Greeting, FYI, we noticed a 2.0% improvement of will-it-scale.per_process_ops due to commit: commit: 80340f8d5f7423c4611fa6e9b07435006aa5834f ("sched: Add saved_state for tasks blocked on sleeping locks") https://git.kernel.org/cgit/linux/kernel/git/bigeasy/staging.git rtmutex in testcase: will-it-scale on test machine: 192 threads Intel(R) Xeon(R) Platinum 9242 CPU @ 2.30GHz with 192G memory with following parameters: nr_task: 16 mode: process test: poll1 cpufreq_governor: performance ucode: 0x5002f01 test-description: Will It Scale takes a testcase and runs it from 1 through to n parallel copies to see if the testcase will scale. It builds both a process and threads based test in order to see any differences between the two. test-url: https://github.com/antonblanchard/will-it-scale Details are as below: --------------------------------------------------------------------------------------------------> To reproduce: git clone https://github.com/intel/lkp-tests.git cd lkp-tests bin/lkp install job.yaml # job file is attached in this email bin/lkp run job.yaml ========================================================================================= compiler/cpufreq_governor/kconfig/mode/nr_task/rootfs/tbox_group/test/testcase/ucode: gcc-9/performance/x86_64-rhel-8.3/process/16/debian-10.4-x86_64-20200603.cgz/lkp-csl-2ap3/poll1/will-it-scale/0x5002f01 commit: 35398ef3be ("locking/rtmutex: export lockdep-less version of rt_mutex's lock, trylock and unlock") 80340f8d5f ("sched: Add saved_state for tasks blocked on sleeping locks") 35398ef3be1827cb 80340f8d5f7423c4611fa6e9b07 ---------------- --------------------------- %stddev %change %stddev \ | \ 6677815 +2.0% 6813971 will-it-scale.per_process_ops 1.068e+08 +2.0% 1.09e+08 will-it-scale.workload 467077 ± 42% +4879.3% 23257159 ± 98% cpuidle.C6.usage 1295308 ± 2% -12.5% 1133062 ± 4% meminfo.DirectMap4k 4.13 ± 8% -1.0 3.14 ± 9% perf-profile.calltrace.cycles-pp.__fget_light.do_sys_poll.__x64_sys_poll.do_syscall_64.entry_SYSCALL_64_after_hwframe 4.13 ± 8% -1.0 3.14 ± 9% perf-profile.children.cycles-pp.__fget_light 3.98 ± 8% -1.0 2.99 ± 9% perf-profile.self.cycles-pp.__fget_light 30899 -1.6% 30390 proc-vmstat.nr_slab_reclaimable 70023 -3.0% 67946 proc-vmstat.nr_slab_unreclaimable 63981 +5.0% 67159 ± 2% proc-vmstat.pgreuse 1927180 ± 12% +22.2% 2355975 ± 3% sched_debug.cfs_rq:/.spread0.max -566638 -75.3% -140233 sched_debug.cfs_rq:/.spread0.min 1064 ± 7% +12.1% 1193 ± 6% sched_debug.cpu.nr_switches.min 40639 ± 4% -29.6% 28615 ± 13% sched_debug.cpu.ttwu_count.max 3004 ± 4% -20.3% 2394 ± 9% sched_debug.cpu.ttwu_count.stddev 34941 ± 3% +16.1% 40583 ± 3% softirqs.CPU0.SCHED 20504 ± 9% -26.3% 15122 ± 31% softirqs.CPU109.RCU 4872 +369.6% 22879 ± 78% softirqs.CPU109.SCHED 20332 ± 9% -28.8% 14481 ± 22% softirqs.CPU111.RCU 12995 ± 10% +25.9% 16362 ± 9% softirqs.CPU41.RCU 