FYI, we noticed a -16.5% regression of netperf.Throughput_Mbps due to commit: commit f85cb5028c273692b2ac4f7f7e73f77ad2d5a3c2 ("mm, page_alloc: remove fair zone allocation policy") https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git master in testcase: netperf on test machine: 16 threads Broadwell-DE with 8G memory with following parameters: ip: ipv4 runtime: 900s nr_threads: 25% cluster: cs-localhost test: TCP_STREAM cpufreq_governor: performance 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. Details are as below: --------------------------------------------------------------------------------------------------> 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 ========================================================================================= cluster/compiler/cpufreq_governor/ip/kconfig/nr_threads/rootfs/runtime/tbox_group/test/testcase: cs-localhost/gcc-6/performance/ipv4/x86_64-rhel/25%/debian-x86_64-2015-02-07.cgz/900s/lkp-bdw-de1/TCP_STREAM/netperf commit: 23c17ad07f ("mm, vmscan: add classzone information to tracepoints") f85cb5028c ("mm, page_alloc: remove fair zone allocation policy") 23c17ad07fba9303 f85cb5028c273692b2ac4f7f7e ---------------- -------------------------- fail:runs %reproduction fail:runs | | | :4 25% 1:4 kmsg.i2c_i2c-#:sendbytes:NAK_bailout 1:4 -25% :4 kmsg.usb#-#:can't_read_configurations,error 1:4 -25% :4 kmsg.usb#-#:unable_to_read_config_index#descriptor/all %stddev %change %stddev \ | \ 29019 ± 1% -16.5% 24223 ± 2% netperf.Throughput_Mbps 12798 ± 6% -17.2% 10593 ± 9% netperf.time.involuntary_context_switches 334.50 ± 1% -13.4% 289.75 ± 1% netperf.time.percent_of_cpu_this_job_got 2941 ± 1% -13.8% 2536 ± 1% netperf.time.system_time 1491248 ± 7% +58.7% 2366463 ± 5% netperf.time.voluntary_context_switches 1896 ± 1% +9.6% 2077 ± 0% slabinfo.kmalloc-2048.active_objs 16.91 ± 4% +9.8% 18.57 ± 2% perf-profile.cycles-pp.copy_user_enhanced_fast_string.tcp_sendmsg.inet_sendmsg.sock_sendmsg.SYSC_sendto 0.83 ± 11% +24.0% 1.03 ± 10% perf-profile.cycles-pp.sk_stream_alloc_skb.tcp_sendmsg.inet_sendmsg.sock_sendmsg.SYSC_sendto 238789 ± 11% -76.8% 55338 ± 49% vmstat.system.cs 16823 ± 0% -1.5% 16570 ± 0% vmstat.system.in 1.884e+08 ± 5% -41.1% 1.11e+08 ± 10% softirqs.NET_RX 218456 ± 0% +18.4% 258550 ± 1% softirqs.RCU 1105427 ± 1% -11.8% 975484 ± 1% softirqs.SCHED 497245 ± 0% -89.0% 54772 ± 1% meminfo.Active 8511 ± 0% +543.5% 54765 ± 1% meminfo.Active(anon) 488733 ± 0% -100.0% 6.00 ± 0% meminfo.Active(file) 54898 ± 1% +790.2% 488733 ± 0% meminfo.Inactive(file) 2.098e+08 ± 12% -79.0% 44033424 ± 52% cpuidle.C1-BDW.time 1.011e+08 ± 11% -79.0% 21193683 ± 55% cpuidle.C1-BDW.usage 3.694e+08 ± 8% +91.5% 7.074e+08 ± 5% cpuidle.C3-BDW.time 1104487 ± 7% +81.0% 1999452 ± 5% cpuidle.C3-BDW.usage 83514559 ± 8% -34.1% 55042921 ± 10% cpuidle.POLL.time 2178797 ± 11% -79.1% 454769 ± 56% cpuidle.POLL.usage 2183 ± 2% -100.0% 0.00 ± -1% proc-vmstat.nr_alloc_batch 3.936e+08 ± 1% -15.7% 3.319e+08 ± 2% proc-vmstat.numa_hit 3.936e+08 ± 1% -15.7% 3.319e+08 ± 2% proc-vmstat.numa_local 7.693e+08 ± 1% -100.0% 0.00 ± 0% proc-vmstat.pgalloc_dma32 2.366e+09 ± 1% +11.5% 2.639e+09 ± 2% proc-vmstat.pgalloc_normal 3.135e+09 ± 1% -15.8% 2.639e+09 ± 2% proc-vmstat.pgfree 46.62 ± 0% -3.6% 44.94 ± 0% turbostat.