Re: [LKP] [function_graph] 8114865ff8: netperf.Throughput_tps -1.1% regression
From: Steven Rostedt
Date: Tue Dec 11 2018 - 09:05:39 EST
On Tue, 11 Dec 2018 17:16:31 +0800
kernel test robot <rong.a.chen@xxxxxxxxx> wrote:
> Greeting,
>
> FYI, we noticed a -1.1% regression of netperf.Throughput_tps due to commit:
>
>
> commit: 8114865ff82e200b383e46821c25cb0625b842b5 ("function_graph: Create function_graph_enter() to consolidate architecture code")
> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master
This must be a false positive, as the commit in question simply adds a
function that nothing calls.
-- Steve
>
> in testcase: netperf
> on test machine: 8 threads Intel(R) Core(TM) i7-4770 CPU @ 3.40GHz with 8G memory
> with following parameters:
>
> ip: ipv4
> runtime: 300s
> nr_threads: 200%
> cluster: cs-localhost
> test: TCP_CRR
> ucode: 0x25
> cpufreq_governor: performance
>
> test-description: Netperf is a benchmark that can be use to measure various aspect of networking performance.
> test-url: http://www.netperf.org/netperf/
>
>
>
> 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
>
> =========================================================================================
> cluster/compiler/cpufreq_governor/ip/kconfig/nr_threads/rootfs/runtime/tbox_group/test/testcase/ucode:
> cs-localhost/gcc-7/performance/ipv4/x86_64-rhel-7.2/200%/debian-x86_64-2018-04-03.cgz/300s/lkp-hsw-d01/TCP_CRR/netperf/0x25
>
> commit:
> v4.20-rc3
> 8114865ff8 ("function_graph: Create function_graph_enter() to consolidate architecture code")
>
> v4.20-rc3 8114865ff82e200b383e46821c
> ---------------- --------------------------
> fail:runs %reproduction fail:runs
> | | |
> 2:5 -40% :4 dmesg.WARNING:at#for_ip_swapgs_restore_regs_and_return_to_usermode/0x
> 2:5 -40% :4 dmesg.WARNING:stack_recursion
> :5 20% 1:4 kmsg.drm:drm_atomic_helper_wait_for_dependencies[drm_kms_helper]]*ERROR*[CONNECTOR:#:VGA-#]flip_done_timed_out
> :5 20% 1:4 kmsg.drm:drm_atomic_helper_wait_for_dependencies[drm_kms_helper]]*ERROR*[CRTC:#:pipe_A]flip_done_timed_out
> :5 20% 1:4 kmsg.drm:drm_atomic_helper_wait_for_dependencies[drm_kms_helper]]*ERROR*[PLANE:#:primary_A]flip_done_timed_out
> :5 20% 1:4 kmsg.drm:drm_atomic_helper_wait_for_flip_done[drm_kms_helper]]*ERROR*[CRTC:#:pipe_A]flip_done_timed_out
> 1:5 -10% 1:4 perf-profile.calltrace.cycles-pp.dequeue_entity.dequeue_task_fair.__sched_text_start.schedule.schedule_timeout
> 26:5 -93% 21:4 perf-profile.children.cycles-pp.schedule_timeout