38366 +6.2% 40738 ± 5% softirqs.CPU43.SCHED 38584 +5.8% 40826 ± 4% softirqs.CPU68.SCHED 36431 ± 4% +10.0% 40056 ± 3% softirqs.CPU73.SCHED 12373 ± 8% +12.7% 13949 ± 8% softirqs.CPU93.RCU 11975 ± 15% -42.0% 6943 ± 20% softirqs.CPU96.SCHED 3409 ± 82% +2989.8% 105337 ± 54% numa-meminfo.node0.AnonHugePages 28904 ± 12% +485.1% 169104 ± 43% numa-meminfo.node0.AnonPages 283677 +12.5% 319149 ± 6% numa-meminfo.node0.FilePages 30171 ± 10% +468.3% 171475 ± 41% numa-meminfo.node0.Inactive 30171 ± 10% +468.3% 171475 ± 41% numa-meminfo.node0.Inactive(anon) 7508 ± 3% +12.9% 8477 ± 6% numa-meminfo.node0.KernelStack 707950 ± 5% +23.6% 874864 ± 7% numa-meminfo.node0.MemUsed 1726 ± 8% +89.8% 3276 ± 16% numa-meminfo.node0.PageTables 281724 +11.2% 313355 ± 5% numa-meminfo.node0.Unevictable 11159 ± 22% -38.1% 6909 ± 3% numa-meminfo.node1.Mapped 82531 ± 67% -90.9% 7522 ±155% numa-meminfo.node3.AnonHugePages 123663 ± 56% -82.0% 22306 ± 67% numa-meminfo.node3.AnonPages 126435 ± 56% -78.3% 27463 ± 61% numa-meminfo.node3.Inactive 126435 ± 56% -78.3% 27463 ± 61% numa-meminfo.node3.Inactive(anon) 808823 ± 12% -20.2% 645340 ± 7% numa-meminfo.node3.MemUsed 7234 ± 12% +484.5% 42283 ± 43% numa-vmstat.node0.nr_anon_pages 70919 +12.5% 79786 ± 6% numa-vmstat.node0.nr_file_pages 7551 ± 10% +467.8% 42876 ± 41% numa-vmstat.node0.nr_inactive_anon 7508 ± 3% +12.9% 8476 ± 6% numa-vmstat.node0.nr_kernel_stack 1763 +30.2% 2295 ± 28% numa-vmstat.node0.nr_mapped 431.50 ± 8% +89.5% 817.50 ± 16% numa-vmstat.node0.nr_page_table_pages 70430 +11.2% 78338 ± 5% numa-vmstat.node0.nr_unevictable 7551 ± 10% +467.8% 42876 ± 41% numa-vmstat.node0.nr_zone_inactive_anon 70430 +11.2% 78338 ± 5% numa-vmstat.node0.nr_zone_unevictable 408102 ± 10% +68.0% 685464 ± 20% numa-vmstat.node0.numa_local 2822 ± 22% -37.7% 1759 ± 6% numa-vmstat.node1.nr_mapped 30923 ± 56% -82.0% 5581 ± 67% numa-vmstat.node3.nr_anon_pages 31618 ± 56% -78.3% 6874 ± 61% numa-vmstat.node3.nr_inactive_anon 31618 ± 56% -78.3% 6874 ± 61% numa-vmstat.node3.nr_zone_inactive_anon 59517 ± 34% +110.1% 125034 numa-vmstat.node3.numa_other 7889 ± 7% -16.0% 6625 ± 5% slabinfo.Acpi-State.active_objs 7889 ± 7% -16.0% 6625 ± 5% slabinfo.Acpi-State.num_objs 2614 ± 6% -16.8% 2175 ± 10% slabinfo.PING.active_objs 2614 ± 6% -16.8% 2175 ± 10% slabinfo.PING.num_objs 5648 ± 9% -16.3% 4729 ± 8% slabinfo.files_cache.active_objs 5648 ± 9% -16.3% 4729 ± 8% slabinfo.files_cache.num_objs 1922 ± 5% -14.9% 1635 ± 12% slabinfo.khugepaged_mm_slot.active_objs 1922 ± 5% -14.9% 1635 ± 12% slabinfo.khugepaged_mm_slot.num_objs 5298 ± 4% -16.3% 4434 ± 8% slabinfo.mm_struct.active_objs 5298 ± 4% -16.3% 4434 ± 8% slabinfo.mm_struct.num_objs 12456 ± 9% -19.5% 10030 ± 17% slabinfo.pde_opener.active_objs 12456 ± 9% -19.5% 10030 ± 17% slabinfo.pde_opener.num_objs 3556 ± 3% -9.7% 3212 ± 6% slabinfo.sighand_cache.num_objs 5417 ± 5% -10.9% 