* [lkp] [locks] 7f3697e24d: +35.1% will-it-scale.per_thread_ops
@ 2016-01-29 1:32 kernel test robot
2016-01-29 2:38 ` Jeff Layton
0 siblings, 1 reply; 5+ messages in thread
From: kernel test robot @ 2016-01-29 1:32 UTC (permalink / raw)
To: Jeff Layton; +Cc: lkp, LKML, J. Bruce Fields, Dmitry Vyukov, Alexander Viro
[-- Attachment #1: Type: text/plain, Size: 18360 bytes --]
FYI, we noticed the below changes on
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
commit 7f3697e24dc3820b10f445a4a7d914fc356012d1 ("locks: fix unlock when fcntl_setlk races with a close")
=========================================================================================
compiler/cpufreq_governor/kconfig/rootfs/tbox_group/test/testcase:
gcc-4.9/performance/x86_64-rhel/debian-x86_64-2015-02-07.cgz/lkp-snb01/lock1/will-it-scale
commit:
9189922675ecca0fab38931d86b676e9d79602dc
7f3697e24dc3820b10f445a4a7d914fc356012d1
9189922675ecca0f 7f3697e24dc3820b10f445a4a7
---------------- --------------------------
%stddev %change %stddev
\ | \
2376432 ± 0% +2.1% 2427484 ± 0% will-it-scale.per_process_ops
807889 ± 0% +35.1% 1091496 ± 4% will-it-scale.per_thread_ops
22.08 ± 2% +89.1% 41.75 ± 5% will-it-scale.time.user_time
1238371 ± 14% +100.4% 2481345 ± 39% cpuidle.C1E-SNB.time
3098 ± 57% -66.6% 1035 ±171% numa-numastat.node1.other_node
379.25 ± 8% -21.4% 298.00 ± 12% numa-vmstat.node0.nr_alloc_batch
22.08 ± 2% +89.1% 41.75 ± 5% time.user_time
1795 ± 4% +7.5% 1930 ± 2% vmstat.system.cs
0.54 ± 5% +136.9% 1.28 ± 10% perf-profile.cycles.___might_sleep.__might_sleep.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file
1.65 ± 57% +245.2% 5.70 ± 29% perf-profile.cycles.__fdget_raw.sys_fcntl.entry_SYSCALL_64_fastpath
1.58 ± 59% +248.3% 5.50 ± 31% perf-profile.cycles.__fget.__fget_light.__fdget_raw.sys_fcntl.entry_SYSCALL_64_fastpath
1.62 ± 58% +246.3% 5.63 ± 30% perf-profile.cycles.__fget_light.__fdget_raw.sys_fcntl.entry_SYSCALL_64_fastpath
0.00 ± -1% +Inf% 5.88 ± 11% perf-profile.cycles.__memset.locks_alloc_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait
2.50 ± 2% -100.0% 0.00 ± -1% perf-profile.cycles.__memset.locks_alloc_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk
1.29 ± 4% +138.8% 3.09 ± 11% perf-profile.cycles.__memset.locks_alloc_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
0.47 ± 9% +144.4% 1.16 ± 11% perf-profile.cycles.__might_fault.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
0.37 ± 12% +140.3% 0.90 ± 9% perf-profile.cycles.__might_sleep.__might_fault.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
0.86 ± 6% +137.7% 2.05 ± 10% perf-profile.cycles.__might_sleep.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file.vfs_lock_file
0.61 ± 14% +56.8% 0.95 ± 14% perf-profile.cycles.__might_sleep.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.sys_fcntl
0.00 ± -1% +Inf% 39.84 ± 12% perf-profile.cycles.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk.sys_fcntl
16.44 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
0.00 ± -1% +Inf% 1.77 ± 11% perf-profile.cycles._raw_spin_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
59.34 ± 1% -72.4% 16.36 ± 33% perf-profile.cycles._raw_spin_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
0.46 ± 11% +144.9% 1.13 ± 19% perf-profile.cycles.avc_has_perm.inode_has_perm.file_has_perm.selinux_file_fcntl.security_file_fcntl
0.87 ± 6% +103.2% 1.77 ± 12% perf-profile.cycles.avc_has_perm.inode_has_perm.file_has_perm.selinux_file_lock.security_file_lock
0.81 ± 4% +135.7% 1.90 ± 10% perf-profile.cycles.copy_user_generic_string.sys_fcntl.entry_SYSCALL_64_fastpath
0.00 ± -1% +Inf% 41.86 ± 12% perf-profile.cycles.do_lock_file_wait.part.29.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
0.88 ± 6% +127.8% 2.00 ± 9% perf-profile.cycles.entry_SYSCALL_64
0.86 ± 4% +122.6% 1.92 ± 12% perf-profile.cycles.entry_SYSCALL_64_after_swapgs
84.98 ± 0% -9.1% 77.20 ± 2% perf-profile.cycles.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
0.76 ± 10% +142.1% 1.84 ± 14% perf-profile.cycles.file_has_perm.selinux_file_fcntl.security_file_fcntl.sys_fcntl.entry_SYSCALL_64_fastpath
1.35 ± 4% +106.3% 2.78 ± 11% perf-profile.cycles.file_has_perm.selinux_file_lock.security_file_lock.fcntl_setlk.sys_fcntl
0.00 ± -1% +Inf% 0.89 ± 12% perf-profile.cycles.flock_to_posix_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
6.90 ± 4% -48.6% 3.55 ± 27% perf-profile.cycles.fput.entry_SYSCALL_64_fastpath
0.51 ± 10% +140.5% 1.23 ± 16% perf-profile.cycles.inode_has_perm.isra.31.file_has_perm.selinux_file_fcntl.security_file_fcntl.sys_fcntl
0.98 ± 4% +97.7% 1.93 ± 11% perf-profile.cycles.inode_has_perm.isra.31.file_has_perm.selinux_file_lock.security_file_lock.fcntl_setlk
0.00 ± -1% +Inf% 6.56 ± 10% perf-profile.cycles.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait
2.75 ± 4% -100.0% 0.00 ± -1% perf-profile.cycles.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk
1.53 ± 7% +119.7% 3.37 ± 13% perf-profile.cycles.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
0.00 ± -1% +Inf% 1.79 ± 11% perf-profile.cycles.kmem_cache_free.locks_free_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait
0.46 ± 14% +257.0% 1.66 ± 11% perf-profile.cycles.kmem_cache_free.locks_free_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
0.40 ± 7% +158.6% 1.05 ± 17% perf-profile.cycles.kmem_cache_free.locks_free_lock.locks_dispose_list.__posix_lock_file.vfs_lock_file
0.00 ± -1% +Inf% 0.96 ± 10% perf-profile.cycles.lg_local_lock.locks_insert_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait
0.00 ± -1% +Inf% 14.69 ± 10% perf-profile.cycles.locks_alloc_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
6.38 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.locks_alloc_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
3.28 ± 6% +127.1% 7.45 ± 12% perf-profile.cycles.locks_alloc_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
0.00 ± -1% +Inf% 9.75 ± 13% perf-profile.cycles.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
3.61 ± 1% -100.0% 0.00 ± -1% perf-profile.cycles.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
0.00 ± -1% +Inf% 1.84 ± 11% perf-profile.cycles.locks_dispose_list.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
0.00 ± -1% +Inf% 2.42 ± 10% perf-profile.cycles.locks_free_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
1.00 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.locks_free_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
0.63 ± 11% +224.1% 2.05 ± 10% perf-profile.cycles.locks_free_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
0.00 ± -1% +Inf% 1.22 ± 14% perf-profile.cycles.locks_free_lock.locks_dispose_list.__posix_lock_file.vfs_lock_file.do_lock_file_wait
0.00 ± -1% +Inf% 6.17 ± 15% perf-profile.cycles.locks_insert_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
2.31 ± 6% -100.0% 0.00 ± -1% perf-profile.cycles.locks_insert_lock_ctx.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
0.00 ± -1% +Inf% 8.96 ± 13% perf-profile.cycles.locks_unlink_lock_ctx.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait
3.27 ± 1% -100.0% 0.00 ± -1% perf-profile.cycles.locks_unlink_lock_ctx.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.fcntl_setlk
53.88 ± 1% -79.7% 10.92 ± 46% perf-profile.cycles.native_queued_spin_lock_slowpath._raw_spin_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
2.75 ± 0% +183.3% 7.79 ± 13% perf-profile.cycles.put_pid.locks_unlink_lock_ctx.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file
1.11 ± 9% +137.2% 2.63 ± 14% perf-profile.cycles.security_file_fcntl.sys_fcntl.entry_SYSCALL_64_fastpath
1.69 ± 4% +118.2% 3.69 ± 11% perf-profile.cycles.security_file_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
0.91 ± 9% +139.0% 2.17 ± 14% perf-profile.cycles.selinux_file_fcntl.security_file_fcntl.sys_fcntl.entry_SYSCALL_64_fastpath
1.39 ± 4% +114.6% 2.97 ± 10% perf-profile.cycles.selinux_file_lock.security_file_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
0.00 ± -1% +Inf% 41.12 ± 12% perf-profile.cycles.vfs_lock_file.do_lock_file_wait.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
17.04 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.vfs_lock_file.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
34.75 ±148% +132.4% 80.75 ± 82% sched_debug.cfs_rq:/.load.8
15.00 ± 9% +198.3% 44.75 ± 72% sched_debug.cfs_rq:/.load_avg.21
25.00 ± 29% +574.0% 168.50 ± 78% sched_debug.cfs_rq:/.load_avg.9
38.47 ± 5% +29.1% 49.65 ± 26% sched_debug.cfs_rq:/.load_avg.avg
63.17 ± 10% +44.3% 91.16 ± 36% sched_debug.cfs_rq:/.load_avg.stddev
893865 ± 12% -12.5% 782455 ± 0% sched_debug.cfs_rq:/.min_vruntime.25
18.25 ± 26% +52.1% 27.75 ± 25% sched_debug.cfs_rq:/.runnable_load_avg.9
-57635 ±-68% -196.4% 55548 ±130% sched_debug.cfs_rq:/.spread0.1
-802264 ±-25% -29.5% -565458 ±-49% sched_debug.cfs_rq:/.spread0.8
-804662 ±-25% -29.4% -567811 ±-48% sched_debug.cfs_rq:/.spread0.min
1233 ± 5% +30.9% 1614 ± 28% sched_debug.cfs_rq:/.tg_load_avg.0
1233 ± 5% +30.9% 1614 ± 28% sched_debug.cfs_rq:/.tg_load_avg.1
1228 ± 5% +30.3% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.10
1228 ± 5% +30.4% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.11
1228 ± 5% +30.3% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.12
1229 ± 5% +30.0% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.13
1228 ± 5% +30.1% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.14
1229 ± 5% +30.0% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.15
1226 ± 5% +30.3% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.16
1226 ± 5% +30.2% 1597 ± 27% sched_debug.cfs_rq:/.tg_load_avg.17
1227 ± 5% +30.1% 1595 ± 27% sched_debug.cfs_rq:/.tg_load_avg.18
1227 ± 5% +29.4% 1588 ± 26% sched_debug.cfs_rq:/.tg_load_avg.19
1233 ± 5% +30.4% 1609 ± 27% sched_debug.cfs_rq:/.tg_load_avg.2
1222 ± 5% +29.9% 1587 ± 26% sched_debug.cfs_rq:/.tg_load_avg.20
1223 ± 5% +24.2% 1519 ± 20% sched_debug.cfs_rq:/.tg_load_avg.21
1223 ± 5% +23.8% 1515 ± 20% sched_debug.cfs_rq:/.tg_load_avg.22
1223 ± 5% +23.9% 1515 ± 20% sched_debug.cfs_rq:/.tg_load_avg.23
1223 ± 5% +23.9% 1515 ± 20% sched_debug.cfs_rq:/.tg_load_avg.24
1223 ± 5% +23.5% 1511 ± 19% sched_debug.cfs_rq:/.tg_load_avg.25
1224 ± 5% +23.5% 1512 ± 19% sched_debug.cfs_rq:/.tg_load_avg.26
1223 ± 5% +23.1% 1506 ± 19% sched_debug.cfs_rq:/.tg_load_avg.27
1223 ± 5% +22.5% 1499 ± 19% sched_debug.cfs_rq:/.tg_load_avg.28
1224 ± 5% +22.5% 1499 ± 19% sched_debug.cfs_rq:/.tg_load_avg.29
1233 ± 5% +30.3% 1607 ± 27% sched_debug.cfs_rq:/.tg_load_avg.3
1223 ± 5% +22.2% 1495 ± 18% sched_debug.cfs_rq:/.tg_load_avg.30
1224 ± 5% +22.0% 1493 ± 19% sched_debug.cfs_rq:/.tg_load_avg.31
1234 ± 5% +30.0% 1604 ± 28% sched_debug.cfs_rq:/.tg_load_avg.4