%Busy 1165 ± 0% -3.6% 1124 ± 0% turbostat.Avg_MHz 1.09 ± 8% +77.2% 1.93 ± 4% turbostat.CPU%c3 10.71 ± 2% +43.7% 15.38 ± 5% turbostat.CPU%c6 39.17 ± 0% -3.9% 37.66 ± 0% turbostat.PkgWatt 19.89 ± 4% +20.1% 23.89 ± 2% turbostat.RAMWatt 1.06e+12 ± 4% -33.3% 7.073e+11 ± 9% perf-stat.branch-instructions 0.40 ± 3% +30.9% 0.52 ± 4% perf-stat.branch-miss-rate 4.228e+09 ± 1% -13.0% 3.68e+09 ± 4% perf-stat.branch-misses 8.141e+11 ± 2% -17.3% 6.732e+11 ± 4% perf-stat.cache-misses 8.141e+11 ± 2% -17.3% 6.732e+11 ± 4% perf-stat.cache-references 2.156e+08 ± 11% -76.8% 49969421 ± 49% perf-stat.context-switches 62562 ± 1% -7.4% 57921 ± 0% perf-stat.cpu-migrations 1.651e+13 ± 1% -3.2% 1.598e+13 ± 2% perf-stat.cycles 0.03 ± 4% +36.5% 0.04 ± 6% perf-stat.dTLB-load-miss-rate 2.614e+12 ± 3% -28.4% 1.871e+12 ± 5% perf-stat.dTLB-loads 9.526e+08 ± 25% -37.7% 5.935e+08 ± 17% perf-stat.dTLB-store-misses 1.958e+12 ± 3% -26.9% 1.432e+12 ± 4% perf-stat.dTLB-stores 1.169e+09 ± 9% -52.6% 5.539e+08 ± 15% perf-stat.iTLB-load-misses 1.186e+13 ± 5% -35.1% 7.702e+12 ± 9% perf-stat.instructions 10183 ± 3% +37.8% 14031 ± 6% perf-stat.instructions-per-iTLB-miss 0.72 ± 6% -33.0% 0.48 ± 8% perf-stat.ipc 246878 ± 25% -77.2% 56283 ±141% sched_debug.cfs_rq:/.MIN_vruntime.max 63597 ± 28% -74.7% 16086 ±146% sched_debug.cfs_rq:/.MIN_vruntime.stddev 0.77 ± 23% -36.7% 0.48 ± 32% sched_debug.cfs_rq:/.load_avg.min 246878 ± 25% -77.2% 56283 ±141% sched_debug.cfs_rq:/.max_vruntime.max 63597 ± 28% -74.7% 16086 ±146% sched_debug.cfs_rq:/.max_vruntime.stddev 322631 ± 2% -13.0% 280763 ± 6% sched_debug.cfs_rq:/.min_vruntime.stddev 320721 ± 33% +44.3% 462865 ± 21% sched_debug.cfs_rq:/.spread0.max 322633 ± 2% -13.0% 280765 ± 6% sched_debug.cfs_rq:/.spread0.stddev 657860 ± 3% +12.2% 738159 ± 1% sched_debug.cpu.avg_idle.avg 84895 ± 24% +239.0% 287807 ± 13% sched_debug.cpu.avg_idle.min 357850 ± 4% -20.4% 284765 ± 3% sched_debug.cpu.avg_idle.stddev 70.30 ± 3% +12.2% 78.85 ± 1% sched_debug.cpu.cpu_load[1].avg 70.58 ± 3% +11.9% 79.00 ± 1% sched_debug.cpu.cpu_load[2].avg 84932 ± 1% -9.1% 77174 ± 3% sched_debug.cpu.nr_load_updates.stddev 7370152 ± 5% -75.2% 1830136 ± 61% sched_debug.cpu.nr_switches.avg 14180734 ± 8% -71.7% 4014680 ± 60% sched_debug.cpu.nr_switches.max 1787548 ± 7% -81.2% 335812 ± 63% sched_debug.cpu.nr_switches.min 4217040 ± 9% -73.3% 1124694 ± 63% sched_debug.cpu.nr_switches.stddev 0.32 ± 23% -58.5% 0.13 ± 49% sched_debug.rt_rq:/.rt_time.max netperf.Throughput_Mbps 35000 ++------------------------------------------------------------------+ | .* .* .* *. | 30000 ++ * ** ***.**** ***.* * ** : * ** **** * *.** | : : : : : : : : : :* | 25000 OO OO O OO:O OO O O O : : : : : : : | | : O O OO OOO OOO : :: : : :: : : : | 20000 ++:: : : :: : : :: : : : | | :: : : :: : : :: : : : | 15000 ++: : : : :: : : :: : : : | | : : : :: : : :: : : :: | 10000 ++: : : :: : : :: : : :: | | : : : :: : : :: : : : | 5000 ++ : : :: : : :: : : : | |: : : :: : : :: : : : | 0 **--***-**-------------------**--***-**----**--****-*---------*-----+ [*] bisect-good sample [O] bisect-bad sample Thanks, Xiaolong