4828 ± 5% slabinfo.signal_cache.num_objs 4544 ± 4% -11.5% 4022 ± 7% slabinfo.sock_inode_cache.active_objs 4544 ± 4% -11.5% 4022 ± 7% slabinfo.sock_inode_cache.num_objs 8411 ± 7% -13.5% 7276 ± 6% slabinfo.task_delay_info.active_objs 8411 ± 7% -13.5% 7276 ± 6% slabinfo.task_delay_info.num_objs 0.28 +135.9% 0.67 ± 93% perf-stat.i.MPKI 1.346e+10 +2.1% 1.375e+10 perf-stat.i.branch-instructions 61401177 +11.6% 68508179 ± 14% perf-stat.i.branch-misses 4100686 +23.1% 5047396 ± 21% perf-stat.i.cache-misses 18651514 +142.1% 45154376 ± 94% perf-stat.i.cache-references 5.593e+10 +2.1% 5.71e+10 perf-stat.i.cpu-cycles 13932 -14.2% 11959 ± 17% perf-stat.i.cycles-between-cache-misses 1.709e+10 +1.6% 1.737e+10 perf-stat.i.dTLB-loads 1.302e+10 +2.0% 1.328e+10 perf-stat.i.dTLB-stores 6.662e+10 +2.1% 6.802e+10 perf-stat.i.instructions 0.29 +2.1% 0.30 perf-stat.i.metric.GHz 0.84 ± 2% +25.9% 1.06 ± 16% perf-stat.i.metric.K/sec 227.02 +2.0% 231.48 perf-stat.i.metric.M/sec 92.68 -3.0 89.71 ± 2% perf-stat.i.node-load-miss-rate% 97.55 -8.0 89.54 ± 7% perf-stat.i.node-store-miss-rate% 3947 ± 8% +211.0% 12275 ± 60% perf-stat.i.node-stores 0.28 +137.3% 0.66 ± 94% perf-stat.overall.MPKI 13639 -14.0% 11729 ± 16% perf-stat.overall.cycles-between-cache-misses 92.39 -3.1 89.32 ± 2% perf-stat.overall.node-load-miss-rate% 1.342e+10 +2.1% 1.37e+10 perf-stat.ps.branch-instructions 61207914 +11.6% 68287200 ± 14% perf-stat.ps.branch-misses 4087407 +23.1% 5030794 ± 21% perf-stat.ps.cache-misses 18592626 +142.1% 45003748 ± 94% perf-stat.ps.cache-references 5.574e+10 +2.1% 5.691e+10 perf-stat.ps.cpu-cycles 1.703e+10 +1.6% 1.731e+10 perf-stat.ps.dTLB-loads 1.297e+10 +2.0% 1.324e+10 perf-stat.ps.dTLB-stores 6.64e+10 +2.1% 6.779e+10 perf-stat.ps.instructions 3944 ± 7% +210.4% 12241 ± 60% perf-stat.ps.node-stores 2.003e+13 +2.3% 2.048e+13 perf-stat.total.instructions 1726 ± 23% +25.6% 2168 ± 5% interrupts.CPU0.CAL:Function_call_interrupts 140.00 ± 22% -63.0% 51.75 ± 52% interrupts.CPU0.RES:Rescheduling_interrupts 1803 ± 18% +52.9% 2757 ± 16% interrupts.CPU115.CAL:Function_call_interrupts 93.25 ± 25% +41.6% 132.00 ± 12% interrupts.CPU120.NMI:Non-maskable_interrupts 93.25 ± 25% +41.6% 132.00 ± 12% interrupts.CPU120.PMI:Performance_monitoring_interrupts 113.25 ± 9% +37.7% 156.00 ± 23% interrupts.CPU121.NMI:Non-maskable_interrupts 113.25 ± 9% +37.7% 156.00 ± 23% interrupts.CPU121.PMI:Performance_monitoring_interrupts 90.00 ± 25% +39.7% 125.75 ± 12% interrupts.CPU139.NMI:Non-maskable_interrupts 90.00 ± 25% +39.7% 125.75 ± 12% interrupts.CPU139.PMI:Performance_monitoring_interrupts 94.00 ± 26% +41.5% 133.00 ± 12% interrupts.CPU144.NMI:Non-maskable_interrupts 94.00 ± 26% +41.5% 133.00 ± 12% interrupts.CPU144.PMI:Performance_monitoring_interrupts 100.25 ± 26% +42.9% 143.25 ± 22% interrupts.CPU168.NMI:Non-maskable_interrupts 100.25 ± 26% +42.9% 143.25 ± 22% interrupts.CPU168.PMI:Performance_monitoring_interrupts 107.00 ± 7% +15.9% 124.00 ± 4% interrupts.CPU17.NMI:Non-maskable_interrupts 107.00 ± 7% +15.9% 124.00 ± 4% interrupts.CPU17.PMI:Performance_monitoring_interrupts 109.00 ± 7% +17.7% 128.25 ± 11% interrupts.CPU189.NMI:Non-maskable_interrupts 109.00 ± 7% +17.7% 128.25 ± 11% interrupts.CPU189.PMI:Performance_monitoring_interrupts 18.75 ± 89% +1310.7% 264.50 ± 60% interrupts.CPU19.RES:Rescheduling_interrupts 1.25 ±131% +43040.0% 539.25 ±168% interrupts.CPU22.RES:Rescheduling_interrupts 106.25 ± 4% +23.5% 131.25 ± 12% interrupts.CPU24.NMI:Non-maskable_interrupts 106.25 ± 4% +23.5% 131.25 ± 12% interrupts.CPU24.PMI:Performance_monitoring_interrupts 107.25 ± 3% +19.3% 128.00 ± 14% interrupts.CPU26.NMI:Non-maskable_interrupts 107.25 ± 3% +19.3% 128.00 ± 14% interrupts.CPU26.PMI:Performance_monitoring_interrupts 107.75 ± 2% +19.3% 128.50 ± 9% interrupts.CPU27.NMI:Non-maskable_interrupts 107.75 ± 2% +19.3% 128.50 ± 9% interrupts.CPU27.PMI:Performance_monitoring_interrupts 111.25 ± 2% +16.6% 129.75 ± 9% interrupts.CPU28.NMI:Non-maskable_interrupts 111.25 ± 2% +16.6% 129.75 ± 9% interrupts.CPU28.PMI:Performance_monitoring_interrupts 108.25 ± 10% +19.4% 129.25 ± 15% interrupts.CPU29.NMI:Non-maskable_interrupts 108.25 ± 10% +19.4% 129.25 ± 15% interrupts.CPU29.PMI:Performance_monitoring_interrupts 90.75 ± 25% +47.4% 133.75 ± 20% interrupts.CPU30.NMI:Non-maskable_interrupts 90.75 ± 25% +47.4% 133.75 ± 20% interrupts.CPU30.PMI:Performance_monitoring_interrupts 104.50 +20.6% 126.00 ± 12% interrupts.CPU31.NMI:Non-maskable_interrupts 104.50 +20.6% 126.00 ± 12% interrupts.CPU31.PMI:Performance_monitoring_interrupts 103.25 ± 5% +34.1% 138.50 ± 25% interrupts.CPU32.NMI:Non-maskable_interrupts 103.25 ± 5% +34.1% 138.50 ± 25% interrupts.CPU32.PMI:Performance_monitoring_interrupts 109.75 ± 9% +15.3% 126.50 ± 13% interrupts.CPU33.NMI:Non-maskable_interrupts 109.75 ± 9% +15.3% 126.50 ± 13% interrupts.CPU33.PMI:Performance_monitoring_interrupts 106.75 ± 8% +19.4% 127.50 ± 14% interrupts.CPU35.NMI:Non-maskable_interrupts 106.75 ± 8% +19.4% 127.50 ± 14% interrupts.CPU35.PMI:Performance_monitoring_interrupts 89.00 ± 25% +42.7% 127.00 ± 12% interrupts.CPU38.NMI:Non-maskable_interrupts 89.00 ± 25% +42.7% 127.00 ± 12% interrupts.CPU38.PMI:Performance_monitoring_interrupts 7661 ± 24% -38.5% 4711 ± 51% interrupts.CPU4.NMI:Non-maskable_interrupts 7661 ± 24% -38.5% 4711 ± 51% interrupts.CPU4.PMI:Performance_monitoring_interrupts 104.25 ± 5% +20.9% 126.00 ± 13% interrupts.CPU41.NMI:Non-maskable_interrupts 104.25 ± 5% +20.9% 126.00 ± 13% interrupts.CPU41.PMI:Performance_monitoring_interrupts 104.25 ± 5% +20.4% 125.50 ± 11% interrupts.CPU44.NMI:Non-maskable_interrupts 104.25 ± 5% +20.4% 125.50 ± 11% interrupts.CPU44.PMI:Performance_monitoring_interrupts 103.00 ± 5% +21.4% 125.00 ± 