29 at 22:36 +0800, kernel test robot wrote:
> hi, Jeff Layton,
> 
> we reported
> "[jlayton:mgtime] [xfs]  4edee232ed:  fio.write_iops -34.9% regression"
> in
> https://lore.kernel.org/all/202406141453.7a44f956-oliver.sang@xxxxxxxxx/
> 
> you asked us to supply further information at that time.
> 
> now we noticed this commit is in linux-next/master, and we observed the
> regression for a different test - filebench. FYI.
> 
> 
> 
> Hello,
> 
> kernel test robot noticed a -85.0% regression of filebench.sum_operations/s on:
> 
> 
> commit: 3062a738d73c866bf50df13bc47a2223b7b47d87 ("xfs: switch to multigrain timestamps")
> https://git.kernel.org/cgit/linux/kernel/git/next/linux-next.git master
> 
> testcase: filebench
> test machine: 96 threads 2 sockets Intel(R) Xeon(R) Platinum 8260L CPU @ 2.40GHz (Cascade Lake) with 128G memory
> parameters:
> 
> 	disk: 1HDD
> 	fs: xfs
> 	fs2: nfsv4
> 	test: filemicro_rwritefsync.f
> 	cpufreq_governor: performance
> 
> 
> 
> 
> If you fix the issue in a separate patch/commit (i.e. not just a new version of
> the same patch/commit), kindly add following tags
> > Reported-by: kernel test robot <oliver.sang@xxxxxxxxx>
> > Closes: https://lore.kernel.org/oe-lkp/202409292200.d8132f52-oliver.sang@xxxxxxxxx
> 
> 
> Details are as below:
> -------------------------------------------------------------------------------------------------->
> 
> 
> The kernel config and materials to reproduce are available at:
> https://download.01.org/0day-ci/archive/20240929/202409292200.d8132f52-oliver.sang@xxxxxxxxx
> 
> =========================================================================================
> compiler/cpufreq_governor/disk/fs2/fs/kconfig/rootfs/tbox_group/test/testcase:
>   gcc-12/performance/1HDD/nfsv4/xfs/x86_64-rhel-8.3/debian-12-x86_64-20240206.cgz/lkp-csl-2sp3/filemicro_rwritefsync.f/filebench
> 
> commit: 
>   42ba4ae657 ("Documentation: add a new file documenting multigrain timestamps")
>   3062a738d7 ("xfs: switch to multigrain timestamps")
> 
> 42ba4ae65752b8cb 3062a738d73c866bf50df13bc47 
> ---------------- --------------------------- 
>          %stddev     %change         %stddev
>              \          |                \  
>       1.48           -13.5%       1.28 ±  5%  iostat.cpu.iowait
>  4.302e+10 ±  2%     -21.9%  3.361e+10 ±  2%  cpuidle..time
>    2316977           -10.5%    2072537        cpuidle..usage
>     763659 ± 17%     -33.4%     508644 ± 15%  numa-numastat.node1.local_node
>     817625 ± 14%     -30.4%     568838 ± 11%  numa-numastat.node1.numa_hit
>       0.32 ± 12%      -0.0        0.27 ±  6%  perf-profile.children.cycles-pp.idle_cpu
>       0.31 ± 12%      -0.0        0.26 ±  6%  perf-profile.self.cycles-pp.idle_cpu
>       0.03 ± 88%    +128.5%       0.08 ±  6%  perf-sched.sch_delay.avg.ms.schedule_timeout.transaction_kthread.kthread.ret_from_fork
>       0.03 ± 88%    +128.5%       0.08 ±  6%  perf-sched.sch_delay.max.ms.schedule_timeout.transaction_kthread.kthread.ret_from_fork
>     523.13           -18.9%     424.32 ±  2%  uptime.boot
>      48962           -18.9%      39691 ±  2%  uptime.idle
>     208.10 ±  7%     -67.4%      67.75 ± 52%  numa-vmstat.node0.nr_mlock
