[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <202408062146.832faa23-oliver.sang@intel.com>
Date: Tue, 6 Aug 2024 21:44:28 +0800
From: kernel test robot <oliver.sang@...el.com>
To: Yu Ma <yu.ma@...el.com>
CC: <oe-lkp@...ts.linux.dev>, <lkp@...el.com>, Jan Kara <jack@...e.cz>, "Tim
Chen" <tim.c.chen@...ux.intel.com>, <linux-fsdevel@...r.kernel.org>,
<ying.huang@...el.com>, <feng.tang@...el.com>, <fengwei.yin@...el.com>,
<brauner@...nel.org>, <mjguzik@...il.com>, <edumazet@...gle.com>,
<yu.ma@...el.com>, <linux-kernel@...r.kernel.org>, <pan.deng@...el.com>,
<tianyou.li@...el.com>, <tim.c.chen@...el.com>, <viro@...iv.linux.org.uk>,
<oliver.sang@...el.com>
Subject: Re: [PATCH v5 1/3] fs/file.c: remove sanity_check and add
likely/unlikely in alloc_fd()
Hello,
kernel test robot noticed a 1.2% improvement of will-it-scale.per_process_ops on:
commit: f1139c8e66d5c618aad04a93a2378ad9586464f9 ("[PATCH v5 1/3] fs/file.c: remove sanity_check and add likely/unlikely in alloc_fd()")
url: https://github.com/intel-lab-lkp/linux/commits/Yu-Ma/fs-file-c-remove-sanity_check-and-add-likely-unlikely-in-alloc_fd/20240717-224830
base: https://git.kernel.org/cgit/linux/kernel/git/vfs/vfs.git vfs.all
patch link: https://lore.kernel.org/all/20240717145018.3972922-2-yu.ma@intel.com/
patch subject: [PATCH v5 1/3] fs/file.c: remove sanity_check and add likely/unlikely in alloc_fd()
testcase: will-it-scale
test machine: 256 threads 2 sockets GENUINE INTEL(R) XEON(R) (Sierra Forest) with 128G memory
parameters:
nr_task: 100%
mode: process
test: dup1
cpufreq_governor: performance
Details are as below:
-------------------------------------------------------------------------------------------------->
The kernel config and materials to reproduce are available at:
https://download.01.org/0day-ci/archive/20240806/202408062146.832faa23-oliver.sang@intel.com
=========================================================================================
compiler/cpufreq_governor/kconfig/mode/nr_task/rootfs/tbox_group/test/testcase:
gcc-13/performance/x86_64-rhel-8.3/process/100%/debian-12-x86_64-20240206.cgz/lkp-srf-2sp1/dup1/will-it-scale
commit:
5f30e082ab ("Merge branch 'vfs.iomap' into vfs.all")
f1139c8e66 ("fs/file.c: remove sanity_check and add likely/unlikely in alloc_fd()")
5f30e082ab8b3431 f1139c8e66d5c618aad04a93a23
---------------- ---------------------------
%stddev %change %stddev
\ | \
377983 ± 69% +74.1% 658036 ± 17% numa-meminfo.node0.AnonPages
18.17 ± 10% -48.6% 9.33 ± 35% perf-c2c.DRAM.local
8.796e+08 +1.2% 8.903e+08 will-it-scale.256.processes
3436082 +1.2% 3477810 will-it-scale.per_process_ops
8.796e+08 +1.2% 8.903e+08 will-it-scale.workload
1.517e+11 -4.3% 1.452e+11 perf-stat.i.branch-instructions
0.03 ± 8% +0.0 0.04 ± 36% perf-stat.i.branch-miss-rate%
0.93 +3.9% 0.96 perf-stat.i.cpi
7.13e+11 -3.5% 6.88e+11 perf-stat.i.instructions
1.08 -3.4% 1.04 perf-stat.i.ipc
0.93 +3.4% 0.96 perf-stat.overall.cpi
1.08 -3.3% 1.04 perf-stat.overall.ipc
245130 -4.4% 234451 perf-stat.overall.path-length
1.512e+11 -4.3% 1.447e+11 perf-stat.ps.branch-instructions
7.106e+11 -3.5% 6.857e+11 perf-stat.ps.instructions
2.156e+14 -3.2% 2.087e+14 perf-stat.total.instructions
14.90 -0.7 14.20 perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.dup
12.01 -0.7 11.32 perf-profile.calltrace.cycles-pp.__x64_sys_dup.do_syscall_64.entry_SYSCALL_64_after_hwframe.dup
16.54 -0.7 15.88 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_after_hwframe.dup
6.44 -0.6 5.89 perf-profile.calltrace.cycles-pp.alloc_fd.__x64_sys_dup.do_syscall_64.entry_SYSCALL_64_after_hwframe.dup
2.86 -0.0 2.82 perf-profile.calltrace.cycles-pp.entry_SYSRETQ_unsafe_stack.__close
8.94 -0.0 8.90 perf-profile.calltrace.cycles-pp.filp_flush.__x64_sys_close.do_syscall_64.entry_SYSCALL_64_after_hwframe.__close
