Skip to contentSkip to conversation

fb7332a9fed: +5.4% tlbflush.mem_acc_time_thread_ms -6.2% will-it-scale.per_thread_ops

LKP Huang Ying 1 participant 1 message Feb 13, 2015

lore
Share & tools
lore

Huang Ying

author
#0 lore raw

FYI, we noticed the below changes on

commit fb7332a9fedfd62b1ba6530c86f39f0fa38afd49 ("mmu_gather: move minimal range calculations into generic code")

testbox/testcase/testparams: ivb42/will-it-scale/performance-brk1

63648dd20fa0780a fb7332a9fedfd62b1ba6530c86
---------------- --------------------------

         %stddev     %change         %stddev
             \          |                \  
    733067 ±  0%      -6.2%     687433 ±  0%  will-it-scale.per_thread_ops
   3728462 ±  0%      -5.7%    3516927 ±  0%  will-it-scale.per_process_ops
      0.46 ±  0%      -3.3%       0.44 ±  0%  will-it-scale.scalability
    140435 ± 22%     -75.2%      34877 ± 32%  sched_debug.cpu#19.ttwu_count
        85 ± 32%    +144.6%        208 ± 30%  sched_debug.cfs_rq[8]:/.blocked_load_avg
    892437 ± 34%     -55.5%     397177 ± 16%  sched_debug.cpu#5.nr_switches
    445593 ± 34%     -55.5%     198460 ± 16%  sched_debug.cpu#5.sched_goidle
        92 ± 29%    +136.7%        219 ± 29%  sched_debug.cfs_rq[8]:/.tg_load_contrib
    374967 ± 14%     -56.3%     163891 ± 35%  sched_debug.cpu#5.ttwu_count
    924309 ± 32%     -51.0%     452960 ±  9%  sched_debug.cpu#5.sched_count
        90 ± 41%     +78.2%        161 ± 18%  sched_debug.cfs_rq[30]:/.tg_load_contrib
    205529 ± 15%     -38.3%     126863 ± 32%  sched_debug.cpu#40.ttwu_count
      1152 ± 33%     +50.4%       1734 ± 22%  sched_debug.cpu#13.ttwu_local
      3.44 ± 38%     +86.6%       6.42 ± 18%  perf-profile.cpu-cycles.rwsem_spin_on_owner.rwsem_down_write_failed.call_rwsem_down_write_failed.sys_brk.system_call_fastpath
         6 ± 39%    +100.0%         13 ±  9%  sched_debug.cpu#2.cpu_load[0]
      1.03 ± 11%     +46.6%       1.52 ± 12%  perf-profile.cpu-cycles.find_vma.do_munmap.sys_brk.system_call_fastpath.brk
      0.76 ± 32%     +65.2%       1.26 ± 14%  perf-profile.cpu-cycles.up_write.vma_adjust.vma_merge.do_brk.sys_brk
    538886 ± 38%     +70.3%     917481 ± 20%  sched_debug.cpu#26.ttwu_count
      0.76 ± 21%     +69.6%       1.28 ± 14%  perf-profile.cpu-cycles.find_vma.sys_brk.system_call_fastpath.brk
        16 ±  5%     +25.8%         20 ± 18%  sched_debug.cpu#30.cpu_load[1]
      2224 ± 11%     +21.8%       2709 ± 16%  sched_debug.cpu#41.curr->pid
      3.94 ±  9%     -30.5%       2.74 ± 17%  perf-profile.cpu-cycles._raw_spin_lock.try_to_wake_up.wake_up_process.__rwsem_do_wake.rwsem_wake
        28 ± 14%     -32.1%         19 ± 14%  sched_debug.cfs_rq[25]:/.load
        16 ±  4%     +25.0%         20 ± 12%  sched_debug.cpu#32.cpu_load[2]
        17 ± 16%     +20.3%         20 ± 10%  sched_debug.cfs_rq[34]:/.runnable_load_avg
    180505 ± 26%     -43.4%     102128 ± 18%  sched_debug.cpu#44.ttwu_count
      2135 ±  7%     +28.7%       2747 ± 22%  sched_debug.cpu#44.curr->pid
     13.14 ± 10%     +20.4%      15.82 ±  5%  perf-profile.cpu-cycles.call_rwsem_down_write_failed.sys_brk.system_call_fastpath.brk
     13.05 ± 10%     +20.4%      15.71 ±  5%  perf-profile.cpu-cycles.rwsem_down_write_failed.call_rwsem_down_write_failed.sys_brk.system_call_fastpath.brk
      2.30 ± 10%     +24.1%       2.86 ±  7%  perf-profile.cpu-cycles.vma_adjust.vma_merge.do_brk.sys_brk.system_call_fastpath
      1.70 ±  6%     -13.2%       1.47 ± 11%  perf-profile.cpu-cycles.clockevents_program_event.tick_program_event.__hrtimer_start_range_ns.hrtimer_start_range_ns.tick_nohz_restart
      5512 ±  1%     +27.7%       7040 ± 22%  sched_debug.cfs_rq[20]:/.exec_clock
        17 ±  5%     -30.9%         11 ± 32%  sched_debug.cpu#40.load
      2.73 ±  9%     +21.0%       3.30 ±  7%  perf-profile.cpu-cycles.vma_merge.do_brk.sys_brk.system_call_fastpath.brk
    505131 ± 13%     +26.4%     638526 ±  7%  sched_debug.cpu#32.ttwu_count
      1.09 ±  7%     -14.9%       0.93 ± 11%  perf-profile.cpu-cycles._raw_spin_unlock_irqrestore.rwsem_wake.call_rwsem_wake.sys_brk.system_call_fastpath
        16 ±  2%     +13.8%         18 ±  8%  sched_debug.cpu#32.cpu_load[3]
      1.73 ±  6%     -12.5%       1.52 ± 11%  perf-profile.cpu-cycles.tick_program_event.__hrtimer_start_range_ns.hrtimer_start_range_ns.tick_nohz_restart.tick_nohz_idle_exit
      1.89 ±  5%     -12.2%       1.66 ±  1%  perf-profile.cpu-cycles.set_next_entity.pick_next_task_fair.__sched_text_start.schedule_preempt_disabled.cpu_startup_entry
     17.50 ±  5%     -14.4%      14.98 ±  6%  perf-profile.cpu-cycles.try_to_wake_up.wake_up_process.__rwsem_do_wake.rwsem_wake.call_rwsem_wake
     18.55 ±  5%     -13.8%      16.00 ±  6%  perf-profile.cpu-cycles.wake_up_process.__rwsem_do_wake.rwsem_wake.call_rwsem_wake.sys_brk
       229 ±  6%     -10.5%        205 ±  0%  sched_debug.cfs_rq[2]:/.tg_runnable_contrib
     10557 ±  6%     -10.2%       9478 ±  0%  sched_debug.cfs_rq[2]:/.avg->runnable_avg_sum
     18.66 ±  5%     -13.4%      16.16 ±  5%  perf-profile.cpu-cycles.__rwsem_do_wake.rwsem_wake.call_rwsem_wake.sys_brk.system_call_fastpath
    745968 ±  4%      +9.1%     813977 ±  5%  sched_debug.cpu#10.avg_idle
      3.53 ±  2%     +13.0%       3.98 ±  5%  perf-profile.cpu-cycles.unmap_region.do_munmap.sys_brk.system_call_fastpath.brk
    738761 ±  5%     +13.5%     838441 ±  3%  sched_debug.cpu#2.avg_idle
     21.09 ±  5%     -13.4%      18.27 ±  6%  perf-profile.cpu-cycles.rwsem_wake.call_rwsem_wake.sys_brk.system_call_fastpath.brk
        16 ±  6%     +12.1%         18 ±  2%  sched_debug.cpu#29.cpu_load[2]
     21.21 ±  5%     -13.2%      18.40 ±  6%  perf-profile.cpu-cycles.call_rwsem_wake.sys_brk.system_call_fastpath.brk
     11.09 ±  3%     +11.3%      12.35 ±  2%  perf-profile.cpu-cycles.do_munmap.sys_brk.system_call_fastpath.brk
    678615 ±  4%      +7.4%     728663 ±  3%  sched_debug.cpu#25.avg_idle
      2757 ±  5%      -7.4%       2551 ±  2%  sched_debug.cpu#25.curr->pid
       885 ± 11%     +22.0%       1080 ± 11%  slabinfo.RAW.num_objs
       885 ± 11%     +22.0%       1080 ± 11%  slabinfo.RAW.active_objs
      2528 ±  3%      +4.7%       2647 ±  5%  sched_debug.cpu#35.curr->pid
       230 ±  4%     -10.0%        207 ±  1%  sched_debug.cfs_rq[8]:/.tg_runnable_contrib

