LinuxLists.cc - [linus:master] [x86/bugs] 6613d82e61: stress-ng.mutex.ops_per

2024-03-04 06:00:22

Subject: [linus:master] [x86/bugs] 6613d82e61: stress-ng.mutex.ops_per_sec -7.9% regression

Hello,

kernel test robot
commit: 6613d82e617dd7eb8b
Details are as below:
--------------------------

The kernel config
noticed a -7.9% regression of stress-ng.mutex.ops_per_sec on:
0c40b2fe3acea655b1d611 ("x86/bugs: Use ALTERNATIVE() instead of mds_user_clear static key")
el.org/cgit/linux/kernel/git/torvalds/linux.git">https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master
threads 2 sockets Intel(R) Xeon(R) Gold 6346 CPU @ 3.10GHz (Ice Lake) with 256G memory
the commit also has impact on the following tests:
-------------------------------------------------------------------------------------+
| stress-ng: stress-ng.ptrace.ops_per_sec -3.9% regression |
| 64 threads 2 sockets Intel(R) Xeon(R) Gold 6346 CPU @ 3.10GHz (Ice Lake) with 256G memory |
| cpufreq_governor=performance |
|
|
|
-------------------------------------------------------------------------------------+
| stress-ng: stress-ng.getdent.ops_per_sec 5.8% improvement |
| 64 threads 2 sockets Intel(R) Xeon(R) Gold 6346 CPU @ 3.10GHz (Ice Lake) with 256G memory |
| cpufreq_governor=performance |
|
|
|
|
|
-------------------------------------------------------------------------------------+
| will-it-scale: will-it-scale.per_thread_ops 4.0% improvement |
| 104 threads 2 sockets (Skylake) with 192G memory |
| cpufreq_governor=performance |
|
|
|
-------------------------------------------------------------------------------------+
| will-it-scale: will-it-scale.per_process_ops -2.1% regression |
| 104 threads 2 sockets (Skylake) with 192G memory |
| cpufreq_governor=performance |
|
|
|
-------------------------------------------------------------------------------------+
| will-it-scale: will-it-scale.per_process_ops 3.7% improvement |
| 104 threads 2 sockets (Skylake) with 192G memory |
| cpufreq_governor=performance |
|
|
|
-------------------------------------------------------------------------------------+
in a separate patch/commit (i.e. not just a new version of
kindly add following tags
test robot <[email protected]>
//lore.kernel.org/oe-lkp/202403041300.a7fb1462-yujie.liu@intel.com">https://lore.kernel.org/oe-lkp/[email protected]
------------------------------------------------------------------------>
and materials to reproduce are available at:
.01.org/0day-ci/archive/20240304/202403041300.a7fb1462-yujie.liu@intel.com">https://download.01.org/0day-ci/archive/20240304/[email protected]
===============================================================
kconfig/nr_threads/rootfs/tbox_group/test/testcase/testtime:
4-rhel-8.3/100%/debian-12-x86_64-20240206.cgz/lkp-icl-2sp7/mutex/stress-ng/60s
Add VERW just before userspace transition")
Use ALTERNATIVE() instead of mds_user_clear static key")
6613d82e617dd7eb8b0c40b2fe3
---------------------------
%stddev
\
8755 ? 10% numa-meminfo.node0.Active
8702 ? 10% numa-meminfo.node0.Active(anon)
-8.0% 384591 vmstat.system.cs
-5.2% 273070 vmstat.system.in
+9.0% 199032 stress-ng.mutex.nanosecs_per_mutex
-7.9% 16702171 stress-ng.mutex.ops
-7.9% 278364 stress-ng.mutex.ops_per_sec
-7.3% 11161921 stress-ng.time.involuntary_context_switches
-7.6% 8707796 stress-ng.time.voluntary_context_switches
1.28 perf-stat.i.MPKI
0.30 perf-stat.i.branch-miss-rate%
-4.4% 10944702 perf-stat.i.branch-misses
-6.7% 19679133 perf-stat.i.cache-misses
-6.7% 53909365 perf-stat.i.cache-references
-7.6% 397018 perf-stat.i.context-switches
-7.3% 111272 perf-stat.i.cpu-migrations
9727 perf-stat.i.cycles-between-cache-misses
7.97 perf-stat.i.metric.K/sec
1.27 perf-stat.overall.MPKI
0.30 perf-stat.overall.branch-miss-rate%
9514 perf-stat.overall.cycles-between-cache-misses
-4.4% 10751121 perf-stat.ps.branch-misses
-6.7% 19302166 perf-stat.ps.cache-misses
-6.7% 52937829 perf-stat.ps.cache-references
-7.6% 390583 perf-stat.ps.context-switches
-7.3% 109477 perf-stat.ps.cpu-migrations
9.54 perf-profile.calltrace.cycles-pp.find_lock_lowest_rq.push_rt_task.push_rt_tasks.finish_task_switch.__schedule
20.04 perf-profile.calltrace.cycles-pp.push_rt_task.push_rt_tasks.finish_task_switch.__schedule.schedule
20.84 perf-profile.calltrace.cycles-pp.__x64_sys_futex.do_syscall_64.entry_SYSCALL_64_after_hwframe
20.83 perf-profile.calltrace.cycles-pp.do_futex.__x64_sys_futex.do_syscall_64.entry_SYSCALL_64_after_hwframe
21.60 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_after_hwframe
21.60 perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe
17.07 perf-profile.calltrace.cycles-pp.__futex_wait.futex_wait.do_futex.__x64_sys_futex.do_syscall_64
17.09 perf-profile.calltrace.cycles-pp.futex_wait.do_futex.__x64_sys_futex.do_syscall_64.entry_SYSCALL_64_after_hwframe
16.93 perf-profile.calltrace.cycles-pp.futex_wait_queue.__futex_wait.futex_wait.do_futex.__x64_sys_futex
16.88 perf-profile.calltrace.cycles-pp.schedule.futex_wait_queue.__futex_wait.futex_wait.do_futex
16.88 perf-profile.calltrace.cycles-pp.__schedule.schedule.futex_wait_queue.__futex_wait.futex_wait
3.98 perf-profile.calltrace.cycles-pp.__sched_yield
3.54 perf-profile.calltrace.cycles-pp.__schedule.schedule.__x64_sys_sched_yield.do_syscall_64.entry_SYSCALL_64_after_hwframe
3.55 perf-profile.calltrace.cycles-pp.schedule.__x64_sys_sched_yield.do_syscall_64.entry_SYSCALL_64_after_hwframe.__sched_yield
3.92 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_after_hwframe.__sched_yield
3.91 perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.__sched_yield
3.81 perf-profile.calltrace.cycles-pp.__x64_sys_sched_yield.do_syscall_64.entry_SYSCALL_64_after_hwframe.__sched_yield
15.46 perf-profile.calltrace.cycles-pp.finish_task_switch.__schedule.schedule.futex_wait_queue.__futex_wait
2.50 perf-profile.calltrace.cycles-pp.push_rt_tasks.finish_task_switch.__schedule.schedule.__x64_sys_sched_yield
14.72 perf-profile.calltrace.cycles-pp.push_rt_tasks.finish_task_switch.__schedule.schedule.futex_wait_queue
2.81 perf-profile.calltrace.cycles-pp.finish_task_switch.__schedule.schedule.__x64_sys_sched_yield.do_syscall_64
0.82 ? 2% perf-profile.calltrace.cycles-pp.cpupri_set.enqueue_task_rt.__sched_setscheduler._sched_setscheduler.do_sched_setscheduler
0.65 ? 2% perf-profile.calltrace.cycles-pp.cpupri_set.dequeue_rt_stack.dequeue_task_rt.__sched_setscheduler._sched_setscheduler
0.66 ? 2% perf-profile.calltrace.cycles-pp.dequeue_rt_stack.dequeue_task_rt.__sched_setscheduler._sched_setscheduler.do_sched_setscheduler
1.04 perf-profile.calltrace.cycles-pp.task_rq_lock.__sched_setscheduler._sched_setscheduler.do_sched_setscheduler.__x64_sys_sched_setscheduler
3.73 perf-profile.calltrace.cycles-pp.futex_wake.do_futex.__x64_sys_futex.do_syscall_64.entry_SYSCALL_64_after_hwframe
0.98 perf-profile.calltrace.cycles-pp._raw_spin_lock.task_rq_lock.__sched_setscheduler._sched_setscheduler.do_sched_setscheduler
0.62 perf-profile.calltrace.cycles-pp.rt_mutex_adjust_pi.__sched_setscheduler._sched_setscheduler.do_sched_setscheduler.__x64_sys_sched_setscheduler
0.71 perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock.__schedule.schedule_idle.do_idle
0.72 perf-profile.calltrace.cycles-pp._raw_spin_lock.__schedule.schedule_idle.do_idle.cpu_startup_entry
5.84 perf-profile.calltrace.cycles-pp.sysvec_call_function_single.asm_sysvec_call_function_single.__sched_setscheduler._sched_setscheduler.do_sched_setscheduler
5.92 perf-profile.calltrace.cycles-pp.asm_sysvec_call_function_single.__sched_setscheduler._sched_setscheduler.do_sched_setscheduler.__x64_sys_sched_setscheduler
5.84 perf-profile.calltrace.cycles-pp.__flush_smp_call_function_queue.__sysvec_call_function_single.sysvec_call_function_single.asm_sysvec_call_function_single.__sched_setscheduler
5.84 perf-profile.calltrace.cycles-pp.__sysvec_call_function_single.sysvec_call_function_single.asm_sysvec_call_function_single.__sched_setscheduler._sched_setscheduler
6.94 perf-profile.calltrace.cycles-pp.sched_ttwu_pending.__flush_smp_call_function_queue.__sysvec_call_function_single.sysvec_call_function_single.asm_sysvec_call_function_single