1233 ± 5% +30.0% 1604 ± 28% sched_debug.cfs_rq:/.tg_load_avg.5
1231 ± 5% +30.3% 1604 ± 28% sched_debug.cfs_rq:/.tg_load_avg.6
1233 ± 5% +30.0% 1603 ± 27% sched_debug.cfs_rq:/.tg_load_avg.7
1231 ± 5% +30.1% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.8
1228 ± 5% +30.3% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.9
1228 ± 5% +27.8% 1569 ± 24% sched_debug.cfs_rq:/.tg_load_avg.avg
1246 ± 5% +30.7% 1628 ± 27% sched_debug.cfs_rq:/.tg_load_avg.max
1212 ± 5% +22.2% 1481 ± 19% sched_debug.cfs_rq:/.tg_load_avg.min
15.00 ± 9% +198.3% 44.75 ± 72% sched_debug.cfs_rq:/.tg_load_avg_contrib.21
25.00 ± 29% +574.0% 168.50 ± 78% sched_debug.cfs_rq:/.tg_load_avg_contrib.9
38.53 ± 5% +29.0% 49.71 ± 26% sched_debug.cfs_rq:/.tg_load_avg_contrib.avg
63.34 ± 10% +44.1% 91.30 ± 36% sched_debug.cfs_rq:/.tg_load_avg_contrib.stddev
532.25 ± 2% +8.5% 577.50 ± 6% sched_debug.cfs_rq:/.util_avg.15
210.75 ± 14% -14.4% 180.50 ± 4% sched_debug.cfs_rq:/.util_avg.29
450.00 ± 22% +50.7% 678.00 ± 18% sched_debug.cfs_rq:/.util_avg.9
955572 ± 4% -10.2% 857813 ± 5% sched_debug.cpu.avg_idle.6
23.99 ± 60% -76.2% 5.71 ± 24% sched_debug.cpu.clock.stddev
23.99 ± 60% -76.2% 5.71 ± 24% sched_debug.cpu.clock_task.stddev
2840 ± 37% -47.4% 1492 ± 65% sched_debug.cpu.curr->pid.25
34.75 ±148% +132.4% 80.75 ± 82% sched_debug.cpu.load.8
61776 ± 7% -7.1% 57380 ± 0% sched_debug.cpu.nr_load_updates.25
6543 ± 2% +20.4% 7879 ± 9% sched_debug.cpu.nr_switches.0
5256 ± 23% +177.1% 14566 ± 52% sched_debug.cpu.nr_switches.27
7915 ± 3% +8.7% 8605 ± 3% sched_debug.cpu.nr_switches.avg
-0.25 ±-519% +1900.0% -5.00 ±-24% sched_debug.cpu.nr_uninterruptible.12
2.00 ± 93% -125.0% -0.50 ±-300% sched_debug.cpu.nr_uninterruptible.24
17468 ± 14% +194.3% 51413 ± 75% sched_debug.cpu.sched_count.15
2112 ± 2% +20.8% 2552 ± 11% sched_debug.cpu.sched_goidle.0
2103 ± 34% +219.0% 6709 ± 55% sched_debug.cpu.sched_goidle.27
3159 ± 3% +8.2% 3418 ± 4% sched_debug.cpu.sched_goidle.avg
1323 ± 64% -72.7% 361.50 ± 15% sched_debug.cpu.ttwu_count.23
3264 ± 12% +94.4% 6347 ± 41% sched_debug.cpu.ttwu_count.27
3860 ± 3% +9.0% 4208 ± 3% sched_debug.cpu.ttwu_count.avg
2358 ± 3% +28.7% 3035 ± 9% sched_debug.cpu.ttwu_local.0
1110 ± 22% +54.6% 1716 ± 28% sched_debug.cpu.ttwu_local.27
1814 ± 8% +16.1% 2106 ± 5% sched_debug.cpu.ttwu_local.stddev
lkp-snb01: Sandy Bridge-EP
Memory: 32G
will-it-scale.per_thread_ops
1.2e+06 ++---------------------------------------------------------------+
| O |
1.15e+06 O+O O O O O O O O |
1.1e+06 ++ |
| O O O O O OO |
1.05e+06 ++ O O |
1e+06 ++ |
| |
950000 ++ |
900000 ++ |
| |
850000 ++ |
800000 *+*.*.*.*.*.*.*.*.*.*.*.*. .*.*. *.*.*.*.*.*.*.*.*.*.*.*.*.*.*.*.*
| * * |
750000 ++---------------------------------------------------------------+
will-it-scale.time.user_time
50 ++---------------------------------------------------------------------+
| |
45 ++ O O O O O O |
O O O O |
| O O O O O |
40 ++ O O O O |
| |
35 ++ |
| |
30 ++ |
| |
| * |
25 ++ + + |
*.*.*.*..*.* *.*.*..*.*.*.*.*.*.*..*.*.*.*.*.*.*..*.*.*.*.*.*..*.*.*.*
20 ++---------------------------------------------------------------------+
[*] bisect-good sample
[O] bisect-bad sample
To reproduce:
git clone git://git.kernel.org/pub/scm/linux/kernel/git/wfg/lkp-tests.git
cd lkp-tests
bin/lkp install job.yaml # job file is attached in this email
bin/lkp run 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,
Ying Huang
[-- Attachment #2: job.yaml --]
[-- Type: text/plain, Size: 3311 bytes --]
---
LKP_SERVER: inn
LKP_CGI_PORT: 80
LKP_CIFS_PORT: 139
testcase: will-it-scale
default-monitors:
wait: activate-monitor
kmsg:
uptime:
iostat:
vmstat:
numa-numastat:
numa-vmstat:
numa-meminfo:
proc-vmstat:
proc-stat:
interval: 10
meminfo:
slabinfo:
interrupts:
lock_stat:
latency_stats:
softirqs:
bdi_dev_mapping:
diskstats:
nfsstat:
cpuidle:
cpufreq-stats:
turbostat:
pmeter:
sched_debug:
interval: 60
cpufreq_governor: performance
default-watchdogs:
oom-killer:
watchdog:
commit: 7f3697e24dc3820b10f445a4a7d914fc356012d1
model: Sandy Bridge-EP
memory: 32G
hdd_partitions: "/dev/sda2"
swap_partitions:
category: benchmark
perf-profile:
freq: 800
will-it-scale:
test: lock1
queue: bisect
testbox: lkp-snb01
tbox_group: lkp-snb01
kconfig: x86_64-rhel
enqueue_time: 2016-01-28 14:50:21.699178965 +08:00
id: c3ed72938d383a211effce7facc978c2cc247aa8
user: lkp
compiler: gcc-4.9
head_commit: 92e963f50fc74041b5e9e744c330dca48e04f08d
base_commit: 5348c1e9e0dc2b62a484c4b74a8d1d59aa9620a4
branch: linus/master
rootfs: debian-x86_64-2015-02-07.cgz
result_root: "/result/will-it-scale/performance-lock1/lkp-snb01/debian-x86_64-2015-02-07.cgz/x86_64-rhel/gcc-4.9/7f3697e24dc3820b10f445a4a7d914fc356012d1/0"
job_file: "/lkp/scheduled/lkp-snb01/bisect_will-it-scale-performance-lock1-debian-x86_64-2015-02-07.cgz-x86_64-rhel-7f3697e24dc3820b10f445a4a7d914fc356012d1-20160128-75911-11hdapy-0.yaml"
nr_cpu: "$(nproc)"
max_uptime: 1500
initrd: "/osimage/debian/debian-x86_64-2015-02-07.cgz"
bootloader_append:
- root=/dev/ram0
- user=lkp
- job=/lkp/scheduled/lkp-snb01/bisect_will-it-scale-performance-lock1-debian-x86_64-2015-02-07.cgz-x86_64-rhel-7f3697e24dc3820b10f445a4a7d914fc356012d1-20160128-75911-11hdapy-0.yaml
- ARCH=x86_64
- kconfig=x86_64-rhel
- branch=linus/master
- commit=7f3697e24dc3820b10f445a4a7d914fc356012d1
- BOOT_IMAGE=/pkg/linux/x86_64-rhel/gcc-4.9/7f3697e24dc3820b10f445a4a7d914fc356012d1/vmlinuz-4.4.0-rc1-00005-g7f3697e
- max_uptime=1500
- RESULT_ROOT=/result/will-it-scale/performance-lock1/lkp-snb01/debian-x86_64-2015-02-07.cgz/x86_64-rhel/gcc-4.9/7f3697e24dc3820b10f445a4a7d914fc356012d1/0
- LKP_SERVER=inn
- |2-
earlyprintk=ttyS0,115200 systemd.log_level=err
debug apic=debug sysrq_always_enabled rcupdate.rcu_cpu_stall_timeout=100
panic=-1 softlockup_panic=1 nmi_watchdog=panic oops=panic load_ramdisk=2 prompt_ramdisk=0
console=ttyS0,115200 console=tty0 vga=normal
rw
lkp_initrd: "/lkp/lkp/lkp-x86_64.cgz"
modules_initrd: "/pkg/linux/x86_64-rhel/gcc-4.9/7f3697e24dc3820b10f445a4a7d914fc356012d1/modules.cgz"
bm_initrd: "/osimage/deps/debian-x86_64-2015-02-07.cgz/lkp.cgz,/osimage/deps/debian-x86_64-2015-02-07.cgz/run-ipconfig.cgz,/osimage/deps/debian-x86_64-2015-02-07.cgz/turbostat.cgz,/lkp/benchmarks/turbostat.cgz,/lkp/benchmarks/will-it-scale.cgz"
linux_headers_initrd: "/pkg/linux/x86_64-rhel/gcc-4.9/7f3697e24dc3820b10f445a4a7d914fc356012d1/linux-headers.cgz"
repeat_to: 2
kernel: "/pkg/linux/x86_64-rhel/gcc-4.9/7f3697e24dc3820b10f445a4a7d914fc356012d1/vmlinuz-4.4.0-rc1-00005-g7f3697e"
dequeue_time: 2016-01-28 15:11:56.910896619 +08:00
job_state: finished
loadavg: 27.88 12.58 4.93 2/368 9093
start_time: '1453965158'
end_time: '1453965468'
version: "/lkp/lkp/.src-20160127-223853"
[-- Attachment #3: reproduce.sh --]
[-- Type: application/x-sh, Size: 3058 bytes --]
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [lkp] [locks] 7f3697e24d: +35.1% will-it-scale.per_thread_ops
2016-01-29 1:32 [lkp] [locks] 7f3697e24d: +35.1% will-it-scale.per_thread_ops kernel test robot
@ 2016-01-29 2:38 ` Jeff Layton
2016-01-29 2:52 ` [LKP] " Huang, Ying
0 siblings, 1 reply; 5+ messages in thread
From: Jeff Layton @ 2016-01-29 2:38 UTC (permalink / raw)
To: kernel test robot
Cc: lkp, LKML, J. Bruce Fields, Dmitry Vyukov, Alexander Viro
On Fri, 29 Jan 2016 09:32:19 +0800
kernel test robot <ying.huang@linux.intel.com> wrote:
> FYI, we noticed the below changes on
>
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
> commit 7f3697e24dc3820b10f445a4a7d914fc356012d1 ("locks: fix unlock when fcntl_setlk races with a close")
>
>
> =========================================================================================
> compiler/cpufreq_governor/kconfig/rootfs/tbox_group/test/testcase:
> gcc-4.9/performance/x86_64-rhel/debian-x86_64-2015-02-07.cgz/lkp-snb01/lock1/will-it-scale
>
> commit:
> 9189922675ecca0fab38931d86b676e9d79602dc
> 7f3697e24dc3820b10f445a4a7d914fc356012d1
>
> 9189922675ecca0f 7f3697e24dc3820b10f445a4a7
> ---------------- --------------------------
> %stddev %change %stddev
> \ | \
> 2376432 ± 0% +2.1% 2427484 ± 0% will-it-scale.per_process_ops
> 807889 ± 0% +35.1% 1091496 ± 4% will-it-scale.per_thread_ops
> 22.08 ± 2% +89.1% 41.75 ± 5% will-it-scale.time.user_time
> 1238371 ± 14% +100.4% 2481345 ± 39% cpuidle.C1E-SNB.time
> 3098 ± 57% -66.6% 1035 ±171% numa-numastat.node1.other_node
> 379.25 ± 8% -21.4% 298.00 ± 12% numa-vmstat.node0.nr_alloc_batch
> 22.08 ± 2% +89.1% 41.75 ± 5% time.user_time
> 1795 ± 4% +7.5% 1930 ± 2% vmstat.system.cs
> 0.54 ± 5% +136.9% 1.28 ± 10% perf-profile.cycles.___might_sleep.__might_sleep.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file
> 1.65 ± 57% +245.2% 5.70 ± 29% perf-profile.cycles.__fdget_raw.sys_fcntl.entry_SYSCALL_64_fastpath
> 1.58 ± 59% +248.3% 5.50 ± 31% perf-profile.cycles.__fget.__fget_light.__fdget_raw.sys_fcntl.entry_SYSCALL_64_fastpath
> 1.62 ± 58% +246.3% 5.63 ± 30% perf-profile.cycles.__fget_light.__fdget_raw.sys_fcntl.entry_SYSCALL_64_fastpath
> 0.00 ± -1% +Inf% 5.88 ± 11% perf-profile.cycles.__memset.locks_alloc_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait
> 2.50 ± 2% -100.0% 0.00 ± -1% perf-profile.cycles.__memset.locks_alloc_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk
> 1.29 ± 4% +138.8% 3.09 ± 11% perf-profile.cycles.__memset.locks_alloc_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> 0.47 ± 9% +144.4% 1.16 ± 11% perf-profile.cycles.__might_fault.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> 0.37 ± 12% +140.3% 0.90 ± 9% perf-profile.cycles.__might_sleep.__might_fault.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> 0.86 ± 6% +137.7% 2.05 ± 10% perf-profile.cycles.__might_sleep.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file.vfs_lock_file
> 0.61 ± 14% +56.8% 0.95 ± 14% perf-profile.cycles.__might_sleep.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.sys_fcntl
> 0.00 ± -1% +Inf% 39.84 ± 12% perf-profile.cycles.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk.sys_fcntl
> 16.44 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> 0.00 ± -1% +Inf% 1.77 ± 11% perf-profile.cycles._raw_spin_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
> 59.34 ± 1% -72.4% 16.36 ± 33% perf-profile.cycles._raw_spin_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> 0.46 ± 11% +144.9% 1.13 ± 19% perf-profile.cycles.avc_has_perm.inode_has_perm.file_has_perm.selinux_file_fcntl.security_file_fcntl
> 0.87 ± 6% +103.2% 1.77 ± 12% perf-profile.cycles.avc_has_perm.inode_has_perm.file_has_perm.selinux_file_lock.security_file_lock
> 0.81 ± 4% +135.7% 1.90 ± 10% perf-profile.cycles.copy_user_generic_string.sys_fcntl.entry_SYSCALL_64_fastpath
> 0.00 ± -1% +Inf% 41.86 ± 12% perf-profile.cycles.do_lock_file_wait.part.29.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> 0.88 ± 6% +127.8% 2.00 ± 9% perf-profile.cycles.entry_SYSCALL_64