testbox/testcase/testparams: lkp-t410/tlbflush/performance-200%-32x-512

63648dd20fa0780a fb7332a9fedfd62b1ba6530c86
---------------- --------------------------

       113 ±  0%      -5.5%        107 ±  0%  tlbflush.mem_acc_cost_ns_time
      8758 ±  0%      +5.4%       9227 ±  0%  tlbflush.mem_acc_time_thread_ms
      3314 ± 11%     -21.2%       2610 ± 15%  slabinfo.anon_vma.num_objs

ivb42: Ivytown Ivy Bridge-EP
Memory: 64G

lkp-t410: Westmere
Memory: 2G

tlbflush.mem_acc_time_thread_ms

9300 ++-O-------------------O--------O-------------------O----------------O

       |                   O        O                            O O     O  |
  9200 O+       O  O    O        O        O O  O  O  O  O     O       O     |
  9100 ++    O                                                              |
       |                                                                    |
  9000 ++            O                                                      |
       |                                                                    |
  8900 ++                                        .*                         |
       |  *                     .*..      *    *.  +                        |
  8800 ++. +        .*..*     *.         + + ..     +  .*..  .*..           |
  8700 *+   +     .*     :   :      *.. +   *        *.    *.    *          |
       |     *..*.        :  :         *                                    |
  8600 ++                 : :                                               |
       |                   *                                                |