2.34 perf-profile.calltrace.cycles-pp._raw_spin_lock.__sched_setscheduler._sched_setscheduler.do_sched_setscheduler.__x64_sys_sched_setscheduler
2.33 perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock.__sched_setscheduler._sched_setscheduler.do_sched_setscheduler
8.61 perf-profile.calltrace.cycles-pp.secondary_startup_64_no_verify
5.52 perf-profile.calltrace.cycles-pp.activate_task.ttwu_do_activate.sched_ttwu_pending.__flush_smp_call_function_queue.__sysvec_call_function_single
8.58 perf-profile.calltrace.cycles-pp.enqueue_task_rt.activate_task.ttwu_do_activate.sched_ttwu_pending.__flush_smp_call_function_queue
8.39 perf-profile.calltrace.cycles-pp._raw_spin_lock.enqueue_task_rt.activate_task.ttwu_do_activate.sched_ttwu_pending
10.16 perf-profile.calltrace.cycles-pp.activate_task.push_rt_task.push_rt_tasks.finish_task_switch.__schedule
10.27 perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock.enqueue_task_rt.activate_task.ttwu_do_activate
9.86 perf-profile.calltrace.cycles-pp.enqueue_task_rt.activate_task.push_rt_task.push_rt_tasks.finish_task_switch
9.48 perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock.enqueue_task_rt.activate_task.push_rt_task
9.56 perf-profile.calltrace.cycles-pp._raw_spin_lock.enqueue_task_rt.activate_task.push_rt_task.push_rt_tasks
48.08 perf-profile.calltrace.cycles-pp.enqueue_task_rt.__sched_setscheduler._sched_setscheduler.do_sched_setscheduler.__x64_sys_sched_setscheduler
64.78 perf-profile.calltrace.cycles-pp.__sched_setscheduler
64.68 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_after_hwframe.__sched_setscheduler
64.68 perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.__sched_setscheduler
60.35 perf-profile.calltrace.cycles-pp.__x64_sys_sched_setscheduler.do_syscall_64.entry_SYSCALL_64_after_hwframe.__sched_setscheduler
60.35 perf-profile.calltrace.cycles-pp.do_sched_setscheduler.__x64_sys_sched_setscheduler.do_syscall_64.entry_SYSCALL_64_after_hwframe.__sched_setscheduler
60.05 perf-profile.calltrace.cycles-pp.__sched_setscheduler._sched_setscheduler.do_sched_setscheduler.__x64_sys_sched_setscheduler.do_syscall_64
60.24 perf-profile.calltrace.cycles-pp._sched_setscheduler.do_sched_setscheduler.__x64_sys_sched_setscheduler.do_syscall_64.entry_SYSCALL_64_after_hwframe
46.45 perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock.enqueue_task_rt.__sched_setscheduler._sched_setscheduler
46.82 perf-profile.calltrace.cycles-pp._raw_spin_lock.enqueue_task_rt.__sched_setscheduler._sched_setscheduler.do_sched_setscheduler
10.60 perf-profile.children.cycles-pp.find_lock_lowest_rq
25.37 perf-profile.children.cycles-pp.schedule
26.60 perf-profile.children.cycles-pp.__schedule
22.28 perf-profile.children.cycles-pp.push_rt_task
22.48 perf-profile.children.cycles-pp.finish_task_switch
21.13 perf-profile.children.cycles-pp.push_rt_tasks
20.84 perf-profile.children.cycles-pp.__x64_sys_futex
20.83 perf-profile.children.cycles-pp.do_futex
17.07 perf-profile.children.cycles-pp.__futex_wait
1.97 ? 2% perf-profile.children.cycles-pp.cpupri_set
17.09 perf-profile.children.cycles-pp.futex_wait
16.93 perf-profile.children.cycles-pp.futex_wait_queue
3.99 perf-profile.children.cycles-pp.__sched_yield
3.81 perf-profile.children.cycles-pp.__x64_sys_sched_yield
0.85 ? 2% perf-profile.children.cycles-pp.dequeue_rt_stack
0.43 ? 3% perf-profile.children.cycles-pp.find_lowest_rq
0.40 ? 4% perf-profile.children.cycles-pp.cpupri_find_fitness
0.78 perf-profile.children.cycles-pp.task_woken_rt
0.28 ? 3% perf-profile.children.cycles-pp.pull_rt_task
0.28 ? 2% perf-profile.children.cycles-pp.pick_next_task_rt
0.55 perf-profile.children.cycles-pp.enqueue_pushable_task
3.73 perf-profile.children.cycles-pp.futex_wake
0.10 ? 3% perf-profile.children.cycles-pp.balance_rt
0.41 perf-profile.children.cycles-pp.rto_push_irq_work_func
0.13 perf-profile.children.cycles-pp.select_task_rq
0.12 perf-profile.children.cycles-pp.select_task_rq_rt
0.06 perf-profile.children.cycles-pp.update_rt_rq_load_avg
0.27 perf-profile.children.cycles-pp.irq_exit_rcu
0.62 perf-profile.children.cycles-pp.rt_mutex_adjust_pi
0.53 perf-profile.children.cycles-pp.scheduler_tick
1.18 perf-profile.children.cycles-pp.update_curr_rt
0.63 perf-profile.children.cycles-pp.tick_nohz_highres_handler
0.62 perf-profile.children.cycles-pp.update_process_times
0.68 perf-profile.children.cycles-pp.hrtimer_interrupt
0.65 perf-profile.children.cycles-pp.__hrtimer_run_queues
0.63 perf-profile.children.cycles-pp.tick_sched_handle
0.68 perf-profile.children.cycles-pp.__sysvec_apic_timer_interrupt
0.93 perf-profile.children.cycles-pp.sysvec_apic_timer_interrupt
1.00 perf-profile.children.cycles-pp.asm_sysvec_apic_timer_interrupt
8.61 perf-profile.children.cycles-pp.cpu_startup_entry
8.61 perf-profile.children.cycles-pp.secondary_startup_64_no_verify
8.61 perf-profile.children.cycles-pp.do_idle
14.44 perf-profile.children.cycles-pp.sched_ttwu_pending
13.76 perf-profile.children.cycles-pp.ttwu_do_activate
24.19 perf-profile.children.cycles-pp.activate_task
60.36 perf-profile.children.cycles-pp.__x64_sys_sched_setscheduler
60.35 perf-profile.children.cycles-pp.do_sched_setscheduler
60.24 perf-profile.children.cycles-pp._sched_setscheduler
89.18 perf-profile.children.cycles-pp._raw_spin_lock
88.41 perf-profile.children.cycles-pp.native_queued_spin_lock_slowpath
71.77 perf-profile.children.cycles-pp.enqueue_task_rt
+0.8 125.02 perf-profile.children.cycles-pp.__sched_setscheduler
1.96 ? 2% perf-profile.self.cycles-pp.cpupri_set
0.27 ? 3% perf-profile.self.cycles-pp.cpupri_find_fitness
0.27 ? 4% perf-profile.self.cycles-pp.pull_rt_task
0.23 ? 3% perf-profile.self.cycles-pp.pick_next_task_rt
0.52 perf-profile.self.cycles-pp.enqueue_pushable_task
0.14 perf-profile.self.cycles-pp.switch_mm_irqs_off
0.67 perf-profile.self.cycles-pp._raw_spin_lock_irqsave
1.04 perf-profile.self.cycles-pp.update_curr_rt
88.40 perf-profile.self.cycles-pp.native_queued_spin_lock_slowpath
*************************************************************************
threads 2 sockets Intel(R) Xeon(R) Gold 6346 CPU @ 3.10GHz (Ice Lake) with 256G memory
===============================================================
kconfig/nr_threads/rootfs/tbox_group/test/testcase/testtime:
4-rhel-8.3/100%/debian-12-x86_64-20240206.cgz/lkp-icl-2sp7/ptrace/stress-ng/60s
Add VERW just before userspace transition")
Use ALTERNATIVE() instead of mds_user_clear static key")
6613d82e617dd7eb8b0c40b2fe3
---------------------------
%stddev
\
-3.9% 1418563 vmstat.system.cs
-3.9% 44765510 stress-ng.ptrace.ops
-3.9% 746080 stress-ng.ptrace.ops_per_sec
-3.9% 89501356 stress-ng.time.voluntary_context_switches
12835 ? 93% proc-vmstat.numa_pages_migrated
-5.7% 343290 ? 3% proc-vmstat.pgfree
12835 ? 93% proc-vmstat.pgmigrate_success
21353 ? 30% proc-vmstat.pgreuse
0.75 ? 2% perf-stat.i.branch-miss-rate%
+7.7% 47672895 ? 3% perf-stat.i.branch-misses
-4.0% 1472825 perf-stat.i.context-switches
-4.0% 23.08 perf-stat.i.metric.K/sec
0.73 ? 2% perf-stat.overall.branch-miss-rate%
+7.7% 46789221 ? 2% perf-stat.ps.branch-misses
-3.9% 1447769 perf-stat.ps.context-switches
45.22 perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irq.cgroup_enter_frozen.ptrace_stop.ptrace_do_notify
45.46 perf-profile.calltrace.cycles-pp._raw_spin_lock_irq.cgroup_enter_frozen.ptrace_stop.ptrace_do_notify.ptrace_notify
22.74 perf-profile.calltrace.cycles-pp.cgroup_enter_frozen.ptrace_stop.ptrace_do_notify.ptrace_notify.syscall_trace_enter
22.78 perf-profile.calltrace.cycles-pp.cgroup_enter_frozen.ptrace_stop.ptrace_do_notify.ptrace_notify.syscall_exit_to_user_mode_prepare
7.88 perf-profile.calltrace.cycles-pp.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe.getgid
0.64 perf-profile.calltrace.cycles-pp.__schedule.schedule.do_wait.kernel_wait4.__do_sys_wait4
0.66 perf-profile.calltrace.cycles-pp.schedule.do_wait.kernel_wait4.__do_sys_wait4.do_syscall_64
1.85 perf-profile.calltrace.cycles-pp.do_wait.kernel_wait4.__do_sys_wait4.do_syscall_64.entry_SYSCALL_64_after_hwframe
1.94 perf-profile.calltrace.cycles-pp.kernel_wait4.__do_sys_wait4.do_syscall_64.entry_SYSCALL_64_after_hwframe.wait4
1.98 perf-profile.calltrace.cycles-pp.__do_sys_wait4.do_syscall_64.entry_SYSCALL_64_after_hwframe.wait4
