* [linus:master] [timers/migration] 61f7fdf8fd: netperf.Throughput_Mbps 2.7% improvement
@ 2024-05-15 8:06 kernel test robot
2024-05-15 14:14 ` Frederic Weisbecker
0 siblings, 1 reply; 3+ messages in thread
From: kernel test robot @ 2024-05-15 8:06 UTC (permalink / raw)
To: Frederic Weisbecker
Cc: oe-lkp, lkp, linux-kernel, Thomas Gleixner, