Skip to contentSkip to conversation

721c21c17ab: +11.7% will-it-scale.per_thread_ops

LKP Huang Ying 3 participants 4 messages Feb 3, 2015 – Feb 12, 2015

lore
Share & tools
lore

Huang Ying

author
#0 lore raw

FYI, we noticed the below changes on

commit 721c21c17ab958abf19a8fc611c3bd4743680e38 ("mm: mmu_gather: use tlb->end != 0 only for TLB invalidation")

testbox/testcase/testparams: nhm4/will-it-scale/performance-readseek1

       v3.19-rc4  721c21c17ab958abf19a8fc611  
----------------  --------------------------  
         %stddev     %change         %stddev
             \          |                \  
      0.56 ±  1%      +5.2%       0.59 ±  1%  will-it-scale.scalability
   1807741 ±  0%      +2.3%    1848641 ±  0%  will-it-scale.per_thread_ops
       740 ± 30%     +40.9%       1043 ± 26%  sched_debug.cpu#4.ttwu_local
      1335 ± 20%     +23.7%       1651 ± 17%  sched_debug.cpu#4.ttwu_count
       506 ±  9%     +40.8%        712 ±  1%  cpuidle.C1-NHM.usage
       120 ±  9%     +33.1%        160 ± 11%  sched_debug.cpu#7.load
       120 ±  9%     +26.2%        151 ± 10%  sched_debug.cfs_rq[7]:/.load
        90 ±  5%     -16.2%         75 ± 16%  sched_debug.cpu#6.cpu_load[4]
        96 ±  7%     +16.7%        112 ± 10%  sched_debug.cfs_rq[2]:/.runnable_load_avg

testbox/testcase/testparams: nhm4/will-it-scale/performance-pread2

       v3.19-rc4  721c21c17ab958abf19a8fc611  
----------------  --------------------------  
    900692 ±  1%     +11.7%    1005724 ±  0%  will-it-scale.per_thread_ops
  28033529 ±  0%      -1.2%   27698665 ±  0%  will-it-scale.time.voluntary_context_switches
       671 ± 22%     +40.4%        942 ± 27%  sched_debug.cfs_rq[7]:/.blocked_load_avg
       802 ± 19%     +30.9%       1049 ± 25%  sched_debug.cfs_rq[7]:/.tg_load_contrib
     44840 ±  6%     +15.6%      51846 ±  6%  meminfo.DirectMap4k
     18284 ±  1%      -7.4%      16926 ±  2%  vmstat.system.in
    378463 ±  0%      -1.2%     373746 ±  0%  vmstat.system.cs

testbox/testcase/testparams: nhm4/will-it-scale/performance-readseek3

       v3.19-rc4  721c21c17ab958abf19a8fc611  
----------------  --------------------------  
      0.55 ±  0%      +9.9%       0.60 ±  5%  will-it-scale.scalability
   1791707 ±  0%      +2.9%    1843202 ±  0%  will-it-scale.per_thread_ops
       187 ± 41%    +167.3%        501 ± 23%  sched_debug.cfs_rq[0]:/.blocked_load_avg
       281 ± 29%    +121.3%        622 ± 18%  sched_debug.cfs_rq[0]:/.tg_load_contrib
       110 ±  9%     +25.5%        138 ± 13%  sched_debug.cfs_rq[5]:/.load
       110 ±  9%     +25.9%        138 ± 13%  sched_debug.cpu#5.load
       178 ±  6%     -19.5%        144 ± 16%  sched_debug.cpu#4.cpu_load[1]
        94 ±  6%     +12.9%        107 ±  8%  sched_debug.cfs_rq[3]:/.runnable_load_avg
      1.78 ±  7%     +17.4%       2.09 ±  0%  perf-profile.cpu-cycles.put_page.shmem_file_read_iter.new_sync_read.__vfs_read.vfs_read
       187 ±  9%     -19.1%        152 ± 16%  sched_debug.cpu#4.cpu_load[2]
       757 ±  5%     +10.6%        838 ±  2%  slabinfo.kmalloc-2048.active_objs
      3064 ±  7%      +7.8%       3302 ±  6%  sched_debug.cpu#1.curr->pid
      5.23 ±  2%      +8.8%       5.69 ±  4%  perf-profile.cpu-cycles.security_file_permission.rw_verify_area.vfs_read.sys_read.system_call_fastpath
      3.23 ±  4%      +8.0%       3.48 ±  5%  perf-profile.cpu-cycles.copy_page_to_iter_iovec.copy_page_to_iter.shmem_file_read_iter.new_sync_read.__vfs_read
      4216 ±  7%      +7.5%       4531 ±  5%  slabinfo.kmalloc-192.active_objs

nhm4: Nehalem
Memory: 4G

lkp-sbx04: Sandy Bridge-EX
Memory: 64G

will-it-scale.per_thread_ops

1.04e+06 ++---------------------------------------------------------------+

           |               O            O      O                            |
  1.02e+06 ++ O   O  O  O      O            O     O      O               O  |
     1e+06 O+                     O  O               O            O         O
           |                                                O  O      O     |
    980000 ++                                                               |
           |                                                                |
    960000 ++                                                               |
           |                                                                |
    940000 ++                                                               |
    920000 ++                                                               |
           *..          *..*                                                |
    900000 ++          +                                                    |
           |  *...*.. +                                                     |

