From: Uladzislau Rezki <urezki@gmail.com>
To: kernel test robot <oliver.sang@intel.com>
Cc: Uladzislau Rezki <urezki@gmail.com>,
oe-lkp@lists.linux.dev, lkp@intel.com,
Linux Memory Management List <linux-mm@kvack.org>,
Andrew Morton <akpm@linux-foundation.org>,
Baoquan He <bhe@redhat.com>, Christoph Hellwig <hch@lst.de>,
Dave Chinner <david@fromorbit.com>,
Joel Fernandes <joel@joelfernandes.org>,
Kazuhito Hagio <k-hagio-ab@nec.com>,
"Liam R. Howlett" <Liam.Howlett@oracle.com>,
Lorenzo Stoakes <lstoakes@gmail.com>,
Matthew Wilcox <willy@infradead.org>,
Oleksiy Avramchenko <oleksiy.avramchenko@sony.com>,
"Paul E. McKenney" <paulmck@kernel.org>,
ying.huang@intel.com, feng.tang@intel.com, fengwei.yin@intel.com
Subject: Re: [linux-next:master] [mm] 8f33a2ff30: stress-ng.resched.ops_per_sec -10.3% regression
Date: Mon, 4 Mar 2024 10:06:58 +0100 [thread overview]
Message-ID: <ZeWPMmwjWjGRsXMj@pc636> (raw)
In-Reply-To: <202402292306.8520763a-oliver.sang@intel.com>
Hello.
>
> Hello,
>
> kernel test robot noticed a -10.3% regression of stress-ng.resched.ops_per_sec on:
>
>
> commit: 8f33a2ff307248c3e55a7696f60b3658b28edb57 ("mm: vmalloc: set nr_nodes based on CPUs in a system")
> https://git.kernel.org/cgit/linux/kernel/git/next/linux-next.git master
>
> testcase: stress-ng
> test machine: 64 threads 2 sockets Intel(R) Xeon(R) Gold 6346 CPU @ 3.10GHz (Ice Lake) with 256G memory
> parameters:
>
> nr_threads: 100%
> testtime: 60s
> test: resched
> cpufreq_governor: performance
>
>
> In addition to that, the commit also has significant impact on the following tests:
>
> +------------------+-------------------------------------------------------------------------------------------+
> | testcase: change | stress-ng: stress-ng.pthread.ops_per_sec 23.0% improvement |
> | test machine | 64 threads 2 sockets Intel(R) Xeon(R) Gold 6346 CPU @ 3.10GHz (Ice Lake) with 256G memory |
> | test parameters | cpufreq_governor=performance |
> | | nr_threads=100% |
> | | test=pthread |
> | | testtime=60s |
> +------------------+-------------------------------------------------------------------------------------------+
> | testcase: change | stress-ng: stress-ng.fstat.ops_per_sec 14.2% improvement |
> | test machine | 64 threads 2 sockets Intel(R) Xeon(R) Gold 6346 CPU @ 3.10GHz (Ice Lake) with 256G memory |
> | test parameters | cpufreq_governor=performance |
> | | disk=1HDD |
> | | fs=xfs |
> | | nr_threads=100% |
> | | test=fstat |
> | | testtime=60s |
> +------------------+-------------------------------------------------------------------------------------------+
>
>
> 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/202402292306.8520763a-oliver.sang@intel.com
>
>
> Details are as below:
> -------------------------------------------------------------------------------------------------->
>
>
> The kernel config and materials to reproduce are available at:
> https://download.01.org/0day-ci/archive/20240229/202402292306.8520763a-oliver.sang@intel.com
>
> =========================================================================================
> compiler/cpufreq_governor/kconfig/nr_threads/rootfs/tbox_group/test/testcase/testtime:
> gcc-12/performance/x86_64-rhel-8.3/100%/debian-12-x86_64-20240206.cgz/lkp-icl-2sp7/resched/stress-ng/60s
>
> commit:
> 8e1d743f2c ("mm: vmalloc: support multiple nodes in vmallocinfo")
> 8f33a2ff30 ("mm: vmalloc: set nr_nodes based on CPUs in a system")
>
8e1d743f2c ("mm: vmalloc: support multiple nodes in vmallocinfo") this
commit has nothing to do with this test.