>     816180 ± 14%     -30.4%     567697 ± 11%  numa-vmstat.node1.numa_hit
>     762194 ± 17%     -33.4%     507502 ± 15%  numa-vmstat.node1.numa_local
>    4823448           +10.8%    5344826        meminfo.Cached
>     244108 ±  6%     -41.9%     141806 ± 29%  meminfo.Dirty
>    2242446           +23.2%    2763206        meminfo.Inactive
>    1552268           +34.0%    2080318        meminfo.Inactive(file)
>       1.49            -0.2        1.29 ±  5%  mpstat.cpu.all.iowait%
>       0.10 ±  2%      +0.0        0.12 ±  4%  mpstat.cpu.all.irq%
>       0.01 ±  4%      +0.0        0.02 ±  2%  mpstat.cpu.all.soft%
>       0.03 ±  2%      +0.0        0.04 ±  2%  mpstat.cpu.all.usr%
>       2603 ±  2%     +16.6%       3036 ±  2%  vmstat.io.bo
>       1.43           -13.6%       1.23 ±  7%  vmstat.procs.b
>       2058           -14.7%       1756        vmstat.system.cs
>       5528           +21.9%       6738 ±  2%  vmstat.system.in
>       2.10           -84.9%       0.32 ± 21%  filebench.sum_bytes_mb/s
>      16385           -85.0%       2456 ± 13%  filebench.sum_operations
>     273.04           -85.0%      40.94 ± 13%  filebench.sum_operations/s
>       0.00 ± 14%  +7.2e+05%      24.14 ± 13%  filebench.sum_time_ms/op
>     273.00           -85.0%      41.00 ± 13%  filebench.sum_writes/s
>     447.44 ±  2%     -21.9%     349.26 ±  2%  filebench.time.elapsed_time
>     447.44 ±  2%     -21.9%     349.26 ±  2%  filebench.time.elapsed_time.max
>    2343344          +207.9%    7214762 ±  9%  filebench.time.file_system_outputs
>       8762 ±  2%     -80.1%       1747 ±  6%  filebench.time.voluntary_context_switches
>     269483 ±  6%     -17.0%     223745 ±  8%  sched_debug.cpu.clock.avg
>     269496 ±  6%     -17.0%     223755 ±  8%  sched_debug.cpu.clock.max
>     269474 ±  6%     -17.0%     223735 ±  8%  sched_debug.cpu.clock.min
>     268974 ±  6%     -17.0%     223268 ±  8%  sched_debug.cpu.clock_task.avg
>     269263 ±  6%     -17.0%     223549 ±  8%  sched_debug.cpu.clock_task.max
>     261595 ±  6%     -17.5%     215932 ±  8%  sched_debug.cpu.clock_task.min
>       8873 ±  4%     -12.9%       7731 ±  5%  sched_debug.cpu.curr->pid.max
>       1033 ±  3%      -9.4%     936.02 ±  5%  sched_debug.cpu.curr->pid.stddev
>       6038 ±  5%     -27.7%       4366 ±  7%  sched_debug.cpu.nr_switches.avg
>     977.53 ±  7%     -17.2%     809.02 ±  8%  sched_debug.cpu.nr_switches.min
>     269486 ±  6%     -17.0%     223746 ±  8%  sched_debug.cpu_clk
>     268914 ±  6%     -17.0%     223174 ±  8%  sched_debug.ktime
>     270076 ±  6%     -16.9%     224334 ±  8%  sched_debug.sched_clk
>      15708            -6.1%      14746 ±  2%  proc-vmstat.nr_active_anon
>     817011          +120.3%    1799485 ±  9%  proc-vmstat.nr_dirtied
>      61152 ±  6%     -42.0%      35491 ± 30%  proc-vmstat.nr_dirty
>    1206117           +10.8%    1335949        proc-vmstat.nr_file_pages
>     388315           +33.9%     519818        proc-vmstat.nr_inactive_file
>      18531            +1.0%      18721        proc-vmstat.nr_kernel_stack
>      16540            -3.8%      15909        proc-vmstat.nr_mapped
>     213.82 ±  4%     -53.7%      99.00 ± 14%  proc-vmstat.nr_mlock
>      25295            -6.6%      23625        proc-vmstat.nr_shmem
>      24447            -1.6%      24066        proc-vmstat.nr_slab_reclaimable
>     817011          +120.3%    1799485 ±  9%  proc-vmstat.nr_written
>      15708            -6.1%      14746 ±  2%  proc-vmstat.nr_zone_active_anon
>     388315           +33.9%     519818        proc-vmstat.nr_zone_inactive_file
>     139175           -19.7%     111713 ±  7%  proc-vmstat.nr_zone_write_pending
>    1299984           -13.1%    1129235        proc-vmstat.numa_hit
>    1200290           -14.2%    1029746        proc-vmstat.numa_local
>    2273118 ±  2%     -10.2%    2042156        proc-vmstat.pgalloc_normal
>    1203036           -18.6%     979402 ±  2%  proc-vmstat.pgfault
>    1743917 ±  2%     -13.4%    1509546        proc-vmstat.pgfree
>    1171848            -8.8%    1069087        proc-vmstat.pgpgout
>      56737           -18.4%      46304 ±  2%  proc-vmstat.pgreuse
>       2.40 ±  2%      +7.8%       2.59 ±  3%  perf-stat.i.MPKI
>   49439015           +14.9%   56804851 ±  2%  perf-stat.i.branch-instructions
>       4.34            +0.1        4.47        perf-stat.i.branch-miss-rate%
>    2634429           +20.4%    3171479 ±  2%  perf-stat.i.branch-misses
>       5.07 ±  2%      +0.4        5.46 ±  3%  perf-stat.i.cache-miss-rate%
>     545304 ±  4%     +22.0%     665397 ±  5%  perf-stat.i.cache-misses
>    7567339           +11.9%    8468261        perf-stat.i.cache-references
>       2021           -15.8%       1702        perf-stat.i.context-switches
>       2.24            +3.8%       2.33        perf-stat.i.cpi
>  4.391e+08           +14.4%  5.022e+08 ±  2%  perf-stat.i.cpu-cycles
>     102.06            +1.4%     103.52        perf-stat.i.cpu-migrations
>  2.401e+08           +15.0%  2.761e+08 ±  2%  perf-stat.i.instructions
>       0.01 ±  6%    -100.0%       0.00        perf-stat.i.metric.K/sec
>       5.33            +0.3        5.58        perf-stat.overall.branch-miss-rate%
>       7.21 ±  4%      +0.7        7.86 ±  5%  perf-stat.overall.cache-miss-rate%
>   49259568           +14.8%   56567448 ±  2%  perf-stat.ps.branch-instructions
>    2625073           +20.3%    3158787 ±  2%  perf-stat.ps.branch-misses
>     543464 ±  4%     +22.0%     662949 ±  5%  perf-stat.ps.cache-misses
>    7540908           +11.8%    8433223        perf-stat.ps.cache-references
>       2017           -15.8%       1697        perf-stat.ps.context-switches
>  4.373e+08           +14.3%  4.999e+08 ±  2%  perf-stat.ps.cpu-cycles
>     101.82            +1.4%     103.21        perf-stat.ps.cpu-migrations
>  2.392e+08           +14.9%   2.75e+08 ±  2%  perf-stat.ps.instructions
>  1.072e+11           -10.2%  9.632e+10        perf-stat.total.instructions
> 
> 
> 
> 
> 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.
> 
> 

-- 
Jeff Layton <jlayton@xxxxxxxxxx>





[Index of Archives]     [XFS Filesystem Development (older mail)]     [Linux Filesystem Development]     [Linux Audio Users]     [Yosemite Trails]     [Linux Kernel]     [Linux RAID]     [Linux SCSI]


  Powered by Linux