> 0.86 ± 4% +122.6% 1.92 ± 12% perf-profile.cycles.entry_SYSCALL_64_after_swapgs
> 84.98 ± 0% -9.1% 77.20 ± 2% perf-profile.cycles.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> 0.76 ± 10% +142.1% 1.84 ± 14% perf-profile.cycles.file_has_perm.selinux_file_fcntl.security_file_fcntl.sys_fcntl.entry_SYSCALL_64_fastpath
> 1.35 ± 4% +106.3% 2.78 ± 11% perf-profile.cycles.file_has_perm.selinux_file_lock.security_file_lock.fcntl_setlk.sys_fcntl
> 0.00 ± -1% +Inf% 0.89 ± 12% perf-profile.cycles.flock_to_posix_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> 6.90 ± 4% -48.6% 3.55 ± 27% perf-profile.cycles.fput.entry_SYSCALL_64_fastpath
> 0.51 ± 10% +140.5% 1.23 ± 16% perf-profile.cycles.inode_has_perm.isra.31.file_has_perm.selinux_file_fcntl.security_file_fcntl.sys_fcntl
> 0.98 ± 4% +97.7% 1.93 ± 11% perf-profile.cycles.inode_has_perm.isra.31.file_has_perm.selinux_file_lock.security_file_lock.fcntl_setlk
> 0.00 ± -1% +Inf% 6.56 ± 10% perf-profile.cycles.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait
> 2.75 ± 4% -100.0% 0.00 ± -1% perf-profile.cycles.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk
> 1.53 ± 7% +119.7% 3.37 ± 13% perf-profile.cycles.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> 0.00 ± -1% +Inf% 1.79 ± 11% perf-profile.cycles.kmem_cache_free.locks_free_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait
> 0.46 ± 14% +257.0% 1.66 ± 11% perf-profile.cycles.kmem_cache_free.locks_free_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> 0.40 ± 7% +158.6% 1.05 ± 17% perf-profile.cycles.kmem_cache_free.locks_free_lock.locks_dispose_list.__posix_lock_file.vfs_lock_file
> 0.00 ± -1% +Inf% 0.96 ± 10% perf-profile.cycles.lg_local_lock.locks_insert_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait
> 0.00 ± -1% +Inf% 14.69 ± 10% perf-profile.cycles.locks_alloc_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
> 6.38 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.locks_alloc_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
> 3.28 ± 6% +127.1% 7.45 ± 12% perf-profile.cycles.locks_alloc_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> 0.00 ± -1% +Inf% 9.75 ± 13% perf-profile.cycles.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
> 3.61 ± 1% -100.0% 0.00 ± -1% perf-profile.cycles.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
> 0.00 ± -1% +Inf% 1.84 ± 11% perf-profile.cycles.locks_dispose_list.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
> 0.00 ± -1% +Inf% 2.42 ± 10% perf-profile.cycles.locks_free_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
> 1.00 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.locks_free_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
> 0.63 ± 11% +224.1% 2.05 ± 10% perf-profile.cycles.locks_free_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> 0.00 ± -1% +Inf% 1.22 ± 14% perf-profile.cycles.locks_free_lock.locks_dispose_list.__posix_lock_file.vfs_lock_file.do_lock_file_wait
> 0.00 ± -1% +Inf% 6.17 ± 15% perf-profile.cycles.locks_insert_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
> 2.31 ± 6% -100.0% 0.00 ± -1% perf-profile.cycles.locks_insert_lock_ctx.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
> 0.00 ± -1% +Inf% 8.96 ± 13% perf-profile.cycles.locks_unlink_lock_ctx.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait
> 3.27 ± 1% -100.0% 0.00 ± -1% perf-profile.cycles.locks_unlink_lock_ctx.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.fcntl_setlk
> 53.88 ± 1% -79.7% 10.92 ± 46% perf-profile.cycles.native_queued_spin_lock_slowpath._raw_spin_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> 2.75 ± 0% +183.3% 7.79 ± 13% perf-profile.cycles.put_pid.locks_unlink_lock_ctx.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file
> 1.11 ± 9% +137.2% 2.63 ± 14% perf-profile.cycles.security_file_fcntl.sys_fcntl.entry_SYSCALL_64_fastpath
> 1.69 ± 4% +118.2% 3.69 ± 11% perf-profile.cycles.security_file_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> 0.91 ± 9% +139.0% 2.17 ± 14% perf-profile.cycles.selinux_file_fcntl.security_file_fcntl.sys_fcntl.entry_SYSCALL_64_fastpath
> 1.39 ± 4% +114.6% 2.97 ± 10% perf-profile.cycles.selinux_file_lock.security_file_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> 0.00 ± -1% +Inf% 41.12 ± 12% perf-profile.cycles.vfs_lock_file.do_lock_file_wait.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> 17.04 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.vfs_lock_file.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> 34.75 ±148% +132.4% 80.75 ± 82% sched_debug.cfs_rq:/.load.8
> 15.00 ± 9% +198.3% 44.75 ± 72% sched_debug.cfs_rq:/.load_avg.21
> 25.00 ± 29% +574.0% 168.50 ± 78% sched_debug.cfs_rq:/.load_avg.9
> 38.47 ± 5% +29.1% 49.65 ± 26% sched_debug.cfs_rq:/.load_avg.avg
> 63.17 ± 10% +44.3% 91.16 ± 36% sched_debug.cfs_rq:/.load_avg.stddev
> 893865 ± 12% -12.5% 782455 ± 0% sched_debug.cfs_rq:/.min_vruntime.25
> 18.25 ± 26% +52.1% 27.75 ± 25% sched_debug.cfs_rq:/.runnable_load_avg.9
> -57635 ±-68% -196.4% 55548 ±130% sched_debug.cfs_rq:/.spread0.1
> -802264 ±-25% -29.5% -565458 ±-49% sched_debug.cfs_rq:/.spread0.8
> -804662 ±-25% -29.4% -567811 ±-48% sched_debug.cfs_rq:/.spread0.min
> 1233 ± 5% +30.9% 1614 ± 28% sched_debug.cfs_rq:/.tg_load_avg.0
> 1233 ± 5% +30.9% 1614 ± 28% sched_debug.cfs_rq:/.tg_load_avg.1
> 1228 ± 5% +30.3% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.10
> 1228 ± 5% +30.4% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.11
> 1228 ± 5% +30.3% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.12
> 1229 ± 5% +30.0% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.13
> 1228 ± 5% +30.1% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.14
> 1229 ± 5% +30.0% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.15
> 1226 ± 5% +30.3% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.16
> 1226 ± 5% +30.2% 1597 ± 27% sched_debug.cfs_rq:/.tg_load_avg.17
> 1227 ± 5% +30.1% 1595 ± 27% sched_debug.cfs_rq:/.tg_load_avg.18
> 1227 ± 5% +29.4% 1588 ± 26% sched_debug.cfs_rq:/.tg_load_avg.19
> 1233 ± 5% +30.4% 1609 ± 27% sched_debug.cfs_rq:/.tg_load_avg.2
> 1222 ± 5% +29.9% 1587 ± 26% sched_debug.cfs_rq:/.tg_load_avg.20
> 1223 ± 5% +24.2% 1519 ± 20% sched_debug.cfs_rq:/.tg_load_avg.21
> 1223 ± 5% +23.8% 1515 ± 20% sched_debug.cfs_rq:/.tg_load_avg.22
> 1223 ± 5% +23.9% 1515 ± 20% sched_debug.cfs_rq:/.tg_load_avg.23
> 1223 ± 5% +23.9% 1515 ± 20% sched_debug.cfs_rq:/.tg_load_avg.24
> 1223 ± 5% +23.5% 1511 ± 19% sched_debug.cfs_rq:/.tg_load_avg.25
> 1224 ± 5% +23.5% 1512 ± 19% sched_debug.cfs_rq:/.tg_load_avg.26
> 1223 ± 5% +23.1% 1506 ± 19% sched_debug.cfs_rq:/.tg_load_avg.27
> 1223 ± 5% +22.5% 1499 ± 19% sched_debug.cfs_rq:/.tg_load_avg.28
> 1224 ± 5% +22.5% 1499 ± 19% sched_debug.cfs_rq:/.tg_load_avg.29
> 1233 ± 5% +30.3% 1607 ± 27% sched_debug.cfs_rq:/.tg_load_avg.3
> 1223 ± 5% +22.2% 1495 ± 18% sched_debug.cfs_rq:/.tg_load_avg.30
> 1224 ± 5% +22.0% 1493 ± 19% sched_debug.cfs_rq:/.tg_load_avg.31
> 1234 ± 5% +30.0% 1604 ± 28% sched_debug.cfs_rq:/.tg_load_avg.4
> 1233 ± 5% +30.0% 1604 ± 28% sched_debug.cfs_rq:/.tg_load_avg.5
> 1231 ± 5% +30.3% 1604 ± 28% sched_debug.cfs_rq:/.tg_load_avg.6
> 1233 ± 5% +30.0% 1603 ± 27% sched_debug.cfs_rq:/.tg_load_avg.7
> 1231 ± 5% +30.1% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.8
> 1228 ± 5% +30.3% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.9
> 1228 ± 5% +27.8% 1569 ± 24% sched_debug.cfs_rq:/.tg_load_avg.avg
> 1246 ± 5% +30.7% 1628 ± 27% sched_debug.cfs_rq:/.tg_load_avg.max
> 1212 ± 5% +22.2% 1481 ± 19% sched_debug.cfs_rq:/.tg_load_avg.min
> 15.00 ± 9% +198.3% 44.75 ± 72% sched_debug.cfs_rq:/.tg_load_avg_contrib.21
> 25.00 ± 29% +574.0% 168.50 ± 78% sched_debug.cfs_rq:/.tg_load_avg_contrib.9
> 38.53 ± 5% +29.0% 49.71 ± 26% sched_debug.cfs_rq:/.tg_load_avg_contrib.avg
> 63.34 ± 10% +44.1% 91.30 ± 36% sched_debug.cfs_rq:/.tg_load_avg_contrib.stddev
> 532.25 ± 2% +8.5% 577.50 ± 6% sched_debug.cfs_rq:/.util_avg.15
> 210.75 ± 14% -14.4% 180.50 ± 4% sched_debug.cfs_rq:/.util_avg.29
> 450.00 ± 22% +50.7% 678.00 ± 18% sched_debug.cfs_rq:/.util_avg.9
> 955572 ± 4% -10.2% 857813 ± 5% sched_debug.cpu.avg_idle.6
> 23.99 ± 60% -76.2% 5.71 ± 24% sched_debug.cpu.clock.stddev
> 23.99 ± 60% -76.2% 5.71 ± 24% sched_debug.cpu.clock_task.stddev
> 2840 ± 37% -47.4% 1492 ± 65% sched_debug.cpu.curr->pid.25
> 34.75 ±148% +132.4% 80.75 ± 82% sched_debug.cpu.load.8
> 61776 ± 7% -7.1% 57380 ± 0% sched_debug.cpu.nr_load_updates.25
> 6543 ± 2% +20.4% 7879 ± 9% sched_debug.cpu.nr_switches.0
> 5256 ± 23% +177.1% 14566 ± 52% sched_debug.cpu.nr_switches.27
> 7915 ± 3% +8.7% 8605 ± 3% sched_debug.cpu.nr_switches.avg
> -0.25 ±-519% +1900.0% -5.00 ±-24% sched_debug.cpu.nr_uninterruptible.12
> 2.00 ± 93% -125.0% -0.50 ±-300% sched_debug.cpu.nr_uninterruptible.24
> 17468 ± 14% +194.3% 51413 ± 75% sched_debug.cpu.sched_count.15
> 2112 ± 2% +20.8% 2552 ± 11% sched_debug.cpu.sched_goidle.0
> 2103 ± 34% +219.0% 6709 ± 55% sched_debug.cpu.sched_goidle.27
> 3159 ± 3% +8.2% 3418 ± 4% sched_debug.cpu.sched_goidle.avg
> 1323 ± 64% -72.7% 361.50 ± 15% sched_debug.cpu.ttwu_count.23
> 3264 ± 12% +94.4% 6347 ± 41% sched_debug.cpu.ttwu_count.27
> 3860 ± 3% +9.0% 4208 ± 3% sched_debug.cpu.ttwu_count.avg
> 2358 ± 3% +28.7% 3035 ± 9% sched_debug.cpu.ttwu_local.0
> 1110 ± 22% +54.6% 1716 ± 28% sched_debug.cpu.ttwu_local.27
> 1814 ± 8% +16.1% 2106 ± 5% sched_debug.cpu.ttwu_local.stddev
>
>
> lkp-snb01: Sandy Bridge-EP
> Memory: 32G
>
> will-it-scale.per_thread_ops
>
> 1.2e+06 ++---------------------------------------------------------------+
> | O |
> 1.15e+06 O+O O O O O O O O |
> 1.1e+06 ++ |
> | O O O O O OO |
> 1.05e+06 ++ O O |
> 1e+06 ++ |
> | |
> 950000 ++ |
> 900000 ++ |
> | |
> 850000 ++ |
> 800000 *+*.*.*.*.*.*.*.*.*.*.*.*. .*.*. *.*.*.*.*.*.*.*.*.*.*.*.*.*.*.*.*
> | * * |
> 750000 ++---------------------------------------------------------------+
>
>
> will-it-scale.time.user_time
>
> 50 ++---------------------------------------------------------------------+
> | |
> 45 ++ O O O O O O |
> O O O O |
> | O O O O O |
> 40 ++ O O O O |
> | |
> 35 ++ |
> | |
> 30 ++ |
> | |
> | * |
> 25 ++ + + |
> *.*.*.*..*.* *.*.*..*.*.*.*.*.*.*..*.*.*.*.*.*.*..*.*.*.*.*.*..*.*.*.*
> 20 ++---------------------------------------------------------------------+
>
>
> [*] bisect-good sample
> [O] bisect-bad sample
>
> To reproduce:
>
> git clone git://git.kernel.org/pub/scm/linux/kernel/git/wfg/lkp-tests.git
> cd lkp-tests
> bin/lkp install job.yaml # job file is attached in this email
> bin/lkp run 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,
> Ying Huang
Thanks...