>
> 8e1d743f2c2671aa 8f33a2ff307248c3e55a7696f60
> ---------------- ---------------------------
> %stddev %change %stddev
> \ | \
> 7.48 -0.8 6.73 mpstat.cpu.all.nice%
> 10439977 -10.4% 9351864 vmstat.system.cs
> 14670714 ± 3% +18.1% 17330709 ± 5% numa-numastat.node0.local_node
> 14688319 ± 3% +18.1% 17348214 ± 5% numa-numastat.node0.numa_hit
> 14538034 ± 3% +15.7% 16824234 ± 4% numa-numastat.node1.local_node
> 14556613 ± 3% +15.6% 16834659 ± 4% numa-numastat.node1.numa_hit
> 14685240 ± 3% +18.0% 17334251 ± 5% numa-vmstat.node0.numa_hit
> 14667635 ± 3% +18.1% 17316745 ± 5% numa-vmstat.node0.numa_local
> 14551744 ± 3% +15.6% 16815047 ± 4% numa-vmstat.node1.numa_hit
> 14533165 ± 3% +15.6% 16804623 ± 4% numa-vmstat.node1.numa_local
> 9.153e+08 -10.3% 8.208e+08 stress-ng.resched.ops
> 15220752 -10.3% 13651349 stress-ng.resched.ops_per_sec
> 6.584e+08 -10.8% 5.871e+08 stress-ng.time.involuntary_context_switches
>
I tested the "resched" use case on my setup to check the commit:
8f33a2ff30 ("mm: vmalloc: set nr_nodes based on CPUs in a system")
n=0; while [ $n -lt 20 ]; do stress-ng --timeout 60 --times --verify --metrics --no-rand-seed --resched 64; n=$(( $n + 1 )); done
1) One socket system 32 CPUS, 64 threads, 128G of memory:
(revert 8f33a2ff30) (with 8f33a2ff30)
resched bogo ops/s resched bogo ops/s resched diff %
1105043856 18404843 1110469441 18491268 -0.49
1094766811 18231572 1117884383 18616359 -2.11
1103621287 18376740 1105661054 18411893 -0.18
1079532022 17973123 1101247950 18337844 -2.01
1099874899 18316050 1089695381 18144556 0.93
1076430974 17921542 1074824321 17899317 0.15
1071025136 17835263 1097552346 18276981 -2.48
1092038983 18182772 1103594553 18377955 -1.06
1099140652 18299703 1080602374 17994387 1.69
1100454122 18324364 1094512741 18227744 0.54
1092551777 18195189 1099387884 18305866 -0.63
1098877800 18297198 1095319518 18240721 0.32
1103042823 18366819 1086364199 18090137 1.51
1083722244 18046970 1073436871 17876677 0.95
1101988080 18350823 1080819704 17996891 1.92
1086171084 18087685 1080936227 17998387 0.48
1106178491 18419226 1078155643 17953565 2.53
1084124963 18053216 1087789728 18111601 -0.34
1076017418 17916972 1090240538 18153644 -1.32
1091438151 18174424 1094233215 18221998 -0.26
no difference.
2) Simulated a NUMA system same as your configuration, two nodes with
16 CPUs each, in total 64 threads:
Do not post result here since no difference.
--
Uladzislau Rezki
prev parent reply other threads:[~2024-03-04 9:07 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-02-29 16:01 [linux-next:master] [mm] 8f33a2ff30: stress-ng.resched.ops_per_sec -10.3% regression kernel test robot
2024-02-29 18:21 ` Uladzislau Rezki
2024-03-04 9:06 ` Uladzislau Rezki [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=ZeWPMmwjWjGRsXMj@pc636 \
--to=urezki@gmail.com \
--cc=Liam.Howlett@oracle.com \
--cc=akpm@linux-foundation.org \
--cc=bhe@redhat.com \
--cc=david@fromorbit.com \
--cc=feng.tang@intel.com \
--cc=fengwei.yin@intel.com \
--cc=hch@lst.de \
--cc=joel@joelfernandes.org \
--cc=k-hagio-ab@nec.com \
--cc=linux-mm@kvack.org \
--cc=lkp@intel.com \
--cc=lstoakes@gmail.com \
--cc=oe-lkp@lists.linux.dev \
--cc=oleksiy.avramchenko@sony.com \
--cc=oliver.sang@intel.com \
--cc=paulmck@kernel.org \
--cc=willy@infradead.org \
--cc=ying.huang@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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 read-only IMAP folder(s) and NNTP newsgroup(s).