> 0:5 -1% 0:4 perf-profile.self.cycles-pp.schedule_timeout
> %stddev %change %stddev
> \ | \
> 131277 -1.1% 129833 netperf.Throughput_total_tps
> 8204 -1.1% 8114 netperf.Throughput_tps
> 104.96 -1.6% 103.25 netperf.time.user_time
> 39321448 -1.1% 38880992 netperf.time.voluntary_context_switches
> 39383220 -1.1% 38949992 netperf.workload
> 327.02 -100.0% 0.00 uptime.boot
> 179.66 Â 8% -100.0% 0.00 uptime.idle
> 4174 Â 10% +12.4% 4692 Â 12% slabinfo.anon_vma.active_objs
> 4174 Â 10% +12.4% 4692 Â 12% slabinfo.anon_vma.num_objs
> 39.00 +229.5% 128.50 Â 27% slabinfo.buffer_head.active_objs
> 39.00 +229.5% 128.50 Â 27% slabinfo.buffer_head.num_objs
> 0.43 Â 9% -19.6% 0.34 Â 13% sched_debug.cfs_rq:/.nr_running.stddev
> 12140 Â114% +330.4% 52248 Â 73% sched_debug.cpu.avg_idle.min
> 0.88 Â 8% +24.1% 1.09 Â 4% sched_debug.cpu.clock.stddev
> 0.88 Â 8% +24.1% 1.09 Â 4% sched_debug.cpu.clock_task.stddev
> 163.75 Â 2% +6.7% 174.67 Â 4% sched_debug.cpu.cpu_load[0].avg
> 238.17 Â 3% +24.7% 296.96 Â 17% sched_debug.cpu.cpu_load[0].max
> 215.03 Â 2% +16.7% 251.04 Â 8% sched_debug.cpu.cpu_load[1].max
> 20.65 Â 12% +19.2% 24.61 Â 7% sched_debug.cpu.cpu_load[2].stddev
> 62804 +2.0% 64053 proc-vmstat.nr_active_anon
> 1.00 +10950.0% 110.50 Â 73% proc-vmstat.nr_active_file
> 61258 +2.0% 62489 proc-vmstat.nr_anon_pages
> 23.00 +303.3% 92.75 Â 38% proc-vmstat.nr_inactive_file
> 3854 +8.2% 4170 proc-vmstat.nr_kernel_stack
> 9474 +2.0% 9666 proc-vmstat.nr_slab_reclaimable
> 8753 +2.7% 8988 proc-vmstat.nr_slab_unreclaimable
> 62807 +2.0% 64055 proc-vmstat.nr_zone_active_anon
> 1.00 +10950.0% 110.50 Â 73% proc-vmstat.nr_zone_active_file
> 23.00 +303.3% 92.75 Â 38% proc-vmstat.nr_zone_inactive_file
> 461366 +4.9% 483771 Â 2% proc-vmstat.numa_hit
> 461366 +4.9% 483771 Â 2% proc-vmstat.numa_local
> 8.459e+11 -1.3% 8.347e+11 perf-stat.branch-instructions
> 81655384 -1.1% 80758043 perf-stat.context-switches
> 2.04 +1.3% 2.07 perf-stat.cpi
> 56426 Â 3% -8.7% 51533 Â 5% perf-stat.cpu-migrations
> 0.10 Â 2% +0.0 0.10 Â 3% perf-stat.dTLB-store-miss-rate%
> 8.433e+11 -1.1% 8.337e+11 perf-stat.dTLB-stores
> 75.15 +1.9 77.05 perf-stat.iTLB-load-miss-rate%
> 8.967e+08 -10.2% 8.056e+08 perf-stat.iTLB-loads
> 4.324e+12 -1.3% 4.267e+12 perf-stat.instructions
> 0.49 -1.2% 0.48 perf-stat.ipc
> 390422 +0.8% 393664 perf-stat.minor-faults
> 17227761 Â 3% +15.0% 19812916 Â 2% perf-stat.node-stores
> 390422 +0.8% 393654 perf-stat.page-faults
> 1.74 Â 15% -0.3 1.43 Â 2% perf-profile.calltrace.cycles-pp.security_sock_rcv_skb.sk_filter_trim_cap.tcp_filter.tcp_v4_rcv.ip_local_deliver_finish