2.19 perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.wait4
2.20 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_after_hwframe.wait4
2.42 perf-profile.calltrace.cycles-pp.wait4
1.34 ? 3% perf-profile.calltrace.cycles-pp.__x64_sys_ptrace.do_syscall_64.entry_SYSCALL_64_after_hwframe.ptrace
1.36 ? 3% perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_after_hwframe.ptrace
1.36 ? 3% perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.ptrace
1.44 ? 2% perf-profile.calltrace.cycles-pp.ptrace
22.84 perf-profile.calltrace.cycles-pp.cgroup_leave_frozen.ptrace_stop.ptrace_do_notify.ptrace_notify.syscall_trace_enter
45.33 perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_irq.cgroup_leave_frozen.ptrace_stop.ptrace_do_notify
45.72 perf-profile.calltrace.cycles-pp._raw_spin_lock_irq.cgroup_leave_frozen.ptrace_stop.ptrace_do_notify.ptrace_notify
45.55 perf-profile.children.cycles-pp.cgroup_enter_frozen
90.57 perf-profile.children.cycles-pp.native_queued_spin_lock_slowpath
0.27 perf-profile.children.cycles-pp.switch_mm_irqs_off
0.17 ? 2% perf-profile.children.cycles-pp.asm_sysvec_apic_timer_interrupt
0.63 ? 2% perf-profile.children.cycles-pp.ttwu_do_activate
1.85 perf-profile.children.cycles-pp.do_wait
1.95 perf-profile.children.cycles-pp.kernel_wait4
1.98 perf-profile.children.cycles-pp.__do_sys_wait4
2.44 perf-profile.children.cycles-pp.wait4
1.56 perf-profile.children.cycles-pp.__schedule
1.58 perf-profile.children.cycles-pp.schedule
1.34 ? 3% perf-profile.children.cycles-pp.__x64_sys_ptrace
1.46 ? 2% perf-profile.children.cycles-pp.ptrace
45.82 perf-profile.children.cycles-pp.cgroup_leave_frozen
90.57 perf-profile.self.cycles-pp.native_queued_spin_lock_slowpath
0.37 perf-profile.self.cycles-pp.ptrace_stop
1.02 perf-profile.self.cycles-pp._raw_spin_lock_irq
*************************************************************************
threads 2 sockets Intel(R) Xeon(R) Gold 6346 CPU @ 3.10GHz (Ice Lake) with 256G memory
===============================================================
disk/fs/kconfig/nr_threads/rootfs/tbox_group/test/testcase/testtime:
btrfs/x86_64-rhel-8.3/100%/debian-12-x86_64-20240206.cgz/lkp-icl-2sp8/getdent/stress-ng/60s
Add VERW just before userspace transition")
Use ALTERNATIVE() instead of mds_user_clear static key")
6613d82e617dd7eb8b0c40b2fe3
---------------------------
%stddev
\
+51.9% 482822 ? 4% cpuidle..usage
3.15 iostat.cpu.user
2.38 ? 4% mpstat.cpu.all.idle%
23954 ? 3% vmstat.system.cs
+2.4% 182761 vmstat.system.in
13598 ? 3% sched_debug.cpu.nr_switches.avg
38014 ? 6% sched_debug.cpu.nr_switches.max
9736 ? 4% sched_debug.cpu.nr_switches.stddev
-40.0% 35337032 numa-numastat.node0.local_node
-40.0% 35382357 numa-numastat.node0.numa_hit
+76.6% 65448120 numa-numastat.node1.local_node
+76.5% 65478720 numa-numastat.node1.numa_hit
77312 ? 28% numa-meminfo.node0.KReclaimable
77312 ? 28% numa-meminfo.node0.SReclaimable
-51.5% 187589 ? 12% numa-meminfo.node0.Slab
274479 ? 8% numa-meminfo.node1.KReclaimable
274479 ? 8% numa-meminfo.node1.SReclaimable
343568 ? 7% numa-meminfo.node1.Slab
19547 ? 28% numa-vmstat.node0.nr_slab_reclaimable
-40.0% 35463515 numa-vmstat.node0.numa_hit
-40.0% 35418189 numa-vmstat.node0.numa_local
69229 ? 9% numa-vmstat.node1.nr_slab_reclaimable
+76.6% 65720661 numa-vmstat.node1.numa_hit
+76.7% 65690060 numa-vmstat.node1.numa_local
8816 stress-ng.getdent.nanosecs_per_getdents_call
+5.8% 1.238e+08 stress-ng.getdent.ops
+5.8% 2063349 stress-ng.getdent.ops_per_sec
+12.9% 109764 stress-ng.time.involuntary_context_switches
+5.8% 90920658 stress-ng.time.minor_page_faults
+6.7% 88.32 stress-ng.time.user_time
+74.4% 649143 ? 3% stress-ng.time.voluntary_context_switches
-1.7% 88797 proc-vmstat.nr_slab_reclaimable
14551 ? 2% proc-vmstat.numa_hint_faults
7560 ? 7% proc-vmstat.numa_hint_faults_local
+5.2% 1.011e+08 proc-vmstat.numa_hit
+5.2% 1.01e+08 proc-vmstat.numa_local
+4.7% 1.059e+08 proc-vmstat.pgalloc_normal
+5.9% 91111926 proc-vmstat.pgfault
+4.7% 1.057e+08 proc-vmstat.pgfree
-8.3% 13744 proc-vmstat.pgreuse
3.15 perf-stat.i.MPKI
+5.0% 1.082e+10 perf-stat.i.branch-instructions
+5.3% 82008784 perf-stat.i.branch-misses
42.98 perf-stat.i.cache-miss-rate%
+6.5% 3.83e+08 perf-stat.i.cache-references
24926 ? 3% perf-stat.i.context-switches
4.27 perf-stat.i.cpi
+30.7% 443.20 ? 4% perf-stat.i.cpu-migrations
+5.0% 5.243e+10 perf-stat.i.instructions
0.25 perf-stat.i.ipc
+5.9% 46.82 perf-stat.i.metric.K/sec
+5.9% 1494386 perf-stat.i.minor-faults
+5.9% 1494386 perf-stat.i.page-faults
3.17 perf-stat.overall.MPKI
43.40 perf-stat.overall.cache-miss-rate%
4.28 perf-stat.overall.cpi
0.23 perf-stat.overall.ipc
+4.9% 1.063e+10 perf-stat.ps.branch-instructions
+5.3% 80174083 perf-stat.ps.branch-misses
+6.3% 3.765e+08 perf-stat.ps.cache-references
24284 ? 3% perf-stat.ps.context-switches
+30.6% 433.03 ? 4% perf-stat.ps.cpu-migrations
+4.9% 5.149e+10 perf-stat.ps.instructions
+5.8% 1468698 perf-stat.ps.minor-faults
+5.8% 1468698 perf-stat.ps.page-faults
+4.2% 3.133e+12 perf-stat.total.instructions
56.25 perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.syscall
56.31 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_after_hwframe.syscall
56.79 perf-profile.calltrace.cycles-pp.syscall
27.70 perf-profile.calltrace.cycles-pp.__x64_sys_getdents64.do_syscall_64.entry_SYSCALL_64_after_hwframe.syscall
27.16 perf-profile.calltrace.cycles-pp.iterate_dir.__x64_sys_getdents64.do_syscall_64.entry_SYSCALL_64_after_hwframe.syscall
28.37 perf-profile.calltrace.cycles-pp.__x64_sys_getdents.do_syscall_64.entry_SYSCALL_64_after_hwframe.syscall
27.58 perf-profile.calltrace.cycles-pp.iterate_dir.__x64_sys_getdents.do_syscall_64.entry_SYSCALL_64_after_hwframe.syscall
8.13 ? 2% perf-profile.calltrace.cycles-pp.proc_readdir_de.proc_tgid_net_readdir.iterate_dir.__x64_sys_getdents64.do_syscall_64
8.19 ? 2% perf-profile.calltrace.cycles-pp.proc_tgid_net_readdir.iterate_dir.__x64_sys_getdents64.do_syscall_64.entry_SYSCALL_64_after_hwframe
8.12 ? 2% perf-profile.calltrace.cycles-pp.proc_readdir_de.proc_tgid_net_readdir.iterate_dir.__x64_sys_getdents.do_syscall_64
8.21 ? 2% perf-profile.calltrace.cycles-pp.proc_tgid_net_readdir.iterate_dir.__x64_sys_getdents.do_syscall_64.entry_SYSCALL_64_after_hwframe
4.89 ? 2% perf-profile.calltrace.cycles-pp.proc_readdir_de.iterate_dir.__x64_sys_getdents64.do_syscall_64.entry_SYSCALL_64_after_hwframe
4.88 ? 2% perf-profile.calltrace.cycles-pp.proc_readdir_de.iterate_dir.__x64_sys_getdents.do_syscall_64.entry_SYSCALL_64_after_hwframe
3.72 ? 2% perf-profile.calltrace.cycles-pp._raw_read_lock.proc_readdir_de.proc_tgid_net_readdir.iterate_dir.__x64_sys_getdents64
3.71 ? 2% perf-profile.calltrace.cycles-pp._raw_read_lock.proc_readdir_de.proc_tgid_net_readdir.iterate_dir.__x64_sys_getdents
2.47 ? 2% perf-profile.calltrace.cycles-pp._raw_read_lock.proc_readdir_de.iterate_dir.__x64_sys_getdents64.do_syscall_64
2.47 ? 2% perf-profile.calltrace.cycles-pp._raw_read_lock.proc_readdir_de.iterate_dir.__x64_sys_getdents.do_syscall_64
0.80 ? 2% perf-profile.calltrace.cycles-pp._raw_read_lock.proc_lookup_de.proc_tgid_net_lookup.lookup_open.open_last_lookups
0.58 perf-profile.calltrace.cycles-pp.alloc_pages_mpol.vma_alloc_folio.do_anonymous_page.__handle_mm_fault.handle_mm_fault
0.64 perf-profile.calltrace.cycles-pp.vma_alloc_folio.do_anonymous_page.__handle_mm_fault.handle_mm_fault.do_user_addr_fault
0.59 ? 2% perf-profile.calltrace.cycles-pp.d_alloc_parallel.lookup_open.open_last_lookups.path_openat.do_filp_open
0.81 perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock.new_inode.proc_get_inode.proc_lookup_de
0.71 ? 5% perf-profile.calltrace.cycles-pp.apparmor_file_free_security.security_file_free.__fput.__x64_sys_close.do_syscall_64
0.72 ? 5% perf-profile.calltrace.cycles-pp.security_file_free.__fput.__x64_sys_close.do_syscall_64.entry_SYSCALL_64_after_hwframe
1.08 perf-profile.calltrace.cycles-pp._raw_spin_lock.new_inode.proc_get_inode.proc_lookup_de.proc_tgid_net_lookup
1.43 perf-profile.calltrace.cycles-pp.new_inode.proc_get_inode.proc_lookup_de.proc_tgid_net_lookup.lookup_open
1.49 perf-profile.calltrace.cycles-pp.proc_get_inode.proc_lookup_de.proc_tgid_net_lookup.lookup_open.open_last_lookups
0.82 perf-profile.calltrace.cycles-pp.may_open.do_open.path_openat.do_filp_open.do_sys_openat2
0.75 perf-profile.calltrace.cycles-pp.inode_permission.may_open.do_open.path_openat.do_filp_open
2.03 ? 2% perf-profile.calltrace.cycles-pp.evict.__dentry_kill.dput.__fput.__x64_sys_close
0.94 ? 4% perf-profile.calltrace.cycles-pp.apparmor_file_alloc_security.security_file_alloc.init_file.alloc_empty_file.path_openat
1.23 ? 2% perf-profile.calltrace.cycles-pp._raw_spin_lock.evict.__dentry_kill.dput.__fput
1.26 ? 3% perf-profile.calltrace.cycles-pp.init_file.alloc_empty_file.path_openat.do_filp_open.do_sys_openat2
0.98 ? 3% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock.evict.__dentry_kill.dput
1.04 ? 3% perf-profile.calltrace.cycles-pp.security_file_alloc.init_file.alloc_empty_file.path_openat.do_filp_open
1.61 ? 3% perf-profile.calltrace.cycles-pp.alloc_empty_file.path_openat.do_filp_open.do_sys_openat2.__x64_sys_openat
0.86 ? 22% perf-profile.calltrace.cycles-pp.__legitimize_path.try_to_unlazy.link_path_walk.path_openat.do_filp_open
0.87 ? 22% perf-profile.calltrace.cycles-pp.try_to_unlazy.link_path_walk.path_openat.do_filp_open.do_sys_openat2
1.10 ? 5% perf-profile.calltrace.cycles-pp.up_read.kernfs_dop_revalidate.lookup_fast.walk_component.link_path_walk
1.73 ? 7% perf-profile.calltrace.cycles-pp.lookup_fast.open_last_lookups.path_openat.do_filp_open.do_sys_openat2
1.56 ? 5% perf-profile.calltrace.cycles-pp.apparmor_file_open.security_file_open.do_dentry_open.do_open.path_openat
1.57 ? 5% perf-profile.calltrace.cycles-pp.security_file_open.do_dentry_open.do_open.path_openat.do_filp_open
2.65 ? 4% perf-profile.calltrace.cycles-pp.do_dentry_open.do_open.path_openat.do_filp_open.do_sys_openat2
1.76 ? 5% perf-profile.calltrace.cycles-pp.up_read.kernfs_iop_permission.inode_permission.link_path_walk.path_openat
1.81 ? 2% perf-profile.calltrace.cycles-pp.down_read.kernfs_iop_permission.inode_permission.link_path_walk.path_openat
1.40 ? 2% perf-profile.calltrace.cycles-pp.up_read.kernfs_fop_readdir.iterate_dir.__x64_sys_getdents.do_syscall_64
1.40 ? 2% perf-profile.calltrace.cycles-pp.up_read.kernfs_fop_readdir.iterate_dir.__x64_sys_getdents64.do_syscall_64
1.56 ? 4% perf-profile.calltrace.cycles-pp.down_read.kernfs_dop_revalidate.lookup_fast.walk_component.link_path_walk
1.44 ? 24% perf-profile.calltrace.cycles-pp.dput.terminate_walk.path_openat.do_filp_open.do_sys_openat2
1.48 ? 23% perf-profile.calltrace.cycles-pp.terminate_walk.path_openat.do_filp_open.do_sys_openat2.__x64_sys_openat
1.81 perf-profile.calltrace.cycles-pp.down_read.kernfs_fop_readdir.iterate_dir.__x64_sys_getdents.do_syscall_64
1.81 ? 2% perf-profile.calltrace.cycles-pp.down_read.kernfs_fop_readdir.iterate_dir.__x64_sys_getdents64.do_syscall_64
0.58 ? 2% perf-profile.calltrace.cycles-pp.kernfs_dop_revalidate.lookup_fast.open_last_lookups.path_openat.do_filp_open
3.93 ? 3% perf-profile.calltrace.cycles-pp.do_open.path_openat.do_filp_open.do_sys_openat2.__x64_sys_openat
0.70 ? 28% perf-profile.calltrace.cycles-pp.lockref_get_not_dead.__legitimize_path.try_to_unlazy.link_path_walk.path_openat
0.56 perf-profile.calltrace.cycles-pp.kernfs_iop_permission.inode_permission.may_open.do_open.path_openat
2.71 ? 4% perf-profile.calltrace.cycles-pp.kernfs_dop_revalidate.lookup_fast.walk_component.link_path_walk.path_openat
3.71 ? 4% perf-profile.calltrace.cycles-pp.kernfs_iop_permission.inode_permission.link_path_walk.path_openat.do_filp_open
4.77 ? 3% perf-profile.calltrace.cycles-pp.inode_permission.link_path_walk.path_openat.do_filp_open.do_sys_openat2
5.50 ? 2% perf-profile.calltrace.cycles-pp.walk_component.link_path_walk.path_openat.do_filp_open.do_sys_openat2
3.93 ? 3% perf-profile.calltrace.cycles-pp.lookup_fast.walk_component.link_path_walk.path_openat.do_filp_open
3.96 perf-profile.calltrace.cycles-pp.kernfs_fop_readdir.iterate_dir.__x64_sys_getdents.do_syscall_64.entry_SYSCALL_64_after_hwframe
3.96 ? 2% perf-profile.calltrace.cycles-pp.kernfs_fop_readdir.iterate_dir.__x64_sys_getdents64.do_syscall_64.entry_SYSCALL_64_after_hwframe
12.34 ? 2% perf-profile.calltrace.cycles-pp.link_path_walk.path_openat.do_filp_open.do_sys_openat2.__x64_sys_openat
27.32 perf-profile.calltrace.cycles-pp.path_openat.do_filp_open.do_sys_openat2.__x64_sys_openat.do_syscall_64
27.39 perf-profile.calltrace.cycles-pp.do_filp_open.do_sys_openat2.__x64_sys_openat.do_syscall_64.entry_SYSCALL_64_after_hwframe
28.52 perf-profile.calltrace.cycles-pp.do_sys_openat2.__x64_sys_openat.do_syscall_64.entry_SYSCALL_64_after_hwframe.open64
28.55 perf-profile.calltrace.cycles-pp.__x64_sys_openat.do_syscall_64.entry_SYSCALL_64_after_hwframe.open64
28.61 perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.open64
28.63 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_after_hwframe.open64
28.83 perf-profile.calltrace.cycles-pp.open64
26.07 ? 2% perf-profile.children.cycles-pp.proc_readdir_de
54.77 perf-profile.children.cycles-pp.iterate_dir
56.97 perf-profile.children.cycles-pp.syscall
16.39 ? 2% perf-profile.children.cycles-pp.proc_tgid_net_readdir
13.70 ? 2% perf-profile.children.cycles-pp._raw_read_lock
27.70 perf-profile.children.cycles-pp.__x64_sys_getdents64
28.38 perf-profile.children.cycles-pp.__x64_sys_getdents
93.85 perf-profile.children.cycles-pp.do_syscall_64
93.94 perf-profile.children.cycles-pp.entry_SYSCALL_64_after_hwframe
0.86 ? 3% perf-profile.children.cycles-pp.proc_readfd_common
0.05 perf-profile.children.cycles-pp.main
0.05 perf-profile.children.cycles-pp.run_builtin
0.05 perf-profile.children.cycles-pp.__cmd_record
0.05 perf-profile.children.cycles-pp.cmd_record
0.10 ? 4% perf-profile.children.cycles-pp._raw_spin_lock_irq
0.11 perf-profile.children.cycles-pp.atime_needs_update
0.06 perf-profile.children.cycles-pp.nd_jump_root
0.10 perf-profile.children.cycles-pp.__init_rwsem
0.18 perf-profile.children.cycles-pp.generic_permission
0.07 perf-profile.children.cycles-pp.proc_pid_readdir
0.07 perf-profile.children.cycles-pp.process_measurement
0.13 perf-profile.children.cycles-pp.uncharge_batch
0.19 perf-profile.children.cycles-pp.vsnprintf
0.24 perf-profile.children.cycles-pp.native_irq_return_iret
0.21 ? 2% perf-profile.children.cycles-pp.stress_getdents_dir
0.18 ? 2% perf-profile.children.cycles-pp.memchr
0.09 ? 5% perf-profile.children.cycles-pp.path_init
0.10 ? 4% perf-profile.children.cycles-pp.locks_remove_posix
0.25 perf-profile.children.cycles-pp.__memcg_slab_post_alloc_hook
0.40 perf-profile.children.cycles-pp.getname_flags
0.12 ? 4% perf-profile.children.cycles-pp.page_counter_uncharge
0.16 perf-profile.children.cycles-pp.__mem_cgroup_uncharge_list
0.12 perf-profile.children.cycles-pp.percpu_counter_add_batch
0.21 ? 2% perf-profile.children.cycles-pp.inode_init_always
0.22 ? 2% perf-profile.children.cycles-pp.mod_objcg_state
0.35 perf-profile.children.cycles-pp.__cond_resched
0.20 ? 2% perf-profile.children.cycles-pp.strlcat
0.58 perf-profile.children.cycles-pp.alloc_inode
0.55 ? 2% perf-profile.children.cycles-pp.d_alloc
0.69 perf-profile.children.cycles-pp.__slab_free
0.15 ? 18% perf-profile.children.cycles-pp.try_to_unlazy_next
0.34 perf-profile.children.cycles-pp.__memcg_slab_free_hook
0.74 perf-profile.children.cycles-pp.d_alloc_parallel
0.81 perf-profile.children.cycles-pp.filldir64
0.82 ? 2% perf-profile.children.cycles-pp.filldir
0.17 ? 7% perf-profile.children.cycles-pp.security_current_getsecid_subj
0.05 ? 8% perf-profile.children.cycles-pp.cpu_startup_entry
0.05 ? 8% perf-profile.children.cycles-pp.do_idle
0.05 ? 8% perf-profile.children.cycles-pp.secondary_startup_64_no_verify
0.05 ? 8% perf-profile.children.cycles-pp.start_secondary
0.16 ? 7% perf-profile.children.cycles-pp.apparmor_current_getsecid_subj
1.09 perf-profile.children.cycles-pp.rcu_do_batch
1.10 perf-profile.children.cycles-pp.rcu_core
1.11 perf-profile.children.cycles-pp.__do_softirq
0.71 ? 4% perf-profile.children.cycles-pp.apparmor_file_free_security
0.72 ? 4% perf-profile.children.cycles-pp.security_file_free
0.25 ? 6% perf-profile.children.cycles-pp.ima_file_check
1.27 perf-profile.children.cycles-pp.kmem_cache_free