880000 ++--------*------------------------------------------------------+

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

To reproduce:

apt-get install ruby ruby-oj
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,
Huang, Ying

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

---
testcase: will-it-scale
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: 634b0bd490b7ebd7a054cea4f7e0d25748bde678
model: Nehalem
nr_cpu: 8
memory: 4G
hdd_partitions: "/dev/disk/by-id/ata-WDC_WD1003FBYZ-010FB0_WD-WCAW36812041-part1"
swap_partitions: "/dev/disk/by-id/ata-WDC_WD1003FBYZ-010FB0_WD-WCAW36812041-part2"
rootfs_partition: "/dev/disk/by-id/ata-WDC_WD1003FBYZ-010FB0_WD-WCAW36812041-part3"
netconsole_port: 6649
perf-profile:
freq: 800
will-it-scale:
test:
- pread2
testbox: nhm4
tbox_group: nhm4
kconfig: x86_64-rhel
enqueue_time: 2015-01-16 19:39:12.848821511 +08:00
head_commit: 634b0bd490b7ebd7a054cea4f7e0d25748bde678
base_commit: eaa27f34e91a14cdceed26ed6c6793ec1d186115
branch: next/master
kernel: "/kernel/x86_64-rhel/634b0bd490b7ebd7a054cea4f7e0d25748bde678/vmlinuz-3.19.0-rc4-next-20150116-g634b0bd"
user: lkp
queue: cyclic
rootfs: debian-x86_64-2015-01-13.cgz
result_root: "/result/nhm4/will-it-scale/performance-pread2/debian-x86_64-2015-01-13.cgz/x86_64-rhel/634b0bd490b7ebd7a054cea4f7e0d25748bde678/0"
job_file: "/lkp/scheduled/nhm4/cyclic_will-it-scale-performance-pread2-x86_64-rhel-HEAD-634b0bd490b7ebd7a054cea4f7e0d25748bde678-0.yaml"
dequeue_time: 2015-01-17 02:30:35.332402855 +08:00
job_state: finished
loadavg: 5.12 3.21 1.34 1/123 6702
start_time: '1421433067'
end_time: '1421433372'
version: "/lkp/lkp/.src-20150116-113525"

./runtest.py pread2 32 1 4 6 8

#1 lore raw
Quoted 4 lines · Huang Ying

FYI, we noticed the below changes on

commit 721c21c17ab958abf19a8fc611c3bd4743680e38 ("mm: mmu_gather: use tlb->end != 0 only for TLB invalidation")

Is this delta relative to the previous commit, or something else? The commit in question restored some 3.18 behaviour on task exit (namely doing a fullmm TLB flush), so it would be interesting to know if you see the same change against 3.18.

Will

author
#2 lore raw
Quoted 10 lines · Will Deacon
Quoted 4 lines · Huang Ying

FYI, we noticed the below changes on

commit 721c21c17ab958abf19a8fc611c3bd4743680e38 ("mm: mmu_gather: use tlb->end != 0 only for TLB invalidation")

Is this delta relative to the previous commit, or something else? The commit in question restored some 3.18 behaviour on task exit (namely doing a fullmm TLB flush), so it would be interesting to know if you see the same change against 3.18.

Sorry for late. The delta is relative to the previous commit. I will queue some test for 3.18 to check.

Best Regards,
Huang, Ying

#3 lore raw
Quoted 7 lines · Huang Ying
Quoted 6 lines · Will Deacon
Quoted 4 lines · Huang Ying

FYI, we noticed the below changes on

commit 721c21c17ab958abf19a8fc611c3bd4743680e38 ("mm: mmu_gather: use tlb->end != 0 only for TLB invalidation")

Quoted 7 lines
Quoted 4 lines

Is this delta relative to the previous commit, or something else? The commit in question restored some 3.18 behaviour on task exit (namely doing a fullmm TLB flush), so it would be interesting to know if you see the same change against 3.18.

Sorry for late. The delta is relative to the previous commit. I will queue some test for 3.18 to check.

The test result for v3.18 compared with the commit 721c21c is as below. You can find that the difference between v3.18 and 721c21c should be less that between v3.18 and the previous commit (parent) for will-it-scale.per_thread_ops. I think this is expected.

testbox/testcase/testparams: nhm4/will-it-scale/performance-pread2

           v3.18  721c21c17ab958abf19a8fc611
----------------  --------------------------
         %stddev     %change         %stddev
             \          |                \  
  31541022 ±  0%     -12.2%   27698665 ±  0%  will-it-scale.time.voluntary_context_switches
   1079302 ±  0%      -6.3%    1011350 ±  2%  will-it-scale.per_process_ops
      0.04 ±  0%      -4.5%       0.03 ±  1%  will-it-scale.scalability
   1038581 ±  0%      -3.2%    1005724 ±  0%  will-it-scale.per_thread_ops

Best Regards,
Huang, Ying