> 1.53 Â 6% -0.2 1.35 Â 2% perf-profile.calltrace.cycles-pp.selinux_socket_sock_rcv_skb.security_sock_rcv_skb.sk_filter_trim_cap.tcp_filter.tcp_v4_rcv
> 1.06 Â 3% -0.1 0.99 Â 2% perf-profile.calltrace.cycles-pp.tcp_send_fin.tcp_close.inet_release.__sock_release.sock_close
> 2.24 +0.1 2.30 perf-profile.calltrace.cycles-pp.tcp_rcv_established.tcp_v4_do_rcv.__release_sock.release_sock.tcp_sendmsg
> 0.88 Â 2% +0.1 0.95 Â 5% perf-profile.calltrace.cycles-pp.__ip_queue_xmit.__tcp_transmit_skb.tcp_fin.tcp_data_queue.tcp_rcv_state_process
> 1.08 Â 2% +0.1 1.17 Â 4% perf-profile.calltrace.cycles-pp.__tcp_transmit_skb.tcp_fin.tcp_data_queue.tcp_rcv_state_process.tcp_v4_do_rcv
> 2.54 +0.1 2.65 perf-profile.calltrace.cycles-pp.__release_sock.release_sock.tcp_sendmsg.sock_sendmsg.__sys_sendto
> 2.68 Â 2% +0.1 2.80 perf-profile.calltrace.cycles-pp.release_sock.tcp_sendmsg.sock_sendmsg.__sys_sendto.__x64_sys_sendto
> 3.03 Â 2% +0.1 3.16 perf-profile.calltrace.cycles-pp.tcp_fin.tcp_data_queue.tcp_rcv_state_process.tcp_v4_do_rcv.tcp_v4_rcv
> 0.71 Â 7% +0.1 0.84 Â 14% perf-profile.calltrace.cycles-pp.ip_output.tcp_v4_send_synack.tcp_conn_request.tcp_rcv_state_process.tcp_v4_do_rcv
> 3.19 Â 2% +0.2 3.36 perf-profile.calltrace.cycles-pp.tcp_data_queue.tcp_rcv_state_process.tcp_v4_do_rcv.tcp_v4_rcv.ip_local_deliver_finish
> 0.33 Â 82% +0.3 0.63 Â 13% perf-profile.calltrace.cycles-pp.ret_from_fork
> 0.33 Â 82% +0.3 0.63 Â 13% perf-profile.calltrace.cycles-pp.kthread.ret_from_fork
> 12.33 +0.4 12.77 Â 2% perf-profile.calltrace.cycles-pp.tcp_v4_connect.__inet_stream_connect.inet_stream_connect.__sys_connect.__x64_sys_connect
> 30.28 +0.5 30.73 perf-profile.calltrace.cycles-pp.ip_local_deliver.ip_rcv.__netif_receive_skb_one_core.process_backlog.net_rx_action
> 7.34 Â 2% +0.5 7.79 Â 2% perf-profile.calltrace.cycles-pp.ip_output.__ip_queue_xmit.__tcp_transmit_skb.tcp_connect.tcp_v4_connect
> 32.67 +0.5 33.13 perf-profile.calltrace.cycles-pp.__netif_receive_skb_one_core.process_backlog.net_rx_action.__softirqentry_text_start.do_softirq_own_stack
> 7.21 Â 2% +0.5 7.69 Â 3% perf-profile.calltrace.cycles-pp.ip_finish_output2.ip_output.__ip_queue_xmit.__tcp_transmit_skb.tcp_connect
> 8.80 Â 2% +0.5 9.28 perf-profile.calltrace.cycles-pp.tcp_rcv_state_process.tcp_v4_do_rcv.tcp_v4_rcv.ip_local_deliver_finish.ip_local_deliver
> 10.34 +0.5 10.88 Â 3% perf-profile.calltrace.cycles-pp.tcp_connect.tcp_v4_connect.__inet_stream_connect.inet_stream_connect.__sys_connect