0.51 ? 5% perf-profile.children.cycles-pp.smpboot_thread_fn
0.52 ? 5% perf-profile.children.cycles-pp.kthread
0.52 ? 5% perf-profile.children.cycles-pp.ret_from_fork
0.52 ? 5% perf-profile.children.cycles-pp.ret_from_fork_asm
0.51 ? 5% perf-profile.children.cycles-pp.run_ksoftirqd
0.83 perf-profile.children.cycles-pp.may_open
2.04 ? 2% perf-profile.children.cycles-pp.evict
1.27 ? 3% perf-profile.children.cycles-pp.init_file
0.95 ? 4% perf-profile.children.cycles-pp.apparmor_file_alloc_security
1.04 ? 3% perf-profile.children.cycles-pp.security_file_alloc
1.61 ? 3% perf-profile.children.cycles-pp.alloc_empty_file
2.34 perf-profile.children.cycles-pp.new_inode
2.42 perf-profile.children.cycles-pp.proc_get_inode
0.70 ? 7% perf-profile.children.cycles-pp.apparmor_file_permission
0.72 ? 6% perf-profile.children.cycles-pp.security_file_permission
1.56 ? 5% perf-profile.children.cycles-pp.apparmor_file_open
1.57 ? 5% perf-profile.children.cycles-pp.security_file_open
2.66 ? 4% perf-profile.children.cycles-pp.do_dentry_open
1.69 ? 19% perf-profile.children.cycles-pp.try_to_unlazy
1.50 ? 23% perf-profile.children.cycles-pp.terminate_walk
1.77 ? 20% perf-profile.children.cycles-pp.__legitimize_path
1.39 ? 26% perf-profile.children.cycles-pp.lockref_get_not_dead
7.42 ? 3% perf-profile.children.cycles-pp.dput
3.94 ? 3% perf-profile.children.cycles-pp.do_open
5.53 ? 2% perf-profile.children.cycles-pp.walk_component
4.29 ? 3% perf-profile.children.cycles-pp.kernfs_iop_permission
5.54 ? 2% perf-profile.children.cycles-pp.inode_permission
3.30 ? 4% perf-profile.children.cycles-pp.kernfs_dop_revalidate
5.69 ? 3% perf-profile.children.cycles-pp.lookup_fast
6.89 ? 3% perf-profile.children.cycles-pp.up_read
12.38 ? 2% perf-profile.children.cycles-pp.link_path_walk
7.95 ? 2% perf-profile.children.cycles-pp.kernfs_fop_readdir
7.75 ? 2% perf-profile.children.cycles-pp.down_read
27.36 perf-profile.children.cycles-pp.path_openat
27.42 perf-profile.children.cycles-pp.do_filp_open
28.56 perf-profile.children.cycles-pp.do_sys_openat2
28.58 perf-profile.children.cycles-pp.__x64_sys_openat
28.88 perf-profile.children.cycles-pp.open64
13.59 ? 2% perf-profile.self.cycles-pp._raw_read_lock
11.95 ? 2% perf-profile.self.cycles-pp.proc_readdir_de
1.46 ? 2% perf-profile.self.cycles-pp.proc_lookup_de
0.10 ? 4% perf-profile.self.cycles-pp._raw_spin_lock_irq
0.11 perf-profile.self.cycles-pp.page_counter_uncharge
0.11 perf-profile.self.cycles-pp.percpu_counter_add_batch
0.06 perf-profile.self.cycles-pp.refill_obj_stock
0.09 perf-profile.self.cycles-pp.number
0.10 perf-profile.self.cycles-pp.pid_revalidate
0.21 ? 2% perf-profile.self.cycles-pp.__cond_resched
0.28 perf-profile.self.cycles-pp.entry_SYSRETQ_unsafe_stack
0.24 perf-profile.self.cycles-pp.native_irq_return_iret
0.13 ? 3% perf-profile.self.cycles-pp.__memcg_slab_free_hook
0.14 ? 3% perf-profile.self.cycles-pp.generic_permission
0.09 ? 5% perf-profile.self.cycles-pp.locks_remove_posix
0.10 perf-profile.self.cycles-pp.__call_rcu_common
0.19 perf-profile.self.cycles-pp.proc_tgid_net_lookup
0.18 ? 2% perf-profile.self.cycles-pp.mod_objcg_state
0.15 ? 3% perf-profile.self.cycles-pp.inode_init_always
0.18 ? 2% perf-profile.self.cycles-pp.do_syscall_64
0.19 ? 2% perf-profile.self.cycles-pp.get_proc_task_net
0.25 perf-profile.self.cycles-pp.syscall
0.68 perf-profile.self.cycles-pp.__slab_free
0.41 ? 6% perf-profile.self.cycles-pp.inode_permission
0.60 perf-profile.self.cycles-pp.filldir
0.89 ? 2% perf-profile.self.cycles-pp.lockref_get_not_dead
0.60 ? 2% perf-profile.self.cycles-pp.filldir64
0.05 perf-profile.self.cycles-pp.proc_tgid_net_readdir
0.16 ? 7% perf-profile.self.cycles-pp.apparmor_current_getsecid_subj
0.71 ? 5% perf-profile.self.cycles-pp.apparmor_file_free_security
0.93 ? 4% perf-profile.self.cycles-pp.apparmor_file_alloc_security
0.66 ? 6% perf-profile.self.cycles-pp.apparmor_file_permission
1.54 ? 5% perf-profile.self.cycles-pp.apparmor_file_open
6.84 ? 3% perf-profile.self.cycles-pp.up_read
7.66 ? 2% perf-profile.self.cycles-pp.down_read
*************************************************************************
threads 2 sockets (Skylake) with 192G memory
===============================================================
kconfig/mode/nr_task/rootfs/tbox_group/test/testcase:
4-rhel-8.3/thread/100%/debian-12-x86_64-20240206.cgz/lkp-skl-fpga01/futex4/will-it-scale
Add VERW just before userspace transition")
Use ALTERNATIVE() instead of mds_user_clear static key")
6613d82e617dd7eb8b0c40b2fe3
---------------------------
%stddev
\
+1.2% 34.01 boot-time.boot
3171 boot-time.idle
+4.0% 73672365 will-it-scale.104.threads
+4.0% 708387 will-it-scale.per_thread_ops
+4.0% 73672365 will-it-scale.workload
-1.7% 88005 proc-vmstat.nr_active_anon
-1.7% 91127 proc-vmstat.nr_shmem
-1.7% 88005 proc-vmstat.nr_zone_active_anon
-1.7% 75654 proc-vmstat.pgactivate
-1.8% 1066713 proc-vmstat.pgalloc_normal
44714 ? 4% proc-vmstat.pgreuse
16392 ? 5% sched_debug.cfs_rq:/.load_avg.max
0.06 ? 3% sched_debug.cfs_rq:/.nr_running.stddev
4.45 ? 30% sched_debug.cfs_rq:/.removed.runnable_avg.avg
18.00 ? 17% sched_debug.cfs_rq:/.removed.runnable_avg.stddev
4.45 ? 30% sched_debug.cfs_rq:/.removed.util_avg.avg
17.99 ? 17% sched_debug.cfs_rq:/.removed.util_avg.stddev
1039 ? 16% sched_debug.cfs_rq:/.util_est.max
4902 ? 12% sched_debug.cpu.max_idle_balance_cost.stddev
0.00 ? 2% sched_debug.cpu.next_balance.stddev
+10.9% 972.39 ? 7% sched_debug.cpu.nr_switches.min
5.60 ? 4% sched_debug.cpu.nr_uninterruptible.stddev
+4.3% 6.376e+09 perf-stat.i.branch-instructions
1.53 perf-stat.i.branch-miss-rate%
+19.2% 97330429 perf-stat.i.branch-misses
5.85 perf-stat.i.cpi
+4.0% 4.944e+10 perf-stat.i.instructions
0.17 perf-stat.i.ipc
1.53 perf-stat.overall.branch-miss-rate%
5.86 perf-stat.overall.cpi
0.17 perf-stat.overall.ipc
+4.3% 6.354e+09 perf-stat.ps.branch-instructions
+19.2% 96977141 perf-stat.ps.branch-misses
+4.0% 4.928e+10 perf-stat.ps.instructions
0.05 ? 9% perf-stat.ps.major-faults
+3.6% 1.491e+13 perf-stat.total.instructions
22.93 perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.syscall
36.49 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_after_hwframe.syscall
1.17 ? 2% perf-profile.calltrace.cycles-pp.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe.syscall
14.24 perf-profile.calltrace.cycles-pp.futex_wait.do_futex.__x64_sys_futex.do_syscall_64.entry_SYSCALL_64_after_hwframe
16.15 perf-profile.calltrace.cycles-pp.__x64_sys_futex.do_syscall_64.entry_SYSCALL_64_after_hwframe.syscall
14.83 perf-profile.calltrace.cycles-pp.do_futex.__x64_sys_futex.do_syscall_64.entry_SYSCALL_64_after_hwframe.syscall
2.31 ? 2% perf-profile.calltrace.cycles-pp.__get_user_nocheck_4.futex_get_value_locked.futex_wait_setup.__futex_wait.futex_wait
3.10 perf-profile.calltrace.cycles-pp.futex_get_value_locked.futex_wait_setup.__futex_wait.futex_wait.do_futex
26.91 perf-profile.calltrace.cycles-pp.syscall_return_via_sysret.syscall
10.22 perf-profile.calltrace.cycles-pp.futex_wait_setup.__futex_wait.futex_wait.do_futex.__x64_sys_futex
11.54 perf-profile.calltrace.cycles-pp.__futex_wait.futex_wait.do_futex.__x64_sys_futex.do_syscall_64
0.69 perf-profile.calltrace.cycles-pp.get_futex_key.futex_wait_setup.__futex_wait.futex_wait.do_futex
0.82 perf-profile.calltrace.cycles-pp.futex_q_unlock.futex_wait_setup.__futex_wait.futex_wait.do_futex
99.65 perf-profile.calltrace.cycles-pp.syscall
1.19 perf-profile.calltrace.cycles-pp.futex_hash.futex_q_lock.futex_wait_setup.__futex_wait.futex_wait
9.13 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64.syscall
2.00 perf-profile.calltrace.cycles-pp._raw_spin_lock.futex_q_lock.futex_wait_setup.__futex_wait.futex_wait
4.58 ? 2% perf-profile.calltrace.cycles-pp.futex_q_lock.futex_wait_setup.__futex_wait.futex_wait.do_futex
10.62 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_safe_stack.syscall
17.48 perf-profile.calltrace.cycles-pp.entry_SYSRETQ_unsafe_stack.syscall
22.46 perf-profile.children.cycles-pp.do_syscall_64
37.23 perf-profile.children.cycles-pp.entry_SYSCALL_64_after_hwframe
1.26 ? 2% perf-profile.children.cycles-pp.syscall_exit_to_user_mode
14.36 perf-profile.children.cycles-pp.futex_wait