Huh...I'm stumped on this one. If anything I would have expected better
performance with this patch since we don't even take the file_lock or
do the fcheck in the F_UNLCK codepath now, or when there is an error.
I'll see if I can reproduce it on my own test rig, but I'd welcome
ideas of where and how this performance regression could have crept in.
--
Jeff Layton <jeff.layton@primarydata.com>
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [LKP] [lkp] [locks] 7f3697e24d: +35.1% will-it-scale.per_thread_ops
2016-01-29 2:38 ` Jeff Layton
@ 2016-01-29 2:52 ` Huang, Ying
2016-01-29 12:13 ` Jeff Layton
2016-02-01 13:39 ` J. Bruce Fields
0 siblings, 2 replies; 5+ messages in thread
From: Huang, Ying @ 2016-01-29 2:52 UTC (permalink / raw)
To: Jeff Layton; +Cc: J. Bruce Fields, lkp, LKML, Dmitry Vyukov, Alexander Viro
Jeff Layton <jeff.layton@primarydata.com> writes:
> On Fri, 29 Jan 2016 09:32:19 +0800
> kernel test robot <ying.huang@linux.intel.com> wrote:
>
>> FYI, we noticed the below changes on
>>
>> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
>> commit 7f3697e24dc3820b10f445a4a7d914fc356012d1 ("locks: fix unlock when fcntl_setlk races with a close")
>>
>>
>> =========================================================================================
>> compiler/cpufreq_governor/kconfig/rootfs/tbox_group/test/testcase:
>> gcc-4.9/performance/x86_64-rhel/debian-x86_64-2015-02-07.cgz/lkp-snb01/lock1/will-it-scale
>>
>> commit:
>> 9189922675ecca0fab38931d86b676e9d79602dc
>> 7f3697e24dc3820b10f445a4a7d914fc356012d1
>>
>> 9189922675ecca0f 7f3697e24dc3820b10f445a4a7
>> ---------------- --------------------------
>> %stddev %change %stddev
>> \ | \
>> 2376432 ± 0% +2.1% 2427484 ± 0% will-it-scale.per_process_ops
>> 807889 ± 0% +35.1% 1091496 ± 4% will-it-scale.per_thread_ops
>> 22.08 ± 2% +89.1% 41.75 ± 5% will-it-scale.time.user_time
>> 1238371 ± 14% +100.4% 2481345 ± 39% cpuidle.C1E-SNB.time
>> 3098 ± 57% -66.6% 1035 ±171% numa-numastat.node1.other_node
>> 379.25 ± 8% -21.4% 298.00 ± 12% numa-vmstat.node0.nr_alloc_batch
>> 22.08 ± 2% +89.1% 41.75 ± 5% time.user_time
>> 1795 ± 4% +7.5% 1930 ± 2% vmstat.system.cs
>> 0.54 ± 5% +136.9% 1.28 ± 10% perf-profile.cycles.___might_sleep.__might_sleep.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file
>> 1.65 ± 57% +245.2% 5.70 ± 29% perf-profile.cycles.__fdget_raw.sys_fcntl.entry_SYSCALL_64_fastpath
>> 1.58 ± 59% +248.3% 5.50 ± 31% perf-profile.cycles.__fget.__fget_light.__fdget_raw.sys_fcntl.entry_SYSCALL_64_fastpath
>> 1.62 ± 58% +246.3% 5.63 ± 30% perf-profile.cycles.__fget_light.__fdget_raw.sys_fcntl.entry_SYSCALL_64_fastpath
>> 0.00 ± -1% +Inf% 5.88 ± 11% perf-profile.cycles.__memset.locks_alloc_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait
>> 2.50 ± 2% -100.0% 0.00 ± -1% perf-profile.cycles.__memset.locks_alloc_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk
>> 1.29 ± 4% +138.8% 3.09 ± 11% perf-profile.cycles.__memset.locks_alloc_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
>> 0.47 ± 9% +144.4% 1.16 ± 11% perf-profile.cycles.__might_fault.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
>> 0.37 ± 12% +140.3% 0.90 ± 9% perf-profile.cycles.__might_sleep.__might_fault.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
>> 0.86 ± 6% +137.7% 2.05 ± 10% perf-profile.cycles.__might_sleep.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file.vfs_lock_file
>> 0.61 ± 14% +56.8% 0.95 ± 14% perf-profile.cycles.__might_sleep.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.sys_fcntl
>> 0.00 ± -1% +Inf% 39.84 ± 12% perf-profile.cycles.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk.sys_fcntl
>> 16.44 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
>> 0.00 ± -1% +Inf% 1.77 ± 11% perf-profile.cycles._raw_spin_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
>> 59.34 ± 1% -72.4% 16.36 ± 33% perf-profile.cycles._raw_spin_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
>> 0.46 ± 11% +144.9% 1.13 ± 19% perf-profile.cycles.avc_has_perm.inode_has_perm.file_has_perm.selinux_file_fcntl.security_file_fcntl
>> 0.87 ± 6% +103.2% 1.77 ± 12% perf-profile.cycles.avc_has_perm.inode_has_perm.file_has_perm.selinux_file_lock.security_file_lock
>> 0.81 ± 4% +135.7% 1.90 ± 10% perf-profile.cycles.copy_user_generic_string.sys_fcntl.entry_SYSCALL_64_fastpath
>> 0.00 ± -1% +Inf% 41.86 ± 12% perf-profile.cycles.do_lock_file_wait.part.29.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
>> 0.88 ± 6% +127.8% 2.00 ± 9% perf-profile.cycles.entry_SYSCALL_64
>> 0.86 ± 4% +122.6% 1.92 ± 12% perf-profile.cycles.entry_SYSCALL_64_after_swapgs
>> 84.98 ± 0% -9.1% 77.20 ± 2% perf-profile.cycles.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
>> 0.76 ± 10% +142.1% 1.84 ± 14% perf-profile.cycles.file_has_perm.selinux_file_fcntl.security_file_fcntl.sys_fcntl.entry_SYSCALL_64_fastpath
>> 1.35 ± 4% +106.3% 2.78 ± 11% perf-profile.cycles.file_has_perm.selinux_file_lock.security_file_lock.fcntl_setlk.sys_fcntl
>> 0.00 ± -1% +Inf% 0.89 ± 12% perf-profile.cycles.flock_to_posix_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
>> 6.90 ± 4% -48.6% 3.55 ± 27% perf-profile.cycles.fput.entry_SYSCALL_64_fastpath
>> 0.51 ± 10% +140.5% 1.23 ± 16% perf-profile.cycles.inode_has_perm.isra.31.file_has_perm.selinux_file_fcntl.security_file_fcntl.sys_fcntl
>> 0.98 ± 4% +97.7% 1.93 ± 11% perf-profile.cycles.inode_has_perm.isra.31.file_has_perm.selinux_file_lock.security_file_lock.fcntl_setlk
>> 0.00 ± -1% +Inf% 6.56 ± 10% perf-profile.cycles.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait
>> 2.75 ± 4% -100.0% 0.00 ± -1% perf-profile.cycles.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk
>> 1.53 ± 7% +119.7% 3.37 ± 13% perf-profile.cycles.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
>> 0.00 ± -1% +Inf% 1.79 ± 11% perf-profile.cycles.kmem_cache_free.locks_free_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait
>> 0.46 ± 14% +257.0% 1.66 ± 11% perf-profile.cycles.kmem_cache_free.locks_free_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
>> 0.40 ± 7% +158.6% 1.05 ± 17% perf-profile.cycles.kmem_cache_free.locks_free_lock.locks_dispose_list.__posix_lock_file.vfs_lock_file
>> 0.00 ± -1% +Inf% 0.96 ± 10% perf-profile.cycles.lg_local_lock.locks_insert_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait
>> 0.00 ± -1% +Inf% 14.69 ± 10% perf-profile.cycles.locks_alloc_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
>> 6.38 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.locks_alloc_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
>> 3.28 ± 6% +127.1% 7.45 ± 12% perf-profile.cycles.locks_alloc_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
>> 0.00 ± -1% +Inf% 9.75 ± 13% perf-profile.cycles.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
>> 3.61 ± 1% -100.0% 0.00 ± -1% perf-profile.cycles.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
>> 0.00 ± -1% +Inf% 1.84 ± 11% perf-profile.cycles.locks_dispose_list.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
>> 0.00 ± -1% +Inf% 2.42 ± 10% perf-profile.cycles.locks_free_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
>> 1.00 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.locks_free_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
>> 0.63 ± 11% +224.1% 2.05 ± 10% perf-profile.cycles.locks_free_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
>> 0.00 ± -1% +Inf% 1.22 ± 14% perf-profile.cycles.locks_free_lock.locks_dispose_list.__posix_lock_file.vfs_lock_file.do_lock_file_wait
>> 0.00 ± -1% +Inf% 6.17 ± 15% perf-profile.cycles.locks_insert_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
>> 2.31 ± 6% -100.0% 0.00 ± -1% perf-profile.cycles.locks_insert_lock_ctx.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
>> 0.00 ± -1% +Inf% 8.96 ± 13% perf-profile.cycles.locks_unlink_lock_ctx.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait
>> 3.27 ± 1% -100.0% 0.00 ± -1% perf-profile.cycles.locks_unlink_lock_ctx.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.fcntl_setlk
>> 53.88 ± 1% -79.7% 10.92 ± 46% perf-profile.cycles.native_queued_spin_lock_slowpath._raw_spin_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
>> 2.75 ± 0% +183.3% 7.79 ± 13% perf-profile.cycles.put_pid.locks_unlink_lock_ctx.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file
>> 1.11 ± 9% +137.2% 2.63 ± 14% perf-profile.cycles.security_file_fcntl.sys_fcntl.entry_SYSCALL_64_fastpath
>> 1.69 ± 4% +118.2% 3.69 ± 11% perf-profile.cycles.security_file_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
>> 0.91 ± 9% +139.0% 2.17 ± 14% perf-profile.cycles.selinux_file_fcntl.security_file_fcntl.sys_fcntl.entry_SYSCALL_64_fastpath
>> 1.39 ± 4% +114.6% 2.97 ± 10% perf-profile.cycles.selinux_file_lock.security_file_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
>> 0.00 ± -1% +Inf% 41.12 ± 12% perf-profile.cycles.vfs_lock_file.do_lock_file_wait.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
>> 17.04 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.vfs_lock_file.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
>> 34.75 ±148% +132.4% 80.75 ± 82% sched_debug.cfs_rq:/.load.8
>> 15.00 ± 9% +198.3% 44.75 ± 72% sched_debug.cfs_rq:/.load_avg.21
>> 25.00 ± 29% +574.0% 168.50 ± 78% sched_debug.cfs_rq:/.load_avg.9
>> 38.47 ± 5% +29.1% 49.65 ± 26% sched_debug.cfs_rq:/.load_avg.avg
>> 63.17 ± 10% +44.3% 91.16 ± 36% sched_debug.cfs_rq:/.load_avg.stddev
>> 893865 ± 12% -12.5% 782455 ± 0% sched_debug.cfs_rq:/.min_vruntime.25
>> 18.25 ± 26% +52.1% 27.75 ± 25% sched_debug.cfs_rq:/.runnable_load_avg.9
>> -57635 ±-68% -196.4% 55548 ±130% sched_debug.cfs_rq:/.spread0.1
>> -802264 ±-25% -29.5% -565458 ±-49% sched_debug.cfs_rq:/.spread0.8
>> -804662 ±-25% -29.4% -567811 ±-48% sched_debug.cfs_rq:/.spread0.min
>> 1233 ± 5% +30.9% 1614 ± 28% sched_debug.cfs_rq:/.tg_load_avg.0
>> 1233 ± 5% +30.9% 1614 ± 28% sched_debug.cfs_rq:/.tg_load_avg.1
>> 1228 ± 5% +30.3% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.10
>> 1228 ± 5% +30.4% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.11
>> 1228 ± 5% +30.3% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.12
>> 1229 ± 5% +30.0% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.13
>> 1228 ± 5% +30.1% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.14
>> 1229 ± 5% +30.0% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.15
>> 1226 ± 5% +30.3% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.16
>> 1226 ± 5% +30.2% 1597 ± 27% sched_debug.cfs_rq:/.tg_load_avg.17
>> 1227 ± 5% +30.1% 1595 ± 27% sched_debug.cfs_rq:/.tg_load_avg.18
>> 1227 ± 5% +29.4% 1588 ± 26% sched_debug.cfs_rq:/.tg_load_avg.19
>> 1233 ± 5% +30.4% 1609 ± 27% sched_debug.cfs_rq:/.tg_load_avg.2
>> 1222 ± 5% +29.9% 1587 ± 26% sched_debug.cfs_rq:/.tg_load_avg.20
>> 1223 ± 5% +24.2% 1519 ± 20% sched_debug.cfs_rq:/.tg_load_avg.21
>> 1223 ± 5% +23.8% 1515 ± 20% sched_debug.cfs_rq:/.tg_load_avg.22
>> 1223 ± 5% +23.9% 1515 ± 20% sched_debug.cfs_rq:/.tg_load_avg.23
>> 1223 ± 5% +23.9% 1515 ± 20% sched_debug.cfs_rq:/.tg_load_avg.24
>> 1223 ± 5% +23.5% 1511 ± 19% sched_debug.cfs_rq:/.tg_load_avg.25
>> 1224 ± 5% +23.5% 1512 ± 19% sched_debug.cfs_rq:/.tg_load_avg.26
>> 1223 ± 5% +23.1% 1506 ± 19% sched_debug.cfs_rq:/.tg_load_avg.27
>> 1223 ± 5% +22.5% 1499 ± 19% sched_debug.cfs_rq:/.tg_load_avg.28
>> 1224 ± 5% +22.5% 1499 ± 19% sched_debug.cfs_rq:/.tg_load_avg.29
>> 1233 ± 5% +30.3% 1607 ± 27% sched_debug.cfs_rq:/.tg_load_avg.3
>> 1223 ± 5% +22.2% 1495 ± 18% sched_debug.cfs_rq:/.tg_load_avg.30
>> 1224 ± 5% +22.0% 1493 ± 19% sched_debug.cfs_rq:/.tg_load_avg.31
>> 1234 ± 5% +30.0% 1604 ± 28% sched_debug.cfs_rq:/.tg_load_avg.4
>> 1233 ± 5% +30.0% 1604 ± 28% sched_debug.cfs_rq:/.tg_load_avg.5
>> 1231 ± 5% +30.3% 1604 ± 28% sched_debug.cfs_rq:/.tg_load_avg.6
>> 1233 ± 5% +30.0% 1603 ± 27% sched_debug.cfs_rq:/.tg_load_avg.7
>> 1231 ± 5% +30.1% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.8
>> 1228 ± 5% +30.3% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.9
>> 1228 ± 5% +27.8% 1569 ± 24% sched_debug.cfs_rq:/.tg_load_avg.avg
>> 1246 ± 5% +30.7% 1628 ± 27% sched_debug.cfs_rq:/.tg_load_avg.max
>> 1212 ± 5% +22.2% 1481 ± 19% sched_debug.cfs_rq:/.tg_load_avg.min
>> 15.00 ± 9% +198.3% 44.75 ± 72% sched_debug.cfs_rq:/.tg_load_avg_contrib.21
>> 25.00 ± 29% +574.0% 168.50 ± 78% sched_debug.cfs_rq:/.tg_load_avg_contrib.9
>> 38.53 ± 5% +29.0% 49.71 ± 26% sched_debug.cfs_rq:/.tg_load_avg_contrib.avg
>> 63.34 ± 10% +44.1% 91.30 ± 36% sched_debug.cfs_rq:/.tg_load_avg_contrib.stddev
>> 532.25 ± 2% +8.5% 577.50 ± 6% sched_debug.cfs_rq:/.util_avg.15
>> 210.75 ± 14% -14.4% 180.50 ± 4% sched_debug.cfs_rq:/.util_avg.29
>> 450.00 ± 22% +50.7% 678.00 ± 18% sched_debug.cfs_rq:/.util_avg.9
>> 955572 ± 4% -10.2% 857813 ± 5% sched_debug.cpu.avg_idle.6
>> 23.99 ± 60% -76.2% 5.71 ± 24% sched_debug.cpu.clock.stddev
>> 23.99 ± 60% -76.2% 5.71 ± 24% sched_debug.cpu.clock_task.stddev
>> 2840 ± 37% -47.4% 1492 ± 65% sched_debug.cpu.curr->pid.25
>> 34.75 ±148% +132.4% 80.75 ± 82% sched_debug.cpu.load.8
>> 61776 ± 7% -7.1% 57380 ± 0% sched_debug.cpu.nr_load_updates.25
>> 6543 ± 2% +20.4% 7879 ± 9% sched_debug.cpu.nr_switches.0
>> 5256 ± 23% +177.1% 14566 ± 52% sched_debug.cpu.nr_switches.27
>> 7915 ± 3% +8.7% 8605 ± 3% sched_debug.cpu.nr_switches.avg
>> -0.25 ±-519% +1900.0% -5.00 ±-24% sched_debug.cpu.nr_uninterruptible.12
>> 2.00 ± 93% -125.0% -0.50 ±-300% sched_debug.cpu.nr_uninterruptible.24
>> 17468 ± 14% +194.3% 51413 ± 75% sched_debug.cpu.sched_count.15
>> 2112 ± 2% +20.8% 2552 ± 11% sched_debug.cpu.sched_goidle.0
>> 2103 ± 34% +219.0% 6709 ± 55% sched_debug.cpu.sched_goidle.27
>> 3159 ± 3% +8.2% 3418 ± 4% sched_debug.cpu.sched_goidle.avg
>> 1323 ± 64% -72.7% 361.50 ± 15% sched_debug.cpu.ttwu_count.23
>> 3264 ± 12% +94.4% 6347 ± 41% sched_debug.cpu.ttwu_count.27
>> 3860 ± 3% +9.0% 4208 ± 3% sched_debug.cpu.ttwu_count.avg
>> 2358 ± 3% +28.7% 3035 ± 9% sched_debug.cpu.ttwu_local.0
>> 1110 ± 22% +54.6% 1716 ± 28% sched_debug.cpu.ttwu_local.27
>> 1814 ± 8% +16.1% 2106 ± 5% sched_debug.cpu.ttwu_local.stddev
>>
>>
>> lkp-snb01: Sandy Bridge-EP
>> Memory: 32G
>>
>> will-it-scale.per_thread_ops
>>
>> 1.2e+06 ++---------------------------------------------------------------+
>> | O |
>> 1.15e+06 O+O O O O O O O O |
>> 1.1e+06 ++ |
>> | O O O O O OO |
>> 1.05e+06 ++ O O |
>> 1e+06 ++ |
>> | |
>> 950000 ++ |
>> 900000 ++ |
>> | |
>> 850000 ++ |
>> 800000 *+*.*.*.*.*.*.*.*.*.*.*.*. .*.*. *.*.*.*.*.*.*.*.*.*.*.*.*.*.*.*.*
>> | * * |
>> 750000 ++---------------------------------------------------------------+
>>
>>
>> will-it-scale.time.user_time
>>
>> 50 ++---------------------------------------------------------------------+
>> | |
>> 45 ++ O O O O O O |
>> O O O O |
>> | O O O O O |
>> 40 ++ O O O O |
>> | |
>> 35 ++ |
>> | |
>> 30 ++ |
>> | |
>> | * |
>> 25 ++ + + |
>> *.*.*.*..*.* *.*.*..*.*.*.*.*.*.*..*.*.*.*.*.*.*..*.*.*.*.*.*..*.*.*.*
>> 20 ++---------------------------------------------------------------------+
>>
>>
>> [*] bisect-good sample
>> [O] bisect-bad sample
>>
>> To reproduce:
>>
>> git clone git://git.kernel.org/pub/scm/linux/kernel/git/wfg/lkp-tests.git
>> cd lkp-tests
>> bin/lkp install job.yaml # job file is attached in this email
>> bin/lkp run 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,
>> Ying Huang
>
> Thanks...
>
> Huh...I'm stumped on this one. If anything I would have expected better
> performance with this patch since we don't even take the file_lock or
> do the fcheck in the F_UNLCK codepath now, or when there is an error.
>
> I'll see if I can reproduce it on my own test rig, but I'd welcome
> ideas of where and how this performance regression could have crept in.
This is a performance increase instead of performance regression.