> 34.75 +0.5 35.30 perf-profile.calltrace.cycles-pp.__softirqentry_text_start.do_softirq_own_stack.do_softirq.__local_bh_enable_ip.ip_finish_output2
> 34.88 +0.6 35.43 perf-profile.calltrace.cycles-pp.do_softirq_own_stack.do_softirq.__local_bh_enable_ip.ip_finish_output2.ip_output
> 16.15 +0.6 16.70 Â 2% perf-profile.calltrace.cycles-pp.tcp_v4_do_rcv.tcp_v4_rcv.ip_local_deliver_finish.ip_local_deliver.ip_rcv
> 8.86 Â 2% +0.6 9.44 Â 2% perf-profile.calltrace.cycles-pp.__tcp_transmit_skb.tcp_connect.tcp_v4_connect.__inet_stream_connect.inet_stream_connect
> 35.25 +0.6 35.84 perf-profile.calltrace.cycles-pp.do_softirq.__local_bh_enable_ip.ip_finish_output2.ip_output.__ip_queue_xmit
> 34.18 +0.6 34.78 perf-profile.calltrace.cycles-pp.net_rx_action.__softirqentry_text_start.do_softirq_own_stack.do_softirq.__local_bh_enable_ip
> 8.31 Â 2% +0.6 8.92 Â 2% perf-profile.calltrace.cycles-pp.__ip_queue_xmit.__tcp_transmit_skb.tcp_connect.tcp_v4_connect.__inet_stream_connect
> 35.35 +0.7 36.01 perf-profile.calltrace.cycles-pp.__local_bh_enable_ip.ip_finish_output2.ip_output.__ip_queue_xmit.__tcp_transmit_skb
> 2.33 Â 5% -0.2 2.10 perf-profile.children.cycles-pp.security_sock_rcv_skb
> 2.21 Â 5% -0.2 1.99 perf-profile.children.cycles-pp.selinux_socket_sock_rcv_skb
> 2.51 Â 5% -0.2 2.30 perf-profile.children.cycles-pp.sk_filter_trim_cap
> 1.59 Â 6% -0.2 1.42 perf-profile.children.cycles-pp.selinux_sock_rcv_skb_compat
> 1.91 Â 6% -0.2 1.75 perf-profile.children.cycles-pp._raw_spin_lock
> 1.49 Â 4% -0.1 1.36 Â 2% perf-profile.children.cycles-pp.__kfree_skb
> 0.61 Â 9% -0.1 0.50 Â 9% perf-profile.children.cycles-pp.sock_def_wakeup
> 0.97 Â 3% -0.1 0.88 Â 6% perf-profile.children.cycles-pp.avc_has_perm
> 0.56 Â 6% -0.1 0.47 Â 4% perf-profile.children.cycles-pp.__x64_sys_close
> 0.37 Â 10% -0.1 0.29 Â 12% perf-profile.children.cycles-pp._copy_from_iter_full
> 0.62 Â 3% -0.1 0.54 Â 3% perf-profile.children.cycles-pp.__fget_light
> 0.26 Â 13% -0.1 0.19 Â 9% perf-profile.children.cycles-pp.security_file_alloc
> 0.42 Â 2% -0.1 0.35 Â 13% perf-profile.children.cycles-pp.copy_user_generic_unrolled
> 0.23 Â 11% -0.1 0.16 Â 9% perf-profile.children.cycles-pp.selinux_file_alloc_security
> 1.06 Â 3% -0.1 1.00 Â 2% perf-profile.children.cycles-pp.tcp_send_fin
> 0.11 Â 18% -0.1 0.05 Â 58% perf-profile.children.cycles-pp.sel_netnode_find
> 0.13 Â 16% -0.1 0.07 Â 6% perf-profile.children.cycles-pp.sel_netnode_sid
> 0.39 Â 6% -0.1 0.34 Â 4% perf-profile.children.cycles-pp.skb_release_data