8500 ++-------------------------------------------------------------------+

[*] bisect-good sample
[O] bisect-bad sample

To reproduce:

apt-get install ruby
git clone git://git.kernel.org/pub/scm/linux/kernel/git/wfg/lkp-tests.git
cd lkp-tests
bin/setup-local job.yaml # the job file attached in this email
bin/run-local job.yaml

Disclaimer:
Results have been estimated based on internal Intel analysis and are provided for informational purposes only. Any difference in system hardware or software design or configuration may affect actual performance.

Thanks,
Fengguang

_______________________________________________
LKP mailing list
LKP(a)linux.intel.com

---
testcase: tlbflush
default_monitors:
wait: pre-test
uptime:
iostat:
vmstat:
numa-numastat:
numa-vmstat:
numa-meminfo:
proc-vmstat:
proc-stat:
meminfo:
slabinfo:
interrupts:
lock_stat:
latency_stats:
softirqs:
bdi_dev_mapping:
diskstats:
cpuidle:
cpufreq:
turbostat:
sched_debug:
interval: 10
pmeter:
default_watchdogs:
watch-oom:
watchdog:
cpufreq_governor:
- performance
commit: 97bf6af1f928216fd6c5a66e8a57bfa95a659672
model: Westmere
memory: 2G
hdd_partitions: "/dev/disk/by-id/ata-FUJITSU_MJA2250BH_G2_K95CT9C2G29W-part6"
swap_partitions:
rootfs_partition: "/dev/disk/by-id/ata-FUJITSU_MJA2250BH_G2_K95CT9C2G29W-part7"
nr_threads: 200%
iterations: 32x
tlbflush:
entries: 512
testbox: lkp-t410
tbox_group: lkp-t410
kconfig: x86_64-rhel
enqueue_time: 2014-12-15 13:40:54.179622910 +08:00
head_commit: 9718143f8eb121a6c53757855427f41da3a5ad7f
base_commit: 97bf6af1f928216fd6c5a66e8a57bfa95a659672
branch: linux-devel/devel-hourly-2014122315
kernel: "/kernel/x86_64-rhel/97bf6af1f928216fd6c5a66e8a57bfa95a659672/vmlinuz-3.19.0-rc1-g97bf6af"
user: lkp
queue: cyclic
rootfs: debian-x86_64.cgz
result_root: "/result/lkp-t410/tlbflush/performance-200%-32x-512/debian-x86_64.cgz/x86_64-rhel/97bf6af1f928216fd6c5a66e8a57bfa95a659672/0"
job_file: "/lkp/scheduled/lkp-t410/cyclic_tlbflush-performance-200%-32x-512-x86_64-rhel-BASE-97bf6af1f928216fd6c5a66e8a57bfa95a659672-0.yaml"
dequeue_time: 2014-12-23 16:43:56.407165575 +08:00
job_state: finished
loadavg: 0.10 0.13 0.08 1/98 7750
start_time: '1419324268'
end_time: '1419324762'
version: "/lkp/lkp/.src-20141223-112927"

echo performance > /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor echo performance > /sys/devices/system/cpu/cpu1/cpufreq/scaling_governor echo performance > /sys/devices/system/cpu/cpu2/cpufreq/scaling_governor echo performance > /sys/devices/system/cpu/cpu3/cpufreq/scaling_governor ./tlbflush -n 91 -t 8
./tlbflush -n 159 -t 8
./tlbflush -n 205 -t 8
./tlbflush -n 314 -t 8
./tlbflush -n 352 -t 8
./tlbflush -n 430 -t 8
./tlbflush -n 479 -t 8
./tlbflush -n 36 -t 8
./tlbflush -n 76 -t 8
./tlbflush -n 166 -t 8
./tlbflush -n 253 -t 8
./tlbflush -n 271 -t 8
./tlbflush -n 372 -t 8
./tlbflush -n 397 -t 8
./tlbflush -n 511 -t 8
./tlbflush -n 50 -t 8
./tlbflush -n 87 -t 8
./tlbflush -n 162 -t 8
./tlbflush -n 223 -t 8
./tlbflush -n 271 -t 8
./tlbflush -n 378 -t 8
./tlbflush -n 437 -t 8
./tlbflush -n 465 -t 8
./tlbflush -n 27 -t 8
./tlbflush -n 97 -t 8
./tlbflush -n 172 -t 8
./tlbflush -n 225 -t 8
./tlbflush -n 289 -t 8
./tlbflush -n 339 -t 8
./tlbflush -n 436 -t 8
./tlbflush -n 449 -t 8
./tlbflush -n 53 -t 8