7.76 -0.0 7.72 perf-profile.calltrace.cycles-pp.locks_remove_posix.filp_flush.__x64_sys_close.do_syscall_64.entry_SYSCALL_64_after_hwframe
2.58 -0.0 2.54 perf-profile.calltrace.cycles-pp.__fput_sync.__x64_sys_close.do_syscall_64.entry_SYSCALL_64_after_hwframe.__close
1.11 -0.0 1.10 perf-profile.calltrace.cycles-pp.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe.__close
1.33 +0.0 1.35 perf-profile.calltrace.cycles-pp.testcase
0.54 +0.0 0.56 perf-profile.calltrace.cycles-pp.x64_sys_call.do_syscall_64.entry_SYSCALL_64_after_hwframe.__close
0.79 +0.0 0.82 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_safe_stack.dup
1.33 +0.0 1.37 perf-profile.calltrace.cycles-pp.close@plt
2.73 +0.1 2.78 perf-profile.calltrace.cycles-pp._raw_spin_lock.file_close_fd.__x64_sys_close.do_syscall_64.entry_SYSCALL_64_after_hwframe
1.05 +0.1 1.11 perf-profile.calltrace.cycles-pp.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe.dup
4.35 +0.1 4.42 perf-profile.calltrace.cycles-pp.file_close_fd.__x64_sys_close.do_syscall_64.entry_SYSCALL_64_after_hwframe.__close
22.18 +0.3 22.51 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64.__close
21.50 ± 2% +1.5 23.02 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64.dup
12.10 -0.7 11.39 perf-profile.children.cycles-pp.__x64_sys_dup
34.79 -0.7 34.12 perf-profile.children.cycles-pp.do_syscall_64
38.04 -0.6 37.42 perf-profile.children.cycles-pp.entry_SYSCALL_64_after_hwframe
6.48 -0.6 5.90 perf-profile.children.cycles-pp.alloc_fd
1.86 -0.5 1.41 perf-profile.children.cycles-pp.syscall_return_via_sysret
0.57 -0.1 0.47 perf-profile.children.cycles-pp.fd_install
9.11 -0.0 9.07 perf-profile.children.cycles-pp.filp_flush
7.93 -0.0 7.89 perf-profile.children.cycles-pp.locks_remove_posix
2.61 -0.0 2.58 perf-profile.children.cycles-pp.__fput_sync
1.16 +0.0 1.18 perf-profile.children.cycles-pp.x64_sys_call
0.05 +0.0 0.07 ± 13% perf-profile.children.cycles-pp.clockevents_program_event
0.51 +0.0 0.53 perf-profile.children.cycles-pp.syscall_exit_to_user_mode_prepare
2.17 +0.0 2.20 perf-profile.children.cycles-pp.syscall_exit_to_user_mode
5.72 +0.0 5.75 perf-profile.children.cycles-pp._raw_spin_lock
2.10 +0.0 2.13 perf-profile.children.cycles-pp.testcase
2.02 +0.0 2.06 perf-profile.children.cycles-pp.entry_SYSCALL_64_safe_stack
0.13 ± 2% +0.0 0.17 perf-profile.children.cycles-pp.dup@plt
4.38 +0.1 4.46 perf-profile.children.cycles-pp.file_close_fd
23.00 +0.1 23.11 perf-profile.children.cycles-pp.entry_SYSRETQ_unsafe_stack
59.27 +0.5 59.73 perf-profile.children.cycles-pp.__close
28.73 +1.1 29.80 perf-profile.children.cycles-pp.entry_SYSCALL_64
1.86 -0.5 1.41 perf-profile.self.cycles-pp.syscall_return_via_sysret
2.28 -0.2 2.12 perf-profile.self.cycles-pp.alloc_fd
0.54 -0.1 0.43 perf-profile.self.cycles-pp.fd_install
7.87 -0.0 7.83 perf-profile.self.cycles-pp.locks_remove_posix
2.47 -0.0 2.44 perf-profile.self.cycles-pp.__fput_sync
1.23 +0.0 1.24 perf-profile.self.cycles-pp.file_close_fd_locked
1.09 +0.0 1.11 perf-profile.self.cycles-pp.x64_sys_call
0.51 +0.0 0.53 perf-profile.self.cycles-pp.syscall_exit_to_user_mode_prepare
1.29 +0.0 1.32 perf-profile.self.cycles-pp.testcase
5.66 +0.0 5.69 perf-profile.self.cycles-pp._raw_spin_lock
1.95 +0.0 1.99 perf-profile.self.cycles-pp.entry_SYSCALL_64_safe_stack
2.85 +0.0 2.90 perf-profile.self.cycles-pp.entry_SYSCALL_64_after_hwframe
0.02 ±141% +0.0 0.06 ± 13% perf-profile.self.cycles-pp.ktime_get
0.00 +0.1 0.07 perf-profile.self.cycles-pp.dup@plt
22.93 +0.1 23.05 perf-profile.self.cycles-pp.entry_SYSRETQ_unsafe_stack
10.11 +0.2 10.34 perf-profile.self.cycles-pp.dup
13.70 +0.3 13.98 perf-profile.self.cycles-pp.entry_SYSCALL_64
9.84 ± 3% +0.7 10.51 perf-profile.self.cycles-pp.__close
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.
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
Powered by blists - more mailing lists