16.21 perf-profile.children.cycles-pp.__x64_sys_futex
14.90 perf-profile.children.cycles-pp.do_futex
2.68 ? 2% perf-profile.children.cycles-pp.__get_user_nocheck_4
3.29 ? 2% perf-profile.children.cycles-pp.futex_get_value_locked
27.09 perf-profile.children.cycles-pp.syscall_return_via_sysret
10.47 perf-profile.children.cycles-pp.futex_wait_setup
11.58 perf-profile.children.cycles-pp.__futex_wait
0.13 ? 6% perf-profile.children.cycles-pp.amd_clear_divider
0.13 ? 3% perf-profile.children.cycles-pp.futex_setup_timer
0.41 ? 3% perf-profile.children.cycles-pp.syscall_exit_to_user_mode_prepare
0.06 perf-profile.children.cycles-pp.syscall@plt
0.16 ? 3% perf-profile.children.cycles-pp.testcase
0.86 perf-profile.children.cycles-pp.futex_q_unlock
0.72 perf-profile.children.cycles-pp.get_futex_key
1.21 perf-profile.children.cycles-pp.futex_hash
2.13 perf-profile.children.cycles-pp._raw_spin_lock
4.75 ? 2% perf-profile.children.cycles-pp.futex_q_lock
5.62 perf-profile.children.cycles-pp.entry_SYSCALL_64_safe_stack
12.94 perf-profile.children.cycles-pp.entry_SYSCALL_64
18.64 perf-profile.children.cycles-pp.entry_SYSRETQ_unsafe_stack
0.77 perf-profile.self.cycles-pp.syscall_exit_to_user_mode
2.51 perf-profile.self.cycles-pp.__get_user_nocheck_4
2.68 perf-profile.self.cycles-pp.futex_wait
27.05 perf-profile.self.cycles-pp.syscall_return_via_sysret
0.10 perf-profile.self.cycles-pp.futex_setup_timer
0.36 ? 4% perf-profile.self.cycles-pp.syscall_exit_to_user_mode_prepare
0.09 ? 5% perf-profile.self.cycles-pp.amd_clear_divider
0.61 perf-profile.self.cycles-pp.entry_SYSCALL_64_safe_stack
0.16 ? 3% perf-profile.self.cycles-pp.testcase
0.69 perf-profile.self.cycles-pp.get_futex_key
0.83 perf-profile.self.cycles-pp.futex_q_unlock
1.07 perf-profile.self.cycles-pp.__futex_wait
0.58 perf-profile.self.cycles-pp.do_futex
1.06 perf-profile.self.cycles-pp.futex_wait_setup
1.17 perf-profile.self.cycles-pp.futex_hash
2.09 perf-profile.self.cycles-pp._raw_spin_lock
11.73 perf-profile.self.cycles-pp.entry_SYSCALL_64
4.37 perf-profile.self.cycles-pp.syscall
4.19 perf-profile.self.cycles-pp.do_syscall_64
15.38 perf-profile.self.cycles-pp.entry_SYSCALL_64_after_hwframe
18.46 perf-profile.self.cycles-pp.entry_SYSRETQ_unsafe_stack
*************************************************************************
threads 2 sockets (Skylake) with 192G memory
===============================================================
kconfig/mode/nr_task/rootfs/tbox_group/test/testcase:
4-rhel-8.3/process/100%/debian-11.1-x86_64-20220510.cgz/lkp-skl-fpga01/futex2/will-it-scale
Add VERW just before userspace transition")
Use ALTERNATIVE() instead of mds_user_clear static key")
6613d82e617dd7eb8b0c40b2fe3
---------------------------
%stddev
\
-23.3% 175.00 ? 15% perf-c2c.HITM.local
501.86 ?132% sched_debug.cfs_rq:/.removed.load_avg.stddev
3154 ? 11% turbostat.C1
20950 ? 4% proc-vmstat.numa_hint_faults_local
+4.9% 64491 ? 2% proc-vmstat.pgactivate
-2.1% 65339777 will-it-scale.104.processes
-2.1% 628266 will-it-scale.per_process_ops
-2.1% 65339777 will-it-scale.workload
0.23 ? 18% perf-sched.sch_delay.avg.ms.do_task_dead.do_exit.do_group_exit.__x64_sys_exit_group.do_syscall_64
0.16 ? 15% perf-sched.wait_and_delay.avg.ms.irqentry_exit_to_user_mode.asm_sysvec_reschedule_ipi.[unknown].[unknown]
-10.5% 478.00 ? 3% perf-sched.wait_and_delay.count.irqentry_exit_to_user_mode.asm_sysvec_apic_timer_interrupt.[unknown].[unknown]
-16.7% 156.00 ? 10% perf-sched.wait_and_delay.count.irqentry_exit_to_user_mode.asm_sysvec_reschedule_ipi.[unknown].[unknown]
0.16 ? 15% perf-sched.wait_time.avg.ms.irqentry_exit_to_user_mode.asm_sysvec_reschedule_ipi.[unknown].[unknown]
-1.9% 1.271e+10 perf-stat.i.branch-instructions
1.06 perf-stat.i.branch-miss-rate%
+4.8% 1.348e+08 perf-stat.i.branch-misses
3.41 perf-stat.i.cpi
-1.9% 65315938 perf-stat.i.dTLB-load-misses
-1.9% 1.909e+10 perf-stat.i.dTLB-loads
0.00 perf-stat.i.dTLB-store-miss-rate%
-6.6% 54625 perf-stat.i.dTLB-store-misses
-1.9% 1.411e+10 perf-stat.i.dTLB-stores
-8.8% 67017808 ? 4% perf-stat.i.iTLB-load-misses
-2.0% 8.443e+10 perf-stat.i.instructions
1266 ? 4% perf-stat.i.instructions-per-iTLB-miss
0.29 perf-stat.i.ipc
-1.9% 441.52 perf-stat.i.metric.M/sec
-6.8% 179259 ? 6% perf-stat.i.node-load-misses
1.06 perf-stat.overall.branch-miss-rate%
3.41 perf-stat.overall.cpi
0.00 perf-stat.overall.dTLB-store-miss-rate%
1262 ? 4% perf-stat.overall.instructions-per-iTLB-miss
0.29 perf-stat.overall.ipc
-1.9% 1.267e+10 perf-stat.ps.branch-instructions
+4.8% 1.344e+08 perf-stat.ps.branch-misses
-1.9% 65097246 perf-stat.ps.dTLB-load-misses
-1.9% 1.903e+10 perf-stat.ps.dTLB-loads
-6.6% 54460 perf-stat.ps.dTLB-store-misses
-1.9% 1.407e+10 perf-stat.ps.dTLB-stores
-8.8% 66790734 ? 4% perf-stat.ps.iTLB-load-misses
-2.0% 8.415e+10 perf-stat.ps.instructions
-6.8% 178656 ? 6% perf-stat.ps.node-load-misses
-2.2% 2.541e+13 perf-stat.total.instructions
0.83 perf-profile.calltrace.cycles-pp.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe.syscall
51.77 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_after_hwframe.syscall
42.84 perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.syscall
7.61 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_safe_stack.syscall
19.98 perf-profile.calltrace.cycles-pp.syscall_return_via_sysret.syscall
7.68 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64.syscall
99.74 perf-profile.calltrace.cycles-pp.syscall
1.13 perf-profile.calltrace.cycles-pp.try_grab_folio.gup_pte_range.gup_pgd_range.lockless_pages_from_mm.internal_get_user_pages_fast
0.96 perf-profile.calltrace.cycles-pp.futex_hash.futex_q_lock.futex_wait_setup.__futex_wait.futex_wait
1.25 perf-profile.calltrace.cycles-pp._raw_spin_lock.futex_q_lock.futex_wait_setup.__futex_wait.futex_wait
0.87 perf-profile.calltrace.cycles-pp.__pte_offset_map.gup_pte_range.gup_pgd_range.lockless_pages_from_mm.internal_get_user_pages_fast
3.97 perf-profile.calltrace.cycles-pp.futex_q_lock.futex_wait_setup.__futex_wait.futex_wait.do_futex
5.43 perf-profile.calltrace.cycles-pp.gup_pte_range.gup_pgd_range.lockless_pages_from_mm.internal_get_user_pages_fast.get_user_pages_fast
4.47 perf-profile.calltrace.cycles-pp.__get_user_nocheck_4.futex_get_value_locked.futex_wait_setup.__futex_wait.futex_wait
9.79 perf-profile.calltrace.cycles-pp.gup_pgd_range.lockless_pages_from_mm.internal_get_user_pages_fast.get_user_pages_fast.get_futex_key
6.00 perf-profile.calltrace.cycles-pp.futex_get_value_locked.futex_wait_setup.__futex_wait.futex_wait.do_futex
11.39 perf-profile.calltrace.cycles-pp.lockless_pages_from_mm.internal_get_user_pages_fast.get_user_pages_fast.get_futex_key.futex_wait_setup
12.95 perf-profile.calltrace.cycles-pp.internal_get_user_pages_fast.get_user_pages_fast.get_futex_key.futex_wait_setup.__futex_wait
14.50 perf-profile.calltrace.cycles-pp.get_user_pages_fast.get_futex_key.futex_wait_setup.__futex_wait.futex_wait
17.64 perf-profile.calltrace.cycles-pp.get_futex_key.futex_wait_setup.__futex_wait.futex_wait.do_futex
32.70 perf-profile.calltrace.cycles-pp.__x64_sys_futex.do_syscall_64.entry_SYSCALL_64_after_hwframe.syscall
31.60 perf-profile.calltrace.cycles-pp.do_futex.__x64_sys_futex.do_syscall_64.entry_SYSCALL_64_after_hwframe.syscall
31.06 perf-profile.calltrace.cycles-pp.futex_wait.do_futex.__x64_sys_futex.do_syscall_64.entry_SYSCALL_64_after_hwframe
29.60 perf-profile.calltrace.cycles-pp.__futex_wait.futex_wait.do_futex.__x64_sys_futex.do_syscall_64
28.57 perf-profile.calltrace.cycles-pp.futex_wait_setup.__futex_wait.futex_wait.do_futex.__x64_sys_futex
13.64 perf-profile.calltrace.cycles-pp.entry_SYSRETQ_unsafe_stack.syscall
0.89 perf-profile.children.cycles-pp.syscall_exit_to_user_mode
52.89 perf-profile.children.cycles-pp.entry_SYSCALL_64_after_hwframe
41.92 perf-profile.children.cycles-pp.do_syscall_64
10.14 perf-profile.children.cycles-pp.entry_SYSCALL_64
4.09 perf-profile.children.cycles-pp.entry_SYSCALL_64_safe_stack
20.13 perf-profile.children.cycles-pp.syscall_return_via_sysret
0.29 perf-profile.children.cycles-pp.syscall_exit_to_user_mode_prepare
0.06 ? 6% perf-profile.children.cycles-pp.amd_clear_divider
0.39 ? 2% perf-profile.children.cycles-pp.is_valid_gup_args
0.10 ? 4% perf-profile.children.cycles-pp.pud_huge
1.13 perf-profile.children.cycles-pp.try_grab_folio
0.16 ? 3% perf-profile.children.cycles-pp.pmd_huge
1.30 perf-profile.children.cycles-pp._raw_spin_lock
0.98 perf-profile.children.cycles-pp.futex_hash
0.88 perf-profile.children.cycles-pp.__pte_offset_map
4.06 perf-profile.children.cycles-pp.futex_q_lock
5.52 perf-profile.children.cycles-pp.gup_pte_range
5.08 perf-profile.children.cycles-pp.__get_user_nocheck_4
5.49 perf-profile.children.cycles-pp.futex_get_value_locked
9.90 perf-profile.children.cycles-pp.gup_pgd_range
11.44 perf-profile.children.cycles-pp.lockless_pages_from_mm
13.12 perf-profile.children.cycles-pp.internal_get_user_pages_fast
14.70 perf-profile.children.cycles-pp.get_user_pages_fast
17.70 perf-profile.children.cycles-pp.get_futex_key
32.76 perf-profile.children.cycles-pp.__x64_sys_futex
31.67 perf-profile.children.cycles-pp.do_futex
31.16 perf-profile.children.cycles-pp.futex_wait
29.65 perf-profile.children.cycles-pp.__futex_wait
28.70 perf-profile.children.cycles-pp.futex_wait_setup
14.65 perf-profile.children.cycles-pp.entry_SYSRETQ_unsafe_stack
0.54 perf-profile.self.cycles-pp.syscall_exit_to_user_mode
9.08 perf-profile.self.cycles-pp.entry_SYSCALL_64
0.95 perf-profile.self.cycles-pp.__futex_wait
12.00 perf-profile.self.cycles-pp.entry_SYSCALL_64_after_hwframe
20.10 perf-profile.self.cycles-pp.syscall_return_via_sysret
1.43 perf-profile.self.cycles-pp.futex_wait
0.26 perf-profile.self.cycles-pp.syscall_exit_to_user_mode_prepare
1.07 perf-profile.self.cycles-pp.__x64_sys_futex
0.08 perf-profile.self.cycles-pp.futex_setup_timer
0.08 ? 5% perf-profile.self.cycles-pp.pud_huge
0.35 ? 2% perf-profile.self.cycles-pp.is_valid_gup_args
0.05 perf-profile.self.cycles-pp.syscall@plt
1.09 perf-profile.self.cycles-pp.try_grab_folio
0.13 ? 2% perf-profile.self.cycles-pp.pmd_huge
2.46 perf-profile.self.cycles-pp.syscall
0.42 ? 5% perf-profile.self.cycles-pp.futex_get_value_locked
0.84 perf-profile.self.cycles-pp.futex_wait_setup
1.26 perf-profile.self.cycles-pp._raw_spin_lock
0.95 perf-profile.self.cycles-pp.futex_hash
0.87 perf-profile.self.cycles-pp.__pte_offset_map
1.41 perf-profile.self.cycles-pp.lockless_pages_from_mm
1.24 perf-profile.self.cycles-pp.get_user_pages_fast
1.73 perf-profile.self.cycles-pp.internal_get_user_pages_fast
1.81 ? 2% perf-profile.self.cycles-pp.futex_q_lock
2.98 perf-profile.self.cycles-pp.get_futex_key
7.13 perf-profile.self.cycles-pp.do_syscall_64
4.16 perf-profile.self.cycles-pp.gup_pgd_range
3.46 perf-profile.self.cycles-pp.gup_pte_range
5.01 perf-profile.self.cycles-pp.__get_user_nocheck_4
14.51 perf-profile.self.cycles-pp.entry_SYSRETQ_unsafe_stack
*************************************************************************
threads 2 sockets (Skylake) with 192G memory
===============================================================
kconfig/mode/nr_task/rootfs/tbox_group/test/testcase:
4-rhel-8.3/process/100%/debian-11.1-x86_64-20220510.cgz/lkp-skl-fpga01/futex3/will-it-scale
Add VERW just before userspace transition")
Use ALTERNATIVE() instead of mds_user_clear static key")
6613d82e617dd7eb8b0c40b2fe3
---------------------------
%stddev
\
17.39 ? 9% sched_debug.cfs_rq:/.removed.runnable_avg.stddev
17.19 ? 9% sched_debug.cfs_rq:/.removed.util_avg.stddev
0.35 ? 17% perf-sched.wait_and_delay.avg.ms.irqentry_exit_to_user_mode.asm_sysvec_apic_timer_interrupt.[unknown].[unknown]
+14.2% 323.67 ? 4% perf-sched.wait_and_delay.count.irqentry_exit_to_user_mode.asm_sysvec_reschedule_ipi.[unknown].[unknown]
0.35 ? 17% perf-sched.wait_time.avg.ms.irqentry_exit_to_user_mode.asm_sysvec_apic_timer_interrupt.[unknown].[unknown]
+3.7% 79094461 will-it-scale.104.processes
+3.7% 760523 will-it-scale.per_process_ops
+3.7% 79094461 will-it-scale.workload
-2.2% 54361 proc-vmstat.nr_active_anon
-2.3% 56792 proc-vmstat.nr_shmem
-2.2% 54361 proc-vmstat.nr_zone_active_anon
-3.5% 55794 proc-vmstat.pgactivate
+3.7% 4.793e+09 perf-stat.i.branch-instructions
2.10 perf-stat.i.branch-miss-rate%
+23.8% 1.009e+08 perf-stat.i.branch-misses
7.59 perf-stat.i.cpi
+3.7% 79030797 perf-stat.i.dTLB-load-misses
+3.7% 9.18e+09 perf-stat.i.dTLB-loads
0.00 perf-stat.i.dTLB-store-miss-rate%
+2.1% 76523 perf-stat.i.dTLB-store-misses
+3.6% 6.954e+09 perf-stat.i.dTLB-stores
+3.3% 3.794e+10 perf-stat.i.instructions
0.13 perf-stat.i.ipc
+3.6% 201.22 perf-stat.i.metric.M/sec
78.12 perf-stat.i.node-store-miss-rate%
2.10 perf-stat.overall.branch-miss-rate%
7.59 perf-stat.overall.cpi
0.00 perf-stat.overall.dTLB-store-miss-rate%
0.13 perf-stat.overall.ipc
+3.6% 4.778e+09 perf-stat.ps.branch-instructions
+23.8% 1.005e+08 perf-stat.ps.branch-misses
+3.7% 78766248 perf-stat.ps.dTLB-load-misses
+3.7% 9.15e+09 perf-stat.ps.dTLB-loads
+2.1% 76323 perf-stat.ps.dTLB-store-misses
+3.6% 6.931e+09 perf-stat.ps.dTLB-stores
+3.3% 3.781e+10 perf-stat.ps.instructions
+3.2% 1.141e+13 perf-stat.total.instructions
13.91 perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe.syscall
29.00 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_after_hwframe.syscall
1.24 perf-profile.calltrace.cycles-pp.syscall_exit_to_user_mode.do_syscall_64.entry_SYSCALL_64_after_hwframe.syscall
6.22 perf-profile.calltrace.cycles-pp.__x64_sys_futex.do_syscall_64.entry_SYSCALL_64_after_hwframe.syscall
4.85 perf-profile.calltrace.cycles-pp.do_futex.__x64_sys_futex.do_syscall_64.entry_SYSCALL_64_after_hwframe.syscall
4.18 perf-profile.calltrace.cycles-pp.futex_wake.do_futex.__x64_sys_futex.do_syscall_64.entry_SYSCALL_64_after_hwframe
1.17 perf-profile.calltrace.cycles-pp.get_futex_key.futex_wake.do_futex.__x64_sys_futex.do_syscall_64
1.31 perf-profile.calltrace.cycles-pp.futex_hash.futex_wake.do_futex.__x64_sys_futex.do_syscall_64
28.48 perf-profile.calltrace.cycles-pp.syscall_return_via_sysret.syscall
96.55 perf-profile.calltrace.cycles-pp.syscall
6.99 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_safe_stack.syscall
9.32 perf-profile.calltrace.cycles-pp.entry_SYSCALL_64.syscall
3.94 ? 7% perf-profile.calltrace.cycles-pp.testcase
21.12 perf-profile.calltrace.cycles-pp.entry_SYSRETQ_unsafe_stack.syscall
13.28 perf-profile.children.cycles-pp.do_syscall_64
29.91 perf-profile.children.cycles-pp.entry_SYSCALL_64_after_hwframe
1.32 perf-profile.children.cycles-pp.syscall_exit_to_user_mode
4.32 perf-profile.children.cycles-pp.futex_wake
4.90 perf-profile.children.cycles-pp.do_futex
6.27 perf-profile.children.cycles-pp.__x64_sys_futex
1.20 perf-profile.children.cycles-pp.get_futex_key
1.32 perf-profile.children.cycles-pp.futex_hash
28.68 perf-profile.children.cycles-pp.syscall_return_via_sysret
97.28 perf-profile.children.cycles-pp.syscall
3.96 perf-profile.children.cycles-pp.entry_SYSCALL_64_safe_stack
11.08 perf-profile.children.cycles-pp.entry_SYSCALL_64
0.36 perf-profile.children.cycles-pp.syscall_exit_to_user_mode_prepare
0.09 ? 4% perf-profile.children.cycles-pp.amd_clear_divider
0.10 ? 3% perf-profile.children.cycles-pp.syscall@plt
3.30 ? 8% perf-profile.children.cycles-pp.testcase
22.37 perf-profile.children.cycles-pp.entry_SYSRETQ_unsafe_stack
0.88 perf-profile.self.cycles-pp.syscall_exit_to_user_mode
1.16 perf-profile.self.cycles-pp.get_futex_key
1.23 perf-profile.self.cycles-pp.futex_hash
28.57 perf-profile.self.cycles-pp.syscall_return_via_sysret
1.87 perf-profile.self.cycles-pp.futex_wake
9.80 perf-profile.self.cycles-pp.entry_SYSCALL_64
0.32 ? 3% perf-profile.self.cycles-pp.syscall_exit_to_user_mode_prepare
0.07 ? 5% perf-profile.self.cycles-pp.amd_clear_divider
0.10 ? 4% perf-profile.self.cycles-pp.syscall@plt
0.62 perf-profile.self.cycles-pp.do_futex
0.91 ? 2% perf-profile.self.cycles-pp.entry_SYSCALL_64_safe_stack
2.63 ? 9% perf-profile.self.cycles-pp.testcase
4.80 perf-profile.self.cycles-pp.do_syscall_64
5.51 ? 3% perf-profile.self.cycles-pp.syscall
17.39 perf-profile.self.cycles-pp.entry_SYSCALL_64_after_hwframe
22.18 perf-profile.self.cycles-pp.entry_SYSRETQ_unsafe_stack
estimated based on internal Intel analysis and are provided
purposes only. Any difference in system hardware or software
may affect actual performance.
om/intel/lkp-tests/wiki">https://github.com/intel/lkp-tests/wiki