Best Regards,
Huang, Ying
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [LKP] [lkp] [locks] 7f3697e24d: +35.1% will-it-scale.per_thread_ops
2016-01-29 2:52 ` [LKP] " Huang, Ying
@ 2016-01-29 12:13 ` Jeff Layton
2016-02-01 13:39 ` J. Bruce Fields
1 sibling, 0 replies; 5+ messages in thread
From: Jeff Layton @ 2016-01-29 12:13 UTC (permalink / raw)
To: Huang, Ying; +Cc: J. Bruce Fields, lkp, LKML, Dmitry Vyukov, Alexander Viro
On Fri, 29 Jan 2016 10:52:20 +0800
"Huang\, Ying" <ying.huang@intel.com> wrote:
> Jeff Layton <jeff.layton@primarydata.com> writes:
>
> > On Fri, 29 Jan 2016 09:32:19 +0800
> > kernel test robot <ying.huang@linux.intel.com> wrote:
> >
> >> FYI, we noticed the below changes on
> >>
> >> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
> >> commit 7f3697e24dc3820b10f445a4a7d914fc356012d1 ("locks: fix unlock when fcntl_setlk races with a close")
> >>
> >>
> >> =========================================================================================
> >> compiler/cpufreq_governor/kconfig/rootfs/tbox_group/test/testcase:
> >> gcc-4.9/performance/x86_64-rhel/debian-x86_64-2015-02-07.cgz/lkp-snb01/lock1/will-it-scale
> >>
> >> commit:
> >> 9189922675ecca0fab38931d86b676e9d79602dc
> >> 7f3697e24dc3820b10f445a4a7d914fc356012d1
> >>
> >> 9189922675ecca0f 7f3697e24dc3820b10f445a4a7
> >> ---------------- --------------------------
> >> %stddev %change %stddev
> >> \ | \
> >> 2376432 ± 0% +2.1% 2427484 ± 0% will-it-scale.per_process_ops
> >> 807889 ± 0% +35.1% 1091496 ± 4% will-it-scale.per_thread_ops
> >> 22.08 ± 2% +89.1% 41.75 ± 5% will-it-scale.time.user_time
> >> 1238371 ± 14% +100.4% 2481345 ± 39% cpuidle.C1E-SNB.time
> >> 3098 ± 57% -66.6% 1035 ±171% numa-numastat.node1.other_node
> >> 379.25 ± 8% -21.4% 298.00 ± 12% numa-vmstat.node0.nr_alloc_batch
> >> 22.08 ± 2% +89.1% 41.75 ± 5% time.user_time
> >> 1795 ± 4% +7.5% 1930 ± 2% vmstat.system.cs
> >> 0.54 ± 5% +136.9% 1.28 ± 10% perf-profile.cycles.___might_sleep.__might_sleep.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file
> >> 1.65 ± 57% +245.2% 5.70 ± 29% perf-profile.cycles.__fdget_raw.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 1.58 ± 59% +248.3% 5.50 ± 31% perf-profile.cycles.__fget.__fget_light.__fdget_raw.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 1.62 ± 58% +246.3% 5.63 ± 30% perf-profile.cycles.__fget_light.__fdget_raw.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.00 ± -1% +Inf% 5.88 ± 11% perf-profile.cycles.__memset.locks_alloc_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait
> >> 2.50 ± 2% -100.0% 0.00 ± -1% perf-profile.cycles.__memset.locks_alloc_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk
> >> 1.29 ± 4% +138.8% 3.09 ± 11% perf-profile.cycles.__memset.locks_alloc_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.47 ± 9% +144.4% 1.16 ± 11% perf-profile.cycles.__might_fault.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.37 ± 12% +140.3% 0.90 ± 9% perf-profile.cycles.__might_sleep.__might_fault.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.86 ± 6% +137.7% 2.05 ± 10% perf-profile.cycles.__might_sleep.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file.vfs_lock_file
> >> 0.61 ± 14% +56.8% 0.95 ± 14% perf-profile.cycles.__might_sleep.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.sys_fcntl
> >> 0.00 ± -1% +Inf% 39.84 ± 12% perf-profile.cycles.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk.sys_fcntl
> >> 16.44 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.00 ± -1% +Inf% 1.77 ± 11% perf-profile.cycles._raw_spin_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
> >> 59.34 ± 1% -72.4% 16.36 ± 33% perf-profile.cycles._raw_spin_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.46 ± 11% +144.9% 1.13 ± 19% perf-profile.cycles.avc_has_perm.inode_has_perm.file_has_perm.selinux_file_fcntl.security_file_fcntl
> >> 0.87 ± 6% +103.2% 1.77 ± 12% perf-profile.cycles.avc_has_perm.inode_has_perm.file_has_perm.selinux_file_lock.security_file_lock
> >> 0.81 ± 4% +135.7% 1.90 ± 10% perf-profile.cycles.copy_user_generic_string.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.00 ± -1% +Inf% 41.86 ± 12% perf-profile.cycles.do_lock_file_wait.part.29.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.88 ± 6% +127.8% 2.00 ± 9% perf-profile.cycles.entry_SYSCALL_64
> >> 0.86 ± 4% +122.6% 1.92 ± 12% perf-profile.cycles.entry_SYSCALL_64_after_swapgs
> >> 84.98 ± 0% -9.1% 77.20 ± 2% perf-profile.cycles.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.76 ± 10% +142.1% 1.84 ± 14% perf-profile.cycles.file_has_perm.selinux_file_fcntl.security_file_fcntl.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 1.35 ± 4% +106.3% 2.78 ± 11% perf-profile.cycles.file_has_perm.selinux_file_lock.security_file_lock.fcntl_setlk.sys_fcntl
> >> 0.00 ± -1% +Inf% 0.89 ± 12% perf-profile.cycles.flock_to_posix_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 6.90 ± 4% -48.6% 3.55 ± 27% perf-profile.cycles.fput.entry_SYSCALL_64_fastpath
> >> 0.51 ± 10% +140.5% 1.23 ± 16% perf-profile.cycles.inode_has_perm.isra.31.file_has_perm.selinux_file_fcntl.security_file_fcntl.sys_fcntl
> >> 0.98 ± 4% +97.7% 1.93 ± 11% perf-profile.cycles.inode_has_perm.isra.31.file_has_perm.selinux_file_lock.security_file_lock.fcntl_setlk
> >> 0.00 ± -1% +Inf% 6.56 ± 10% perf-profile.cycles.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait
> >> 2.75 ± 4% -100.0% 0.00 ± -1% perf-profile.cycles.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk
> >> 1.53 ± 7% +119.7% 3.37 ± 13% perf-profile.cycles.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.00 ± -1% +Inf% 1.79 ± 11% perf-profile.cycles.kmem_cache_free.locks_free_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait
> >> 0.46 ± 14% +257.0% 1.66 ± 11% perf-profile.cycles.kmem_cache_free.locks_free_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.40 ± 7% +158.6% 1.05 ± 17% perf-profile.cycles.kmem_cache_free.locks_free_lock.locks_dispose_list.__posix_lock_file.vfs_lock_file
> >> 0.00 ± -1% +Inf% 0.96 ± 10% perf-profile.cycles.lg_local_lock.locks_insert_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait
> >> 0.00 ± -1% +Inf% 14.69 ± 10% perf-profile.cycles.locks_alloc_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
> >> 6.38 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.locks_alloc_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
> >> 3.28 ± 6% +127.1% 7.45 ± 12% perf-profile.cycles.locks_alloc_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.00 ± -1% +Inf% 9.75 ± 13% perf-profile.cycles.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
> >> 3.61 ± 1% -100.0% 0.00 ± -1% perf-profile.cycles.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
> >> 0.00 ± -1% +Inf% 1.84 ± 11% perf-profile.cycles.locks_dispose_list.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
> >> 0.00 ± -1% +Inf% 2.42 ± 10% perf-profile.cycles.locks_free_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
> >> 1.00 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.locks_free_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
> >> 0.63 ± 11% +224.1% 2.05 ± 10% perf-profile.cycles.locks_free_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.00 ± -1% +Inf% 1.22 ± 14% perf-profile.cycles.locks_free_lock.locks_dispose_list.__posix_lock_file.vfs_lock_file.do_lock_file_wait
> >> 0.00 ± -1% +Inf% 6.17 ± 15% perf-profile.cycles.locks_insert_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
> >> 2.31 ± 6% -100.0% 0.00 ± -1% perf-profile.cycles.locks_insert_lock_ctx.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
> >> 0.00 ± -1% +Inf% 8.96 ± 13% perf-profile.cycles.locks_unlink_lock_ctx.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait
> >> 3.27 ± 1% -100.0% 0.00 ± -1% perf-profile.cycles.locks_unlink_lock_ctx.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.fcntl_setlk
> >> 53.88 ± 1% -79.7% 10.92 ± 46% perf-profile.cycles.native_queued_spin_lock_slowpath._raw_spin_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 2.75 ± 0% +183.3% 7.79 ± 13% perf-profile.cycles.put_pid.locks_unlink_lock_ctx.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file
> >> 1.11 ± 9% +137.2% 2.63 ± 14% perf-profile.cycles.security_file_fcntl.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 1.69 ± 4% +118.2% 3.69 ± 11% perf-profile.cycles.security_file_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.91 ± 9% +139.0% 2.17 ± 14% perf-profile.cycles.selinux_file_fcntl.security_file_fcntl.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 1.39 ± 4% +114.6% 2.97 ± 10% perf-profile.cycles.selinux_file_lock.security_file_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.00 ± -1% +Inf% 41.12 ± 12% perf-profile.cycles.vfs_lock_file.do_lock_file_wait.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 17.04 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.vfs_lock_file.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 34.75 ±148% +132.4% 80.75 ± 82% sched_debug.cfs_rq:/.load.8
> >> 15.00 ± 9% +198.3% 44.75 ± 72% sched_debug.cfs_rq:/.load_avg.21
> >> 25.00 ± 29% +574.0% 168.50 ± 78% sched_debug.cfs_rq:/.load_avg.9
> >> 38.47 ± 5% +29.1% 49.65 ± 26% sched_debug.cfs_rq:/.load_avg.avg
> >> 63.17 ± 10% +44.3% 91.16 ± 36% sched_debug.cfs_rq:/.load_avg.stddev
> >> 893865 ± 12% -12.5% 782455 ± 0% sched_debug.cfs_rq:/.min_vruntime.25
> >> 18.25 ± 26% +52.1% 27.75 ± 25% sched_debug.cfs_rq:/.runnable_load_avg.9
> >> -57635 ±-68% -196.4% 55548 ±130% sched_debug.cfs_rq:/.spread0.1
> >> -802264 ±-25% -29.5% -565458 ±-49% sched_debug.cfs_rq:/.spread0.8
> >> -804662 ±-25% -29.4% -567811 ±-48% sched_debug.cfs_rq:/.spread0.min
> >> 1233 ± 5% +30.9% 1614 ± 28% sched_debug.cfs_rq:/.tg_load_avg.0
> >> 1233 ± 5% +30.9% 1614 ± 28% sched_debug.cfs_rq:/.tg_load_avg.1
> >> 1228 ± 5% +30.3% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.10
> >> 1228 ± 5% +30.4% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.11
> >> 1228 ± 5% +30.3% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.12
> >> 1229 ± 5% +30.0% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.13
> >> 1228 ± 5% +30.1% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.14
> >> 1229 ± 5% +30.0% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.15
> >> 1226 ± 5% +30.3% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.16
> >> 1226 ± 5% +30.2% 1597 ± 27% sched_debug.cfs_rq:/.tg_load_avg.17
> >> 1227 ± 5% +30.1% 1595 ± 27% sched_debug.cfs_rq:/.tg_load_avg.18
> >> 1227 ± 5% +29.4% 1588 ± 26% sched_debug.cfs_rq:/.tg_load_avg.19
> >> 1233 ± 5% +30.4% 1609 ± 27% sched_debug.cfs_rq:/.tg_load_avg.2
> >> 1222 ± 5% +29.9% 1587 ± 26% sched_debug.cfs_rq:/.tg_load_avg.20
> >> 1223 ± 5% +24.2% 1519 ± 20% sched_debug.cfs_rq:/.tg_load_avg.21
> >> 1223 ± 5% +23.8% 1515 ± 20% sched_debug.cfs_rq:/.tg_load_avg.22
> >> 1223 ± 5% +23.9% 1515 ± 20% sched_debug.cfs_rq:/.tg_load_avg.23