> 0.34 Â 10% -0.1 0.29 Â 11% perf-profile.children.cycles-pp.filp_close
> 0.20 Â 5% -0.0 0.15 Â 12% perf-profile.children.cycles-pp.tcp_update_skb_after_send
> 0.36 Â 7% -0.0 0.32 Â 5% perf-profile.children.cycles-pp.__update_load_avg_se
> 0.12 Â 14% -0.0 0.09 Â 16% perf-profile.children.cycles-pp.__tcp_ack_snd_check
> 0.15 Â 14% -0.0 0.11 Â 9% perf-profile.children.cycles-pp.init_timer_key
> 0.15 Â 4% -0.0 0.12 Â 10% perf-profile.children.cycles-pp.remove_wait_queue
> 0.08 Â 17% -0.0 0.06 perf-profile.children.cycles-pp.ip_build_and_send_pkt
> 0.15 Â 6% +0.0 0.17 Â 3% perf-profile.children.cycles-pp.__usecs_to_jiffies
> 0.04 Â 51% +0.0 0.07 Â 6% perf-profile.children.cycles-pp.find_exception
> 0.09 Â 11% +0.0 0.11 Â 17% perf-profile.children.cycles-pp.tcp_newly_delivered
> 0.09 Â 15% +0.0 0.11 Â 7% perf-profile.children.cycles-pp.xfrm_lookup_with_ifid
> 0.12 Â 15% +0.0 0.15 Â 11% perf-profile.children.cycles-pp.get_nohz_timer_target
> 0.17 Â 11% +0.0 0.20 Â 11% perf-profile.children.cycles-pp.tcp_md5_do_lookup
> 0.02 Â123% +0.0 0.06 Â 16% perf-profile.children.cycles-pp.security_sk_classify_flow
> 0.08 Â 21% +0.0 0.11 Â 4% perf-profile.children.cycles-pp._raw_spin_trylock
> 0.04 Â 83% +0.0 0.08 Â 14% perf-profile.children.cycles-pp.activate_task
> 0.05 Â 54% +0.0 0.09 Â 24% perf-profile.children.cycles-pp.tcp_urg
> 0.19 Â 8% +0.0 0.23 Â 8% perf-profile.children.cycles-pp.tcp_rearm_rto
> 0.01 Â200% +0.0 0.06 Â 9% perf-profile.children.cycles-pp.skb_free_head
> 0.10 Â 9% +0.0 0.15 Â 8% perf-profile.children.cycles-pp.xfrm_lookup_route
> 0.39 Â 5% +0.1 0.44 Â 8% perf-profile.children.cycles-pp.ip_rcv_core
> 0.14 Â 19% +0.1 0.19 Â 22% perf-profile.children.cycles-pp.__inet_lookup_listener
> 0.01 Â200% +0.1 0.06 Â 20% perf-profile.children.cycles-pp.blkcg_maybe_throttle_current
> 0.00 +0.1 0.05 Â 8% perf-profile.children.cycles-pp.ip_mc_drop_socket
> 0.58 Â 4% +0.1 0.64 Â 3% perf-profile.children.cycles-pp.sk_prot_alloc
> 0.60 Â 4% +0.1 0.66 Â 4% perf-profile.children.cycles-pp.__might_sleep
> 0.15 Â 25% +0.1 0.21 Â 20% perf-profile.children.cycles-pp.eth_type_trans
> 1.13 Â 2% +0.1 1.21 Â 2% perf-profile.children.cycles-pp.inet_csk_destroy_sock
> 0.78 Â 2% +0.1 0.87 Â 8% perf-profile.children.cycles-pp.tcp_mstamp_refresh
> 0.52 Â 8% +0.1 0.63 Â 13% perf-profile.children.cycles-pp.kthread
> 0.52 Â 8% +0.1 0.64 Â 13% perf-profile.children.cycles-pp.ret_from_fork
> 0.56 Â 11% +0.1 0.68 Â 6% perf-profile.children.cycles-pp.ip_finish_output
