FYI, we noticed a -3.8% regression of netperf.Throughput_Mbps due to commit: commit 99a544df069c325e161d4f5246586161d84e9e82 ("timer: Switch to a non cascading wheel") https://github.com/yyu168/linux.git WIP.timers in testcase: netperf on test machine: 48 threads Ivytown Ivy Bridge-EP with 64G memory with following parameters: cluster=cs-localhost/cpufreq_governor=performance/ip=ipv4/nr_threads=200%/runtime=300s/test=SCTP_STREAM 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-4.9/performance/ipv4/x86_64-rhel/200%/debian-x86_64-2015-02-07.cgz/300s/ivb43/SCTP_STREAM/netperf commit: 672573b753 ("timer: Reduce the CPU index space to 256k") 99a544df06 ("timer: Switch to a non cascading wheel") 672573b753e951a6 99a544df069c325e161d4f5246 ---------------- -------------------------- fail:runs %reproduction fail:runs | | | :4 25% 1:4 kmsg.DHCP/BOOTP:Reply_not_for_us,op[#]xid[#] 1:4 -25% :4 kmsg.Spurious_LAPIC_timer_interrupt_on_cpu %stddev %change %stddev \ | \ 4.25 ± 0% -3.8% 4.09 ± 0% netperf.Throughput_Mbps 73653 ± 0% -2.9% 71503 ± 0% netperf.time.voluntary_context_switches 748.50 ± 9% -22.5% 580.25 ± 3% cpuidle.POLL.usage 2745 ± 2% -5.1% 2605 ± 1% vmstat.system.cs 5.58 ± 0% +0.8% 5.62 ± 0% turbostat.%Busy 86.00 ± 0% +2.0% 87.75 ± 0% turbostat.Avg_MHz 1265 ± 6% -8.4% 1158 ± 4% sched_debug.cfs_rq:/.min_vruntime.avg 3948 ± 2% -8.7% 3603 ± 1% sched_debug.cpu.nr_load_updates.stddev 0.23 ± 1% +8.0% 0.25 ± 3% sched_debug.cpu.nr_running.stddev 4.66 ± 7% +29.9% 6.05 ± 7% perf-profile.cycles-pp.__tick_nohz_idle_enter.tick_nohz_irq_exit.irq_exit.smp_apic_timer_interrupt.apic_timer_interrupt 2.29 ± 27% -26.2% 1.69 ± 5% perf-profile.cycles-pp.call_cpuidle.cpu_startup_entry.rest_init.start_kernel.x86_64_start_reservations 2.29 ± 27% -26.2% 1.69 ± 5% perf-profile.cycles-pp.cpuidle_enter.call_cpuidle.cpu_startup_entry.rest_init.start_kernel 1.33 ± 3% -20.5% 1.06 ± 15% perf-profile.cycles-pp.cpuidle_enter_state.cpuidle_enter.call_cpuidle.cpu_startup_entry.rest_init 0.00 ± -1% +Inf% 1.58 ± 18% perf-profile.cycles-pp.get_next_timer_interrupt.tick_nohz_stop_sched_tick.__tick_nohz_idle_enter.tick_nohz_irq_exit.irq_exit 8.03 ± 5% +16.3% 9.34 ± 3% perf-profile.cycles-pp.irq_exit.smp_apic_timer_interrupt.apic_timer_interrupt.cpuidle_enter.call_cpuidle 2.73 ± 44% -31.3% 1.87 ± 1% perf-profile.cycles-pp.rest_init.start_kernel.x86_64_start_reservations.x86_64_start_kernel 2.73 ± 44% -31.3% 1.87 ± 1% perf-profile.cycles-pp.start_kernel.x86_64_start_reservations.x86_64_start_kernel 4.83 ± 6% +29.9% 6.28 ± 6% perf-profile.cycles-pp.tick_nohz_irq_exit.irq_exit.smp_apic_timer_interrupt.apic_timer_interrupt.cpuidle_enter 3.53 ± 8% +38.3% 4.88 ± 7% perf-profile.cycles-pp.tick_nohz_stop_sched_tick.__tick_nohz_idle_enter.tick_nohz_irq_exit.irq_exit.smp_apic_timer_interrupt 2.73 ± 44% -31.3% 1.87 ± 1% perf-profile.cycles-pp.x86_64_start_kernel 2.73 ± 44% -31.3% 1.87 ± 1% perf-profile.cycles-pp.x86_64_start_reservations.x86_64_start_kernel 3.797e+10 ± 1% +6.7% 4.052e+10 ± 3% perf-stat.L1-dcache-loads 2.954e+08 ± 1% +9.0% 3.219e+08 ± 2% perf-stat.L1-dcache-prefetch-misses 1.406e+09 ± 2% +4.3% 1.466e+09 ± 1% perf-stat.LLC-prefetches 1.897e+10 ± 3% +16.1% 2.202e+10 ± 3% perf-stat.branch-instructions 2.176e+09 ± 3% +5.2% 2.29e+09 ± 3% perf-stat.branch-load-misses 1.84e+10 ± 3% +17.5% 2.161e+10 ± 4% perf-stat.branch-loads 4.727e+10 ± 0% +1.7% 4.805e+10 ± 1% perf-stat.bus-cycles 831778 ± 2% -5.0% 789901 ± 1% perf-stat.context-switches 9.632e+11 ± 1% +3.5% 9.97e+11 ± 0% perf-stat.cpu-cycles 174418 ± 0% -1.7% 171529 ± 0% perf-stat.cpu-migrations 3.815e+10 ± 2% +8.2% 4.129e+10 ± 5% perf-stat.dTLB-loads 20335861 ± 1% -16.1% 17057586 ± 2% perf-stat.iTLB-loads 9.431e+10 ± 2% +14.8% 1.083e+11 ± 2% perf-stat.instructions 3.957e+08 ± 0% -6.2% 3.711e+08 ± 2% perf-stat.node-loads netperf.Throughput_Mbps 4.26 ++-------------------------------------------------------------------+ *.*.*.*..*.*.*.*.*.*..*.*.*.*.*.*..*.*.*.*.*.*.*..*.*.*.*.*.*..*.*.*.* 4.24 ++ | 4.22 ++ | | | 4.2 ++ | 4.18 ++ | | | 4.16 ++ | 4.14 ++ | | | 4.12 ++ | 4.1 O+O O O O O O | | O O O O O O O O O O O O O O O O O O O O O O O O O 4.08 ++---------------O---------------------------------------------------+ netperf.time.voluntary_context_switches 74500 ++-----------------*------------------------------------------------+ | : + .*.. .*. | 74000 *+*. .*.. .*. : * *. .*.*.* *.. .*.*.*. .*..*. .* 73500 ++ * *.* *.* *.* *.* *.* *.*.* | | | 73000 ++ | 72500 ++ | | | 72000 ++ O O O O | 71500 ++ O O O O O O | | O O O O O O O O O O O O O O O O 71000 ++ | 70500 ++ | O O O O O O O | 70000 ++------------------------------------------------------------------+ [*] bisect-good sample [O] bisect-bad sample Thanks, Xiaolong