12% interrupts.CPU45.NMI:Non-maskable_interrupts 103.00 ± 5% +21.4% 125.00 ± 12% interrupts.CPU45.PMI:Performance_monitoring_interrupts 91.50 ± 28% +36.1% 124.50 ± 12% interrupts.CPU46.NMI:Non-maskable_interrupts 91.50 ± 28% +36.1% 124.50 ± 12% interrupts.CPU46.PMI:Performance_monitoring_interrupts 92.75 ± 29% +35.3% 125.50 ± 13% interrupts.CPU47.NMI:Non-maskable_interrupts 92.75 ± 29% +35.3% 125.50 ± 13% interrupts.CPU47.PMI:Performance_monitoring_interrupts 78.00 ± 33% +65.7% 129.25 ± 8% interrupts.CPU49.NMI:Non-maskable_interrupts 78.00 ± 33% +65.7% 129.25 ± 8% interrupts.CPU49.PMI:Performance_monitoring_interrupts 108.25 ± 5% +19.4% 129.25 ± 12% interrupts.CPU50.NMI:Non-maskable_interrupts 108.25 ± 5% +19.4% 129.25 ± 12% interrupts.CPU50.PMI:Performance_monitoring_interrupts 107.50 ± 2% +28.6% 138.25 ± 11% interrupts.CPU51.NMI:Non-maskable_interrupts 107.50 ± 2% +28.6% 138.25 ± 11% interrupts.CPU51.PMI:Performance_monitoring_interrupts 7660 ± 24% -38.5% 4710 ± 51% interrupts.CPU6.NMI:Non-maskable_interrupts 7660 ± 24% -38.5% 4710 ± 51% interrupts.CPU6.PMI:Performance_monitoring_interrupts 92.50 ± 22% +35.7% 125.50 ± 11% interrupts.CPU70.NMI:Non-maskable_interrupts 92.50 ± 22% +35.7% 125.50 ± 11% interrupts.CPU70.PMI:Performance_monitoring_interrupts 106.00 ± 7% +18.4% 125.50 ± 12% interrupts.CPU71.NMI:Non-maskable_interrupts 106.00 ± 7% +18.4% 125.50 ± 12% interrupts.CPU71.PMI:Performance_monitoring_interrupts 97.75 ± 25% +37.9% 134.75 ± 14% interrupts.CPU72.NMI:Non-maskable_interrupts 97.75 ± 25% +37.9% 134.75 ± 14% interrupts.CPU72.PMI:Performance_monitoring_interrupts 108.75 ± 8% +17.9% 128.25 ± 10% interrupts.CPU79.NMI:Non-maskable_interrupts 108.75 ± 8% +17.9% 128.25 ± 10% interrupts.CPU79.PMI:Performance_monitoring_interrupts 108.75 ± 7% +19.8% 130.25 ± 10% interrupts.CPU82.NMI:Non-maskable_interrupts 108.75 ± 7% +19.8% 130.25 ± 10% interrupts.CPU82.PMI:Performance_monitoring_interrupts 97.00 ± 27% +40.5% 136.25 ± 12% interrupts.CPU83.NMI:Non-maskable_interrupts 97.00 ± 27% +40.5% 136.25 ± 12% interrupts.CPU83.PMI:Performance_monitoring_interrupts 114.25 ± 4% +19.3% 136.25 ± 10% interrupts.CPU95.NMI:Non-maskable_interrupts 114.25 ± 4% +19.3% 136.25 ± 10% interrupts.CPU95.PMI:Performance_monitoring_interrupts 2851 ± 7% +73.3% 4940 ± 19% interrupts.RES:Rescheduling_interrupts will-it-scale.per_process_ops 6.9e+06 +----------------------------------------------------------------+ | O O | | OO O O | 6.85e+06 |-+ O O O | | O O O O O O O OO O O O | | O O O O | 6.8e+06 |-+ O | | | 6.75e+06 |-+ | | | | +.+. .+. .+.+.+.+.+.+ | 6.7e+06 |-+ .+ + + : | | .+. .+ : .+. .+. .+. .+. .+ .+.+.+.+.| |.+ + ++ + +.+ + + + | 6.65e+06 +----------------------------------------------------------------+ [*] bisect-good sample [O] bisect-bad sample 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. Thanks, Rong Chen