> 3.44 +0.1 3.57 perf-profile.children.cycles-pp.tcp_fin
> 4.81 +0.1 4.95 perf-profile.children.cycles-pp.tcp_data_queue
> 12.35 +0.4 12.78 Â 2% perf-profile.children.cycles-pp.tcp_v4_connect
> 32.75 +0.5 33.20 perf-profile.children.cycles-pp.__netif_receive_skb_one_core
> 10.35 +0.5 10.89 Â 3% perf-profile.children.cycles-pp.tcp_connect
> 34.90 +0.6 35.45 perf-profile.children.cycles-pp.do_softirq_own_stack
> 34.29 +0.6 34.86 perf-profile.children.cycles-pp.net_rx_action
> 45.59 +0.6 46.19 perf-profile.children.cycles-pp.__tcp_transmit_skb
> 35.37 +0.6 35.99 perf-profile.children.cycles-pp.__softirqentry_text_start
> 42.97 +0.7 43.69 perf-profile.children.cycles-pp.__ip_queue_xmit
> 35.89 +0.7 36.62 perf-profile.children.cycles-pp.__local_bh_enable_ip
> 41.33 +0.8 42.10 perf-profile.children.cycles-pp.ip_output
> 39.17 +0.8 39.94 perf-profile.children.cycles-pp.ip_finish_output2
> 0.88 Â 4% -0.1 0.74 Â 3% perf-profile.self.cycles-pp.tcp_write_xmit
> 0.61 Â 2% -0.1 0.54 Â 3% perf-profile.self.cycles-pp.__fget_light
> 0.23 Â 5% -0.1 0.15 Â 18% perf-profile.self.cycles-pp.tcp_v4_connect
> 0.93 Â 3% -0.1 0.86 Â 6% perf-profile.self.cycles-pp.avc_has_perm
> 0.41 Â 2% -0.1 0.34 Â 12% perf-profile.self.cycles-pp.copy_user_generic_unrolled
> 0.11 Â 18% -0.1 0.05 Â 58% perf-profile.self.cycles-pp.sel_netnode_find
> 0.39 Â 8% -0.1 0.32 Â 6% perf-profile.self.cycles-pp.selinux_ip_postroute
> 0.11 Â 24% -0.1 0.05 Â 58% perf-profile.self.cycles-pp.security_d_instantiate
> 0.16 Â 7% -0.1 0.11 Â 13% perf-profile.self.cycles-pp.tcp_send_fin
> 0.35 Â 6% -0.1 0.30 Â 6% perf-profile.self.cycles-pp.__update_load_avg_se
> 0.10 Â 15% -0.0 0.06 Â 58% perf-profile.self.cycles-pp.__x64_sys_close
> 0.07 Â 16% -0.0 0.03 Â100% perf-profile.self.cycles-pp.inet_release
> 0.35 Â 7% -0.0 0.31 Â 2% perf-profile.self.cycles-pp.tcp_parse_options
> 0.09 Â 15% -0.0 0.05 Â 58% perf-profile.self.cycles-pp.__sys_socket
> 0.13 Â 8% -0.0 0.10 Â 17% perf-profile.self.cycles-pp.sk_stream_alloc_skb
> 0.12 Â 13% -0.0 0.08 Â 19% perf-profile.self.cycles-pp.__tcp_ack_snd_check
> 0.14 Â 12% -0.0 0.10 Â 12% perf-profile.self.cycles-pp.skb_release_all
> 0.16 Â 10% -0.0 0.13 Â 3% perf-profile.self.cycles-pp.__inet_bind
> 0.18 Â 8% -0.0 0.15 Â 5% perf-profile.self.cycles-pp.sk_clone_lock
> 0.16 Â 5% -0.0 0.14 Â 11% perf-profile.self.cycles-pp.__wake_up_common
> 0.09 Â 8% +0.0 0.11 Â 11% perf-profile.self.cycles-pp.ip_local_out
> 0.08 Â 12% +0.0 0.11 Â 9% perf-profile.self.cycles-pp.xfrm_lookup_with_ifid
