Re: [lkp] [mm/vmstat] 6cdb18ad98: -8.5% will-it-scale.per_thread_ops

[Huang\, Ying]

Heiko Carstens <heiko.carstens@xxxxxxxxxx> writes:

> On Wed, Jan 06, 2016 at 11:20:55AM +0800, kernel test robot wrote:
>> FYI, we noticed the below changes on
>> 
>> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
>> commit 6cdb18ad98a49f7e9b95d538a0614cde827404b8 ("mm/vmstat: fix overflow in mod_zone_page_state()")
>> 
>> 
>> =========================================================================================
>> compiler/cpufreq_governor/kconfig/rootfs/tbox_group/test/testcase:
>>   gcc-4.9/performance/x86_64-rhel/debian-x86_64-2015-02-07.cgz/ivb42/pread1/will-it-scale
>> 
>> commit: 
>>   cc28d6d80f6ab494b10f0e2ec949eacd610f66e3
>>   6cdb18ad98a49f7e9b95d538a0614cde827404b8
>> 
>> cc28d6d80f6ab494 6cdb18ad98a49f7e9b95d538a0 
>> ---------------- -------------------------- 
>>          %stddev     %change         %stddev
>>              \          |                \  
>>    2733943 .  0%      -8.5%    2502129 .  0%  will-it-scale.per_thread_ops
>>       3410 .  0%      -2.0%       3343 .  0%  will-it-scale.time.system_time
>>     340.08 .  0%     +19.7%     406.99 .  0%  will-it-scale.time.user_time
>>   69882822 .  2%     -24.3%   52926191 .  5%  cpuidle.C1-IVT.time
>>     340.08 .  0%     +19.7%     406.99 .  0%  time.user_time
>>     491.25 .  6%     -17.7%     404.25 .  7%  numa-vmstat.node0.nr_alloc_batch
>>       2799 . 20%     -36.6%       1776 .  0%  numa-vmstat.node0.nr_mapped
>>     630.00 .140%    +244.4%       2169 .  1%  numa-vmstat.node1.nr_inactive_anon
>
> Hmm... this is odd. I did review all callers of mod_zone_page_state() and
> couldn't find anything obvious that would go wrong after the int -> long
> change.
>
> I also tried the "pread1_threads" test case from
> https://github.com/antonblanchard/will-it-scale.git
>
> However the results seem to vary a lot after a reboot(!), at least on s390.
>
> So I'm not sure if this is really a regression.

Most part of the regression is restored for v4.4.  But because the changes are
like "V", it is hard to bisect.

=========================================================================================
compiler/cpufreq_governor/kconfig/mode/nr_task/rootfs/tbox_group/test/testcase:
  gcc-4.9/performance/x86_64-rhel/thread/24/debian-x86_64-2015-02-07.cgz/ivb42/pread1/will-it-scale

commit: 
  cc28d6d80f6ab494b10f0e2ec949eacd610f66e3
  6cdb18ad98a49f7e9b95d538a0614cde827404b8
  v4.4

cc28d6d80f6ab494 6cdb18ad98a49f7e9b95d538a0                       v4.4 
---------------- -------------------------- -------------------------- 
         %stddev     %change         %stddev     %change         %stddev
             \          |                \          |                \  
   3083436 Â  0%      -9.6%    2788374 Â  0%      -3.7%    2970130 Â  0%  will-it-scale.per_thread_ops
      6447 Â  0%      -2.2%       6308 Â  0%      -0.3%       6425 Â  0%  will-it-scale.time.system_time
    776.90 Â  0%     +17.9%     915.71 Â  0%      +2.9%     799.12 Â  0%  will-it-scale.time.user_time
    316177 Â  4%      -4.6%     301616 Â  3%     -10.3%     283563 Â  3%  softirqs.RCU
    776.90 Â  0%     +17.9%     915.71 Â  0%      +2.9%     799.12 Â  0%  time.user_time
    777.33 Â  7%     +20.8%     938.67 Â  7%      +7.5%     836.00 Â  8%  slabinfo.blkdev_requests.active_objs
    777.33 Â  7%     +20.8%     938.67 Â  7%      +7.5%     836.00 Â  8%  slabinfo.blkdev_requests.num_objs
  74313962 Â 44%     -16.5%   62053062 Â 41%     -49.9%   37246967 Â  8%  cpuidle.C1-IVT.time
  43381614 Â 79%     +24.4%   53966568 Â111%    +123.9%   97135791 Â 33%  cpuidle.C1E-IVT.time
     97.67 Â 36%     +95.2%     190.67 Â 63%    +122.5%     217.33 Â 41%  cpuidle.C3-IVT.usage
   3679437 Â 69%    -100.0%       0.00 Â -1%    -100.0%       0.00 Â -1%  latency_stats.avg.nfs_wait_on_request.nfs_updatepage.nfs_write_end.generic_perform_write.__generic_file_write_iter.generic_file_write_iter.nfs_file_write.__vfs_write.vfs_write.SyS_write.entry_SYSCALL_64_fastpath
   5177475 Â 82%    -100.0%       0.00 Â -1%    -100.0%       0.00 Â -1%  latency_stats.max.nfs_wait_on_request.nfs_updatepage.nfs_write_end.generic_perform_write.__generic_file_write_iter.generic_file_write_iter.nfs_file_write.__vfs_write.vfs_write.SyS_write.entry_SYSCALL_64_fastpath
  11726393 Â112%    -100.0%       0.00 Â -1%    -100.0%       0.00 Â -1%  latency_stats.sum.nfs_wait_on_request.nfs_updatepage.nfs_write_end.generic_perform_write.__generic_file_write_iter.generic_file_write_iter.nfs_file_write.__vfs_write.vfs_write.SyS_write.entry_SYSCALL_64_fastpath
    178.07 Â  0%      -1.3%     175.79 Â  0%      -0.8%     176.62 Â  0%  turbostat.CorWatt
      0.20 Â  2%     -16.9%       0.16 Â 18%     -11.9%       0.17 Â 17%  turbostat.Pkg%pc6
    207.38 Â  0%      -1.1%     205.13 Â  0%      -0.7%     205.99 Â  0%  turbostat.PkgWatt
      6889 Â 33%     -49.2%       3497 Â 86%     -19.4%       5552 Â 27%  numa-vmstat.node0.nr_active_anon
    483.33 Â 29%     -32.3%     327.00 Â 48%      +0.1%     483.67 Â 29%  numa-vmstat.node0.nr_page_table_pages
     27536 Â 96%     +10.9%      30535 Â 78%    +148.5%      68418 Â  2%  numa-vmstat.node0.numa_other
    214.00 Â 11%     +18.1%     252.67 Â  4%      +2.8%     220.00 Â  9%  numa-vmstat.node1.nr_kernel_stack
    370.67 Â 38%     +42.0%     526.33 Â 30%      -0.2%     370.00 Â 39%  numa-vmstat.node1.nr_page_table_pages
     61177 Â 43%      -5.2%      57976 Â 41%     -66.3%      20644 Â 10%  numa-vmstat.node1.numa_other
     78172 Â 13%     -16.1%      65573 Â 18%      -5.8%      73626 Â  9%  numa-meminfo.node0.Active
     27560 Â 33%     -49.2%      14006 Â 86%     -19.4%      22203 Â 27%  numa-meminfo.node0.Active(anon)
      3891 Â 58%     -38.1%       2407 Â100%     -58.8%       1604 Â110%  numa-meminfo.node0.AnonHugePages
      1934 Â 29%     -32.3%       1309 Â 48%      +0.1%       1936 Â 29%  numa-meminfo.node0.PageTables
     63139 Â 17%     +19.8%      75670 Â 16%      +6.0%      66937 Â 10%  numa-meminfo.node1.Active
      3432 Â 11%     +18.0%       4049 Â  4%      +2.8%       3527 Â  9%  numa-meminfo.node1.KernelStack
      1483 Â 38%     +42.0%       2106 Â 30%      -0.2%       1481 Â 39%  numa-meminfo.node1.PageTables
      1.47 Â  1%     -11.8%       1.30 Â  2%      -7.0%       1.37 Â  3%  perf-profile.cycles-pp.___might_sleep.__might_sleep.find_lock_entry.shmem_getpage_gfp.shmem_file_read_iter
      2.00 Â  2%     -11.3%       1.78 Â  2%      -7.2%       1.86 Â  2%  perf-profile.cycles-pp.__might_sleep.find_lock_entry.shmem_getpage_gfp.shmem_file_read_iter.__vfs_read
      2.30 Â  4%     +33.6%       3.07 Â  0%      -1.9%       2.26 Â  1%  perf-profile.cycles-pp.atime_needs_update.touch_atime.shmem_file_read_iter.__vfs_read.vfs_read
      1.05 Â  1%     -27.7%       0.76 Â  1%      -8.0%       0.96 Â  0%  perf-profile.cycles-pp.current_fs_time.atime_needs_update.touch_atime.shmem_file_read_iter.__vfs_read
      2.21 Â  3%     -11.9%       1.94 Â  2%      -9.4%       2.00 Â  0%  perf-profile.cycles-pp.fput.entry_SYSCALL_64_fastpath
      0.78 Â  2%     +38.5%       1.08 Â  2%     +23.1%       0.96 Â  3%  perf-profile.cycles-pp.fsnotify.vfs_read.sys_pread64.entry_SYSCALL_64_fastpath
      2.87 Â  7%     +42.6%       4.09 Â  1%      -0.3%       2.86 Â  2%  perf-profile.cycles-pp.touch_atime.shmem_file_read_iter.__vfs_read.vfs_read.sys_pread64
      6.68 Â  2%      -7.3%       6.19 Â  1%      -6.7%       6.23 Â  1%  perf-profile.cycles-pp.unlock_page.shmem_file_read_iter.__vfs_read.vfs_read.sys_pread64


Best Regards,
Huang, Ying