[tip:sched/core] [sched/topology] 9e005ed211: stress-ng.kill.ops_per_sec 4.6% regression

[tip:sched/core] [sched/topology] 9e005ed211: stress-ng.kill.ops_per_sec 4.6% regression

[kernel test robot]

Hello,

kernel test robot noticed a 4.6% regression of stress-ng.kill.ops_per_sec on:


commit: 9e005ed21152d4a4bb0ceea71045ff8a642a6feb ("sched/topology: Allow multiple domains to claim sched_domain_shared")
https://git.kernel.org/cgit/linux/kernel/git/tip/tip.git sched/core


testcase: stress-ng
config: x86_64-rhel-9.4
compiler: gcc-14
test machine: 256 threads 4 sockets INTEL(R) XEON(R) PLATINUM 8592+ (Emerald Rapids) with 256G memory
parameters:

	nr_threads: 100%
	testtime: 60s
	test: kill
	cpufreq_governor: performance


If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <oliver.sang@intel.com>
| Closes: https://lore.kernel.org/oe-lkp/202606111053.c484a580-lkp@intel.com


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


The kernel config and materials to reproduce are available at:
https://download.01.org/0day-ci/archive/20260611/202606111053.c484a580-lkp@intel.com

=========================================================================================
compiler/cpufreq_governor/kconfig/nr_threads/rootfs/tbox_group/test/testcase/testtime:
  gcc-14/performance/x86_64-rhel-9.4/100%/debian-13-x86_64-20250902.cgz/lkp-emr-2sp1/kill/stress-ng/60s

commit: 
  a26d9208c1 ("Merge branch 'sched/cache'")
  9e005ed211 ("sched/topology: Allow multiple domains to claim sched_domain_shared")

a26d9208c1376ac3 9e005ed21152d4a4bb0ceea7104 
---------------- --------------------------- 
         %stddev     %change         %stddev
             \          |                \  
     16542            -5.6%      15623        stress-ng.kill.kill_calls_per_sec
  66940342            -4.6%   63830711        stress-ng.kill.ops
   1116630            -4.6%    1064849        stress-ng.kill.ops_per_sec
     50279            +5.7%      53164        meminfo.KernelStack
     14019 ± 46%     -48.8%       7174 ± 70%  numa-vmstat.node2.nr_slab_reclaimable
     10643 ± 21%     +49.0%      15853 ± 19%  numa-vmstat.node3.nr_slab_reclaimable
     50278            +5.7%      53144        proc-vmstat.nr_kernel_stack
     74611            +1.0%      75344        proc-vmstat.nr_vmalloc
    613641            +5.5%     647439 ±  3%  proc-vmstat.pgfree
     56078 ± 46%     -48.8%      28697 ± 70%  numa-meminfo.node2.KReclaimable
     56078 ± 46%     -48.8%      28697 ± 70%  numa-meminfo.node2.SReclaimable
     42574 ± 21%     +49.1%      63467 ± 19%  numa-meminfo.node3.KReclaimable
     42574 ± 21%     +49.1%      63467 ± 19%  numa-meminfo.node3.SReclaimable
      0.03 ± 55%     -64.7%       0.01 ± 99%  sched_debug.cfs_rq:/.spread.avg
      0.26 ± 27%     -55.0%       0.12 ± 84%  sched_debug.cfs_rq:/.spread.stddev
      4715 ± 14%     -28.3%       3380 ± 44%  sched_debug.cpu.curr->pid.min
      0.32 ±  3%      -8.3%       0.30 ±  3%  sched_debug.cpu.nr_running.stddev
     22.17 ± 31%    +128.9%      50.75 ± 12%  sched_debug.cpu.nr_uninterruptible.max
      3.99 ± 14%     +55.6%       6.20 ±  4%  sched_debug.cpu.nr_uninterruptible.stddev




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

Re: [tip:sched/core] [sched/topology] 9e005ed211: stress-ng.kill.ops_per_sec 4.6% regression

[Chen, Yu C]

On 6/11/2026 10:19 AM, kernel test robot wrote:
> 
> 
> Hello,
> 
> kernel test robot noticed a 4.6% regression of stress-ng.kill.ops_per_sec on:
> 
> 
> commit: 9e005ed21152d4a4bb0ceea71045ff8a642a6feb ("sched/topology: Allow multiple domains to claim sched_domain_shared")
> https://git.kernel.org/cgit/linux/kernel/git/tip/tip.git sched/core
> 
> 
> testcase: stress-ng
> config: x86_64-rhel-9.4
> compiler: gcc-14
> test machine: 256 threads 4 sockets INTEL(R) XEON(R) PLATINUM 8592+ (Emerald Rapids) with 256G memory
> parameters:
> 
> 	nr_threads: 100%
> 	testtime: 60s
> 	test: kill
> 	cpufreq_governor: performance
> 
> 

This 4.6% difference is very small and within normal run-to-run variance 
IMO, as
the commit has no functional changes.

BTW the commit log indicates nr_idle_scan is
initialized to a stale value 512, sds->alloc_flags = SD_SHARE_LLC, the 
wakeup
before the next period LB in this LLC domain might have slight higher 
latency though.
But consider your test last for 60 seconds which is much longer than the 
first period LB,
this should not be a problem.

thanks,
Chenyu