> >> 1223 ± 5% +23.9% 1515 ± 20% sched_debug.cfs_rq:/.tg_load_avg.24
> >> 1223 ± 5% +23.5% 1511 ± 19% sched_debug.cfs_rq:/.tg_load_avg.25
> >> 1224 ± 5% +23.5% 1512 ± 19% sched_debug.cfs_rq:/.tg_load_avg.26
> >> 1223 ± 5% +23.1% 1506 ± 19% sched_debug.cfs_rq:/.tg_load_avg.27
> >> 1223 ± 5% +22.5% 1499 ± 19% sched_debug.cfs_rq:/.tg_load_avg.28
> >> 1224 ± 5% +22.5% 1499 ± 19% sched_debug.cfs_rq:/.tg_load_avg.29
> >> 1233 ± 5% +30.3% 1607 ± 27% sched_debug.cfs_rq:/.tg_load_avg.3
> >> 1223 ± 5% +22.2% 1495 ± 18% sched_debug.cfs_rq:/.tg_load_avg.30
> >> 1224 ± 5% +22.0% 1493 ± 19% sched_debug.cfs_rq:/.tg_load_avg.31
> >> 1234 ± 5% +30.0% 1604 ± 28% sched_debug.cfs_rq:/.tg_load_avg.4
> >> 1233 ± 5% +30.0% 1604 ± 28% sched_debug.cfs_rq:/.tg_load_avg.5
> >> 1231 ± 5% +30.3% 1604 ± 28% sched_debug.cfs_rq:/.tg_load_avg.6
> >> 1233 ± 5% +30.0% 1603 ± 27% sched_debug.cfs_rq:/.tg_load_avg.7
> >> 1231 ± 5% +30.1% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.8
> >> 1228 ± 5% +30.3% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.9
> >> 1228 ± 5% +27.8% 1569 ± 24% sched_debug.cfs_rq:/.tg_load_avg.avg
> >> 1246 ± 5% +30.7% 1628 ± 27% sched_debug.cfs_rq:/.tg_load_avg.max
> >> 1212 ± 5% +22.2% 1481 ± 19% sched_debug.cfs_rq:/.tg_load_avg.min
> >> 15.00 ± 9% +198.3% 44.75 ± 72% sched_debug.cfs_rq:/.tg_load_avg_contrib.21
> >> 25.00 ± 29% +574.0% 168.50 ± 78% sched_debug.cfs_rq:/.tg_load_avg_contrib.9
> >> 38.53 ± 5% +29.0% 49.71 ± 26% sched_debug.cfs_rq:/.tg_load_avg_contrib.avg
> >> 63.34 ± 10% +44.1% 91.30 ± 36% sched_debug.cfs_rq:/.tg_load_avg_contrib.stddev
> >> 532.25 ± 2% +8.5% 577.50 ± 6% sched_debug.cfs_rq:/.util_avg.15
> >> 210.75 ± 14% -14.4% 180.50 ± 4% sched_debug.cfs_rq:/.util_avg.29
> >> 450.00 ± 22% +50.7% 678.00 ± 18% sched_debug.cfs_rq:/.util_avg.9
> >> 955572 ± 4% -10.2% 857813 ± 5% sched_debug.cpu.avg_idle.6
> >> 23.99 ± 60% -76.2% 5.71 ± 24% sched_debug.cpu.clock.stddev
> >> 23.99 ± 60% -76.2% 5.71 ± 24% sched_debug.cpu.clock_task.stddev
> >> 2840 ± 37% -47.4% 1492 ± 65% sched_debug.cpu.curr->pid.25
> >> 34.75 ±148% +132.4% 80.75 ± 82% sched_debug.cpu.load.8
> >> 61776 ± 7% -7.1% 57380 ± 0% sched_debug.cpu.nr_load_updates.25
> >> 6543 ± 2% +20.4% 7879 ± 9% sched_debug.cpu.nr_switches.0
> >> 5256 ± 23% +177.1% 14566 ± 52% sched_debug.cpu.nr_switches.27
> >> 7915 ± 3% +8.7% 8605 ± 3% sched_debug.cpu.nr_switches.avg
> >> -0.25 ±-519% +1900.0% -5.00 ±-24% sched_debug.cpu.nr_uninterruptible.12
> >> 2.00 ± 93% -125.0% -0.50 ±-300% sched_debug.cpu.nr_uninterruptible.24
> >> 17468 ± 14% +194.3% 51413 ± 75% sched_debug.cpu.sched_count.15
> >> 2112 ± 2% +20.8% 2552 ± 11% sched_debug.cpu.sched_goidle.0
> >> 2103 ± 34% +219.0% 6709 ± 55% sched_debug.cpu.sched_goidle.27
> >> 3159 ± 3% +8.2% 3418 ± 4% sched_debug.cpu.sched_goidle.avg
> >> 1323 ± 64% -72.7% 361.50 ± 15% sched_debug.cpu.ttwu_count.23
> >> 3264 ± 12% +94.4% 6347 ± 41% sched_debug.cpu.ttwu_count.27
> >> 3860 ± 3% +9.0% 4208 ± 3% sched_debug.cpu.ttwu_count.avg
> >> 2358 ± 3% +28.7% 3035 ± 9% sched_debug.cpu.ttwu_local.0
> >> 1110 ± 22% +54.6% 1716 ± 28% sched_debug.cpu.ttwu_local.27
> >> 1814 ± 8% +16.1% 2106 ± 5% sched_debug.cpu.ttwu_local.stddev
> >>
> >>
> >> lkp-snb01: Sandy Bridge-EP
> >> Memory: 32G
> >>
> >> will-it-scale.per_thread_ops
> >>
> >> 1.2e+06 ++---------------------------------------------------------------+
> >> | O |
> >> 1.15e+06 O+O O O O O O O O |
> >> 1.1e+06 ++ |
> >> | O O O O O OO |
> >> 1.05e+06 ++ O O |
> >> 1e+06 ++ |
> >> | |
> >> 950000 ++ |
> >> 900000 ++ |
> >> | |
> >> 850000 ++ |
> >> 800000 *+*.*.*.*.*.*.*.*.*.*.*.*. .*.*. *.*.*.*.*.*.*.*.*.*.*.*.*.*.*.*.*
> >> | * * |
> >> 750000 ++---------------------------------------------------------------+
> >>
> >>
> >> will-it-scale.time.user_time
> >>
> >> 50 ++---------------------------------------------------------------------+
> >> | |
> >> 45 ++ O O O O O O |
> >> O O O O |
> >> | O O O O O |
> >> 40 ++ O O O O |
> >> | |
> >> 35 ++ |
> >> | |
> >> 30 ++ |
> >> | |
> >> | * |
> >> 25 ++ + + |
> >> *.*.*.*..*.* *.*.*..*.*.*.*.*.*.*..*.*.*.*.*.*.*..*.*.*.*.*.*..*.*.*.*
> >> 20 ++---------------------------------------------------------------------+
> >>
> >>
> >> [*] bisect-good sample
> >> [O] bisect-bad sample
> >>
> >> To reproduce:
> >>
> >> git clone git://git.kernel.org/pub/scm/linux/kernel/git/wfg/lkp-tests.git
> >> cd lkp-tests
> >> bin/lkp install job.yaml # job file is attached in this email
> >> bin/lkp run 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,
> >> Ying Huang
> >
> > Thanks...
> >
> > Huh...I'm stumped on this one. If anything I would have expected better
> > performance with this patch since we don't even take the file_lock or
> > do the fcheck in the F_UNLCK codepath now, or when there is an error.
> >
> > I'll see if I can reproduce it on my own test rig, but I'd welcome
> > ideas of where and how this performance regression could have crept in.
>
> This is a performance increase instead of performance regression.
>
Hah, no wonder that made no sense. Ok, thanks for letting me know!
--
Jeff Layton <jeff.layton@primarydata.com>
^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [LKP] [lkp] [locks] 7f3697e24d: +35.1% will-it-scale.per_thread_ops
2016-01-29 2:52 ` [LKP] " Huang, Ying
2016-01-29 12:13 ` Jeff Layton
@ 2016-02-01 13:39 ` J. Bruce Fields
1 sibling, 0 replies; 5+ messages in thread
From: J. Bruce Fields @ 2016-02-01 13:39 UTC (permalink / raw)
To: Huang, Ying; +Cc: Jeff Layton, lkp, LKML, Dmitry Vyukov, Alexander Viro
On Fri, Jan 29, 2016 at 10:52:20AM +0800, Huang, Ying wrote:
> Jeff Layton <jeff.layton@primarydata.com> writes:
>
> > On Fri, 29 Jan 2016 09:32:19 +0800
> > kernel test robot <ying.huang@linux.intel.com> wrote:
> >
> >> FYI, we noticed the below changes on
> >>
> >> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
> >> commit 7f3697e24dc3820b10f445a4a7d914fc356012d1 ("locks: fix unlock when fcntl_setlk races with a close")
> >>
> >>
> >> =========================================================================================
> >> compiler/cpufreq_governor/kconfig/rootfs/tbox_group/test/testcase:
> >> gcc-4.9/performance/x86_64-rhel/debian-x86_64-2015-02-07.cgz/lkp-snb01/lock1/will-it-scale
> >>
> >> commit:
> >> 9189922675ecca0fab38931d86b676e9d79602dc
> >> 7f3697e24dc3820b10f445a4a7d914fc356012d1
> >>
> >> 9189922675ecca0f 7f3697e24dc3820b10f445a4a7
> >> ---------------- --------------------------
> >> %stddev %change %stddev
> >> \ | \
> >> 2376432 ± 0% +2.1% 2427484 ± 0% will-it-scale.per_process_ops
> >> 807889 ± 0% +35.1% 1091496 ± 4% will-it-scale.per_thread_ops
> >> 22.08 ± 2% +89.1% 41.75 ± 5% will-it-scale.time.user_time
> >> 1238371 ± 14% +100.4% 2481345 ± 39% cpuidle.C1E-SNB.time
> >> 3098 ± 57% -66.6% 1035 ±171% numa-numastat.node1.other_node
> >> 379.25 ± 8% -21.4% 298.00 ± 12% numa-vmstat.node0.nr_alloc_batch
> >> 22.08 ± 2% +89.1% 41.75 ± 5% time.user_time
> >> 1795 ± 4% +7.5% 1930 ± 2% vmstat.system.cs
> >> 0.54 ± 5% +136.9% 1.28 ± 10% perf-profile.cycles.___might_sleep.__might_sleep.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file
> >> 1.65 ± 57% +245.2% 5.70 ± 29% perf-profile.cycles.__fdget_raw.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 1.58 ± 59% +248.3% 5.50 ± 31% perf-profile.cycles.__fget.__fget_light.__fdget_raw.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 1.62 ± 58% +246.3% 5.63 ± 30% perf-profile.cycles.__fget_light.__fdget_raw.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.00 ± -1% +Inf% 5.88 ± 11% perf-profile.cycles.__memset.locks_alloc_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait
> >> 2.50 ± 2% -100.0% 0.00 ± -1% perf-profile.cycles.__memset.locks_alloc_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk
> >> 1.29 ± 4% +138.8% 3.09 ± 11% perf-profile.cycles.__memset.locks_alloc_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.47 ± 9% +144.4% 1.16 ± 11% perf-profile.cycles.__might_fault.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.37 ± 12% +140.3% 0.90 ± 9% perf-profile.cycles.__might_sleep.__might_fault.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.86 ± 6% +137.7% 2.05 ± 10% perf-profile.cycles.__might_sleep.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file.vfs_lock_file
> >> 0.61 ± 14% +56.8% 0.95 ± 14% perf-profile.cycles.__might_sleep.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.sys_fcntl
> >> 0.00 ± -1% +Inf% 39.84 ± 12% perf-profile.cycles.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk.sys_fcntl
> >> 16.44 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.00 ± -1% +Inf% 1.77 ± 11% perf-profile.cycles._raw_spin_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
> >> 59.34 ± 1% -72.4% 16.36 ± 33% perf-profile.cycles._raw_spin_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.46 ± 11% +144.9% 1.13 ± 19% perf-profile.cycles.avc_has_perm.inode_has_perm.file_has_perm.selinux_file_fcntl.security_file_fcntl
> >> 0.87 ± 6% +103.2% 1.77 ± 12% perf-profile.cycles.avc_has_perm.inode_has_perm.file_has_perm.selinux_file_lock.security_file_lock
> >> 0.81 ± 4% +135.7% 1.90 ± 10% perf-profile.cycles.copy_user_generic_string.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.00 ± -1% +Inf% 41.86 ± 12% perf-profile.cycles.do_lock_file_wait.part.29.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.88 ± 6% +127.8% 2.00 ± 9% perf-profile.cycles.entry_SYSCALL_64
> >> 0.86 ± 4% +122.6% 1.92 ± 12% perf-profile.cycles.entry_SYSCALL_64_after_swapgs
> >> 84.98 ± 0% -9.1% 77.20 ± 2% perf-profile.cycles.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.76 ± 10% +142.1% 1.84 ± 14% perf-profile.cycles.file_has_perm.selinux_file_fcntl.security_file_fcntl.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 1.35 ± 4% +106.3% 2.78 ± 11% perf-profile.cycles.file_has_perm.selinux_file_lock.security_file_lock.fcntl_setlk.sys_fcntl
> >> 0.00 ± -1% +Inf% 0.89 ± 12% perf-profile.cycles.flock_to_posix_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 6.90 ± 4% -48.6% 3.55 ± 27% perf-profile.cycles.fput.entry_SYSCALL_64_fastpath
> >> 0.51 ± 10% +140.5% 1.23 ± 16% perf-profile.cycles.inode_has_perm.isra.31.file_has_perm.selinux_file_fcntl.security_file_fcntl.sys_fcntl
> >> 0.98 ± 4% +97.7% 1.93 ± 11% perf-profile.cycles.inode_has_perm.isra.31.file_has_perm.selinux_file_lock.security_file_lock.fcntl_setlk
> >> 0.00 ± -1% +Inf% 6.56 ± 10% perf-profile.cycles.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait
> >> 2.75 ± 4% -100.0% 0.00 ± -1% perf-profile.cycles.kmem_cache_alloc.locks_alloc_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk
> >> 1.53 ± 7% +119.7% 3.37 ± 13% perf-profile.cycles.kmem_cache_alloc.locks_alloc_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.00 ± -1% +Inf% 1.79 ± 11% perf-profile.cycles.kmem_cache_free.locks_free_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait
> >> 0.46 ± 14% +257.0% 1.66 ± 11% perf-profile.cycles.kmem_cache_free.locks_free_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.40 ± 7% +158.6% 1.05 ± 17% perf-profile.cycles.kmem_cache_free.locks_free_lock.locks_dispose_list.__posix_lock_file.vfs_lock_file
> >> 0.00 ± -1% +Inf% 0.96 ± 10% perf-profile.cycles.lg_local_lock.locks_insert_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait
> >> 0.00 ± -1% +Inf% 14.69 ± 10% perf-profile.cycles.locks_alloc_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
> >> 6.38 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.locks_alloc_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
> >> 3.28 ± 6% +127.1% 7.45 ± 12% perf-profile.cycles.locks_alloc_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.00 ± -1% +Inf% 9.75 ± 13% perf-profile.cycles.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
> >> 3.61 ± 1% -100.0% 0.00 ± -1% perf-profile.cycles.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
> >> 0.00 ± -1% +Inf% 1.84 ± 11% perf-profile.cycles.locks_dispose_list.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
> >> 0.00 ± -1% +Inf% 2.42 ± 10% perf-profile.cycles.locks_free_lock.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
> >> 1.00 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.locks_free_lock.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
> >> 0.63 ± 11% +224.1% 2.05 ± 10% perf-profile.cycles.locks_free_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.00 ± -1% +Inf% 1.22 ± 14% perf-profile.cycles.locks_free_lock.locks_dispose_list.__posix_lock_file.vfs_lock_file.do_lock_file_wait
> >> 0.00 ± -1% +Inf% 6.17 ± 15% perf-profile.cycles.locks_insert_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait.fcntl_setlk
> >> 2.31 ± 6% -100.0% 0.00 ± -1% perf-profile.cycles.locks_insert_lock_ctx.__posix_lock_file.vfs_lock_file.fcntl_setlk.sys_fcntl
> >> 0.00 ± -1% +Inf% 8.96 ± 13% perf-profile.cycles.locks_unlink_lock_ctx.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.do_lock_file_wait
> >> 3.27 ± 1% -100.0% 0.00 ± -1% perf-profile.cycles.locks_unlink_lock_ctx.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file.fcntl_setlk
> >> 53.88 ± 1% -79.7% 10.92 ± 46% perf-profile.cycles.native_queued_spin_lock_slowpath._raw_spin_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 2.75 ± 0% +183.3% 7.79 ± 13% perf-profile.cycles.put_pid.locks_unlink_lock_ctx.locks_delete_lock_ctx.__posix_lock_file.vfs_lock_file
> >> 1.11 ± 9% +137.2% 2.63 ± 14% perf-profile.cycles.security_file_fcntl.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 1.69 ± 4% +118.2% 3.69 ± 11% perf-profile.cycles.security_file_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.91 ± 9% +139.0% 2.17 ± 14% perf-profile.cycles.selinux_file_fcntl.security_file_fcntl.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 1.39 ± 4% +114.6% 2.97 ± 10% perf-profile.cycles.selinux_file_lock.security_file_lock.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 0.00 ± -1% +Inf% 41.12 ± 12% perf-profile.cycles.vfs_lock_file.do_lock_file_wait.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 17.04 ± 3% -100.0% 0.00 ± -1% perf-profile.cycles.vfs_lock_file.fcntl_setlk.sys_fcntl.entry_SYSCALL_64_fastpath
> >> 34.75 ±148% +132.4% 80.75 ± 82% sched_debug.cfs_rq:/.load.8
> >> 15.00 ± 9% +198.3% 44.75 ± 72% sched_debug.cfs_rq:/.load_avg.21
> >> 25.00 ± 29% +574.0% 168.50 ± 78% sched_debug.cfs_rq:/.load_avg.9
> >> 38.47 ± 5% +29.1% 49.65 ± 26% sched_debug.cfs_rq:/.load_avg.avg
> >> 63.17 ± 10% +44.3% 91.16 ± 36% sched_debug.cfs_rq:/.load_avg.stddev
> >> 893865 ± 12% -12.5% 782455 ± 0% sched_debug.cfs_rq:/.min_vruntime.25
> >> 18.25 ± 26% +52.1% 27.75 ± 25% sched_debug.cfs_rq:/.runnable_load_avg.9
> >> -57635 ±-68% -196.4% 55548 ±130% sched_debug.cfs_rq:/.spread0.1
> >> -802264 ±-25% -29.5% -565458 ±-49% sched_debug.cfs_rq:/.spread0.8
> >> -804662 ±-25% -29.4% -567811 ±-48% sched_debug.cfs_rq:/.spread0.min
> >> 1233 ± 5% +30.9% 1614 ± 28% sched_debug.cfs_rq:/.tg_load_avg.0
> >> 1233 ± 5% +30.9% 1614 ± 28% sched_debug.cfs_rq:/.tg_load_avg.1
> >> 1228 ± 5% +30.3% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.10
> >> 1228 ± 5% +30.4% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.11
> >> 1228 ± 5% +30.3% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.12
> >> 1229 ± 5% +30.0% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.13
> >> 1228 ± 5% +30.1% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.14
> >> 1229 ± 5% +30.0% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.15
> >> 1226 ± 5% +30.3% 1598 ± 27% sched_debug.cfs_rq:/.tg_load_avg.16
> >> 1226 ± 5% +30.2% 1597 ± 27% sched_debug.cfs_rq:/.tg_load_avg.17
> >> 1227 ± 5% +30.1% 1595 ± 27% sched_debug.cfs_rq:/.tg_load_avg.18
> >> 1227 ± 5% +29.4% 1588 ± 26% sched_debug.cfs_rq:/.tg_load_avg.19
> >> 1233 ± 5% +30.4% 1609 ± 27% sched_debug.cfs_rq:/.tg_load_avg.2
> >> 1222 ± 5% +29.9% 1587 ± 26% sched_debug.cfs_rq:/.tg_load_avg.20
> >> 1223 ± 5% +24.2% 1519 ± 20% sched_debug.cfs_rq:/.tg_load_avg.21
> >> 1223 ± 5% +23.8% 1515 ± 20% sched_debug.cfs_rq:/.tg_load_avg.22
> >> 1223 ± 5% +23.9% 1515 ± 20% sched_debug.cfs_rq:/.tg_load_avg.23
> >> 1223 ± 5% +23.9% 1515 ± 20% sched_debug.cfs_rq:/.tg_load_avg.24
> >> 1223 ± 5% +23.5% 1511 ± 19% sched_debug.cfs_rq:/.tg_load_avg.25
> >> 1224 ± 5% +23.5% 1512 ± 19% sched_debug.cfs_rq:/.tg_load_avg.26
> >> 1223 ± 5% +23.1% 1506 ± 19% sched_debug.cfs_rq:/.tg_load_avg.27
> >> 1223 ± 5% +22.5% 1499 ± 19% sched_debug.cfs_rq:/.tg_load_avg.28
> >> 1224 ± 5% +22.5% 1499 ± 19% sched_debug.cfs_rq:/.tg_load_avg.29
> >> 1233 ± 5% +30.3% 1607 ± 27% sched_debug.cfs_rq:/.tg_load_avg.3
> >> 1223 ± 5% +22.2% 1495 ± 18% sched_debug.cfs_rq:/.tg_load_avg.30
> >> 1224 ± 5% +22.0% 1493 ± 19% sched_debug.cfs_rq:/.tg_load_avg.31
> >> 1234 ± 5% +30.0% 1604 ± 28% sched_debug.cfs_rq:/.tg_load_avg.4
> >> 1233 ± 5% +30.0% 1604 ± 28% sched_debug.cfs_rq:/.tg_load_avg.5
> >> 1231 ± 5% +30.3% 1604 ± 28% sched_debug.cfs_rq:/.tg_load_avg.6
> >> 1233 ± 5% +30.0% 1603 ± 27% sched_debug.cfs_rq:/.tg_load_avg.7
> >> 1231 ± 5% +30.1% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.8
> >> 1228 ± 5% +30.3% 1601 ± 27% sched_debug.cfs_rq:/.tg_load_avg.9
> >> 1228 ± 5% +27.8% 1569 ± 24% sched_debug.cfs_rq:/.tg_load_avg.avg
> >> 1246 ± 5% +30.7% 1628 ± 27% sched_debug.cfs_rq:/.tg_load_avg.max
> >> 1212 ± 5% +22.2% 1481 ± 19% sched_debug.cfs_rq:/.tg_load_avg.min
> >> 15.00 ± 9% +198.3% 44.75 ± 72% sched_debug.cfs_rq:/.tg_load_avg_contrib.21
> >> 25.00 ± 29% +574.0% 168.50 ± 78% sched_debug.cfs_rq:/.tg_load_avg_contrib.9
> >> 38.53 ± 5% +29.0% 49.71 ± 26% sched_debug.cfs_rq:/.tg_load_avg_contrib.avg
> >> 63.34 ± 10% +44.1% 91.30 ± 36% sched_debug.cfs_rq:/.tg_load_avg_contrib.stddev
> >> 532.25 ± 2% +8.5% 577.50 ± 6% sched_debug.cfs_rq:/.util_avg.15
> >> 210.75 ± 14% -14.4% 180.50 ± 4% sched_debug.cfs_rq:/.util_avg.29
> >> 450.00 ± 22% +50.7% 678.00 ± 18% sched_debug.cfs_rq:/.util_avg.9
> >> 955572 ± 4% -10.2% 857813 ± 5% sched_debug.cpu.avg_idle.6
> >> 23.99 ± 60% -76.2% 5.71 ± 24% sched_debug.cpu.clock.stddev
> >> 23.99 ± 60% -76.2% 5.71 ± 24% sched_debug.cpu.clock_task.stddev
> >> 2840 ± 37% -47.4% 1492 ± 65% sched_debug.cpu.curr->pid.25
> >> 34.75 ±148% +132.4% 80.75 ± 82% sched_debug.cpu.load.8
> >> 61776 ± 7% -7.1% 57380 ± 0% sched_debug.cpu.nr_load_updates.25
> >> 6543 ± 2% +20.4% 7879 ± 9% sched_debug.cpu.nr_switches.0
> >> 5256 ± 23% +177.1% 14566 ± 52% sched_debug.cpu.nr_switches.27
> >> 7915 ± 3% +8.7% 8605 ± 3% sched_debug.cpu.nr_switches.avg
> >> -0.25 ±-519% +1900.0% -5.00 ±-24% sched_debug.cpu.nr_uninterruptible.12
> >> 2.00 ± 93% -125.0% -0.50 ±-300% sched_debug.cpu.nr_uninterruptible.24
> >> 17468 ± 14% +194.3% 51413 ± 75% sched_debug.cpu.sched_count.15
> >> 2112 ± 2% +20.8% 2552 ± 11% sched_debug.cpu.sched_goidle.0
> >> 2103 ± 34% +219.0% 6709 ± 55% sched_debug.cpu.sched_goidle.27
> >> 3159 ± 3% +8.2% 3418 ± 4% sched_debug.cpu.sched_goidle.avg
> >> 1323 ± 64% -72.7% 361.50 ± 15% sched_debug.cpu.ttwu_count.23
> >> 3264 ± 12% +94.4% 6347 ± 41% sched_debug.cpu.ttwu_count.27
> >> 3860 ± 3% +9.0% 4208 ± 3% sched_debug.cpu.ttwu_count.avg
> >> 2358 ± 3% +28.7% 3035 ± 9% sched_debug.cpu.ttwu_local.0
> >> 1110 ± 22% +54.6% 1716 ± 28% sched_debug.cpu.ttwu_local.27
> >> 1814 ± 8% +16.1% 2106 ± 5% sched_debug.cpu.ttwu_local.stddev
> >>
> >>
> >> lkp-snb01: Sandy Bridge-EP
> >> Memory: 32G
> >>
> >> will-it-scale.per_thread_ops
> >>
> >> 1.2e+06 ++---------------------------------------------------------------+
> >> | O |
> >> 1.15e+06 O+O O O O O O O O |
> >> 1.1e+06 ++ |
> >> | O O O O O OO |
> >> 1.05e+06 ++ O O |
> >> 1e+06 ++ |
> >> | |
> >> 950000 ++ |
> >> 900000 ++ |
> >> | |
> >> 850000 ++ |
> >> 800000 *+*.*.*.*.*.*.*.*.*.*.*.*. .*.*. *.*.*.*.*.*.*.*.*.*.*.*.*.*.*.*.*
> >> | * * |
> >> 750000 ++---------------------------------------------------------------+
> >>
> >>
> >> will-it-scale.time.user_time
> >>
> >> 50 ++---------------------------------------------------------------------+
> >> | |
> >> 45 ++ O O O O O O |
> >> O O O O |
> >> | O O O O O |
> >> 40 ++ O O O O |
> >> | |
> >> 35 ++ |
> >> | |
> >> 30 ++ |
> >> | |
> >> | * |
> >> 25 ++ + + |
> >> *.*.*.*..*.* *.*.*..*.*.*.*.*.*.*..*.*.*.*.*.*.*..*.*.*.*.*.*..*.*.*.*
> >> 20 ++---------------------------------------------------------------------+
> >>
> >>
> >> [*] bisect-good sample
> >> [O] bisect-bad sample
> >>
> >> To reproduce:
> >>
> >> git clone git://git.kernel.org/pub/scm/linux/kernel/git/wfg/lkp-tests.git
> >> cd lkp-tests
> >> bin/lkp install job.yaml # job file is attached in this email
> >> bin/lkp run 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,
> >> Ying Huang
> >
> > Thanks...
> >
> > Huh...I'm stumped on this one. If anything I would have expected better
> > performance with this patch since we don't even take the file_lock or
> > do the fcheck in the F_UNLCK codepath now, or when there is an error.
> >
> > I'll see if I can reproduce it on my own test rig, but I'd welcome
> > ideas of where and how this performance regression could have crept in.
>
> This is a performance increase instead of performance regression.
Could you provide any help reading the above graphs? For example, is
the "bisect-good" case before or after the given commit? And are lower
or higher numbers better on the graph?
Thanks for doing this testing. I'm impressed that it's happening and
curious to learn anything more about it.
--b.
^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2016-02-01 13:39 UTC | newest]
Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2016-01-29 1:32 [lkp] [locks] 7f3697e24d: +35.1% will-it-scale.per_thread_ops kernel test robot
2016-01-29 2:38 ` Jeff Layton
2016-01-29 2:52 ` [LKP] " Huang, Ying
2016-01-29 12:13 ` Jeff Layton
2016-02-01 13:39 ` J. Bruce Fields
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).