[frederic-dynticks:timers/anna] [timers] 5bbb5400a9: unixbench.throughput 2.1% improvement

[frederic-dynticks:timers/anna] [timers] 5bbb5400a9: unixbench.throughput 2.1% improvement

[kernel test robot]

Hello,

kernel test robot noticed a 2.1% improvement of unixbench.throughput on:


commit: 5bbb5400a9f70dae4069290b581b1d2c934c932a ("timers: Keep the pinned timers separate from the others")
https://git.kernel.org/cgit/linux/kernel/git/frederic/linux-dynticks.git timers/anna

testcase: unixbench
test machine: 36 threads 1 sockets Intel(R) Core(TM) i9-10980XE CPU @ 3.00GHz (Cascade Lake) with 128G memory
parameters:

	runtime: 300s
	nr_task: 1
	test: syscall
	cpufreq_governor: performance






Details are as below:
-------------------------------------------------------------------------------------------------->


The kernel config and materials to reproduce are available at:
https://download.01.org/0day-ci/archive/20240222/202402221047.8338ced3-oliver.sang@intel.com

=========================================================================================
compiler/cpufreq_governor/kconfig/nr_task/rootfs/runtime/tbox_group/test/testcase:
  gcc-12/performance/x86_64-rhel-8.3/1/debian-11.1-x86_64-20220510.cgz/300s/lkp-csl-d02/syscall/unixbench

commit: 
  f2234c1add ("timers: Split next timer interrupt logic")
  5bbb5400a9 ("timers: Keep the pinned timers separate from the others")

f2234c1add92fe1b 5bbb5400a9f70dae4069290b581 
---------------- --------------------------- 
         %stddev     %change         %stddev
             \          |                \  
      3.99 ±  8%     +14.4%       4.57 ±  8%  perf-sched.wait_time.avg.ms.rcu_gp_kthread.kthread.ret_from_fork.ret_from_fork_asm
   2744609            +2.1%    2801712        unixbench.throughput
 1.071e+09            +2.0%  1.093e+09        unixbench.workload
      2.65 ± 10%      -0.5        2.12 ±  8%  perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.umask
      0.29 ±100%      +0.4        0.67 ±  7%  perf-profile.calltrace.cycles-pp.syscall_return_via_sysret.syscall
      0.29 ±101%      +0.4        0.70 ± 10%  perf-profile.calltrace.cycles-pp.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe.getuid
      0.00            +0.2        0.21 ± 15%  perf-profile.children.cycles-pp.__get_next_timer_interrupt
      5.67 ± 12%      -2.4        3.24 ±  6%  perf-profile.self.cycles-pp.do_syscall_64
      0.00            +0.1        0.09 ± 14%  perf-profile.self.cycles-pp.__get_next_timer_interrupt
 6.425e+08            +1.8%  6.539e+08        perf-stat.i.branch-instructions
 9.814e+08            +1.9%  9.997e+08        perf-stat.i.dTLB-loads
 6.276e+08            +1.9%  6.397e+08        perf-stat.i.dTLB-stores
    325643 ±  2%      +7.6%     350447 ±  3%  perf-stat.i.iTLB-loads
 3.314e+09            +1.8%  3.372e+09        perf-stat.i.instructions
    555.44            +3.8%     576.68        perf-stat.i.instructions-per-iTLB-miss
      0.75            +1.8%       0.76        perf-stat.i.ipc
     62.51            +1.9%      63.67        perf-stat.i.metric.M/sec
      1.25            -1.7%       1.22        perf-stat.overall.cpi
      0.80            +1.7%       0.82        perf-stat.overall.ipc
 6.408e+08            +1.8%  6.522e+08        perf-stat.ps.branch-instructions
 9.789e+08            +1.9%  9.971e+08        perf-stat.ps.dTLB-loads
 6.259e+08            +1.9%  6.381e+08        perf-stat.ps.dTLB-stores
    324768 ±  2%      +7.6%     349509 ±  3%  perf-stat.ps.iTLB-loads
 3.305e+09            +1.8%  3.363e+09        perf-stat.ps.instructions
 1.292e+12            +1.8%  1.315e+12        perf-stat.total.instructions




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