2024-03-04 17:59:07

by Dave Hansen

[permalink] [raw]

Subject: Re: [linus:master] [x86/bugs] 6613d82e61: stress-ng.mutex.ops_per_sec -7.9% regression

On 3/3/24 21:53, kernel test robot wrote:
> kernel test robot noticed a -7.9% regression of stress-ng.mutex.ops_per_sec on:
>
> commit: 6613d82e617dd7eb8b0c40b2fe3acea655b1d611 ("x86/bugs: Use ALTERNATIVE() instead of mds_user_clear static key")
> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master

This _looks_ like noise to me.

Some benchmarks went up, some went down. The differential profile shows
random gunk that basically amounts to "my computer is slow" because it's
mostly things that change when the result changes, like:

> 182670 +9.0% 199032 stress-ng.mutex.nanosecs_per_mutex

Does anyone think there's something substantial to chase after here?

2024-03-05 05:51:40

by Feng Tang

[permalink] [raw]

Subject: Re: [linus:master] [x86/bugs] 6613d82e61: stress-ng.mutex.ops_per_sec -7.9% regression

Hi Dave,

On Mon, Mar 04, 2024 at 09:58:53AM -0800, Dave Hansen wrote:
> On 3/3/24 21:53, kernel test robot wrote:
> > kernel test robot noticed a -7.9% regression of stress-ng.mutex.ops_per_sec on:
> >
> > commit: 6613d82e617dd7eb8b0c40b2fe3acea655b1d611 ("x86/bugs: Use ALTERNATIVE() instead of mds_user_clear static key")
> > https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master
>
> This _looks_ like noise to me.
>
> Some benchmarks went up, some went down. The differential profile shows
> random gunk that basically amounts to "my computer is slow" because it's
> mostly things that change when the result changes, like:
>
> > 182670 +9.0% 199032 stress-ng.mutex.nanosecs_per_mutex
>
> Does anyone think there's something substantial to chase after here?

We further checked this, and it seems to be another case of data/text
alignment effect, that 6613d82e617d removes staic key 'mds_user_clear'
which sits in '.bss' section and change the address alignment of
following data in that section.

With below debug patch to restore the alignment, we can see the
performance is recovered:

a0e2dab44d22b913 6613d82e617dd7eb8b0c40b2fe3 398e7f0da8595354dc330938831
---------------- --------------------------- ---------------------------

302318 -7.9% 278364 +0.3% 303161 stress-ng.mutex.ops_per_sec

---
diff --git a/arch/x86/kernel/cpu/bugs.c b/arch/x86/kernel/cpu/bugs.c
index 48d049cd74e7..1876865dc954 100644
--- a/arch/x86/kernel/cpu/bugs.c
+++ b/arch/x86/kernel/cpu/bugs.c
@@ -111,6 +111,9 @@ DEFINE_STATIC_KEY_FALSE(switch_mm_cond_ibpb);
/* Control unconditional IBPB in switch_mm() */
DEFINE_STATIC_KEY_FALSE(switch_mm_always_ibpb);

+DEFINE_STATIC_KEY_FALSE(test_static_key);
+EXPORT_SYMBOL_GPL(test_static_key);
+
/* Control MDS CPU buffer clear before idling (halt, mwait) */
DEFINE_STATIC_KEY_FALSE(mds_idle_clear);
EXPORT_SYMBOL_GPL(mds_idle_clear);
---

There was another similar case which changed the alignment of
percpu section:
https://lore.kernel.org/lkml/ZSeF6T0mkrH5pOgD@feng-clx/

Thanks,
Feng