> 0.11 Â 8% +0.0 0.14 Â 8% perf-profile.self.cycles-pp.__usecs_to_jiffies
> 0.16 Â 10% +0.0 0.19 Â 4% perf-profile.self.cycles-pp.tcp_rearm_rto
> 0.07 Â 6% +0.0 0.11 Â 14% perf-profile.self.cycles-pp.__dentry_kill
> 0.24 Â 7% +0.0 0.28 Â 6% perf-profile.self.cycles-pp.tcp_ack_update_rtt
> 0.08 Â 12% +0.0 0.11 Â 15% perf-profile.self.cycles-pp.tcp_newly_delivered
> 0.14 Â 16% +0.0 0.17 Â 13% perf-profile.self.cycles-pp.skb_push
> 0.06 Â 54% +0.0 0.10 Â 11% perf-profile.self.cycles-pp._raw_spin_unlock_bh
> 0.08 Â 11% +0.0 0.12 Â 17% perf-profile.self.cycles-pp.tcp_init_transfer
> 0.07 Â 23% +0.0 0.11 Â 4% perf-profile.self.cycles-pp._raw_spin_trylock
> 0.14 Â 13% +0.0 0.18 Â 9% perf-profile.self.cycles-pp.account_entity_dequeue
> 0.01 Â200% +0.0 0.06 Â 9% perf-profile.self.cycles-pp.skb_free_head
> 0.38 Â 5% +0.1 0.43 Â 7% perf-profile.self.cycles-pp.ip_rcv_core
> 0.06 Â 12% +0.1 0.11 Â 25% perf-profile.self.cycles-pp.tcp_filter
> 0.14 Â 19% +0.1 0.19 Â 22% perf-profile.self.cycles-pp.__inet_lookup_listener
> 0.00 +0.1 0.05 Â 8% perf-profile.self.cycles-pp.ip_mc_drop_socket
> 0.10 Â 26% +0.1 0.15 Â 11% perf-profile.self.cycles-pp.tcp_make_synack
> 0.03 Â 82% +0.1 0.09 Â 24% perf-profile.self.cycles-pp.tcp_urg
> 0.51 Â 6% +0.1 0.57 Â 3% perf-profile.self.cycles-pp.__might_sleep
> 0.14 Â 28% +0.1 0.21 Â 20% perf-profile.self.cycles-pp.eth_type_trans
> 0.54 Â 5% +0.1 0.61 Â 4% perf-profile.self.cycles-pp.__kmalloc_node_track_caller
> 0.44 Â 5% +0.1 0.51 Â 4% perf-profile.self.cycles-pp.enqueue_to_backlog
> 0.45 Â 10% +0.1 0.53 perf-profile.self.cycles-pp.net_rx_action
> 0.28 Â 8% +0.1 0.36 Â 6% perf-profile.self.cycles-pp.ip_local_deliver_finish
> 0.70 Â 7% +0.1 0.79 Â 6% perf-profile.self.cycles-pp.__dev_queue_xmit
> 0.52 Â 4% +0.1 0.64 Â 9% perf-profile.self.cycles-pp.mod_timer
> 1.54 Â 3% +0.1 1.67 Â 3% perf-profile.self.cycles-pp.__tcp_transmit_skb
> 0.28 Â 5% +0.1 0.41 Â 8% perf-profile.self.cycles-pp.ip_finish_output
> 0.68 Â 5% +0.1 0.83 Â 4% perf-profile.self.cycles-pp.__local_bh_enable_ip
> 0.22 Â 10% +0.3 0.49 Â 3% perf-profile.self.cycles-pp.dev_hard_start_xmit
>
>
>
> netperf.time.voluntary_context_switches
>
> 4e+07 O-+O-O--O--O-O----O--O------------O--O--O-O--O--O-O--O-O--O-------+
> | O O |
> 3.5e+07 +-+ |
> 3e+07 +-+ |
> | |
> 2.5e+07 +-+ |
> | |
> 2e+07 +-+ |
> | |
> 1.5e+07 +-+ |
> 1e+07 +-+ |
> | |
> 5e+06 +-+ |
> | |
> 0 +-+-------------O-------O----O------------------------------------+
>
>
> [*] 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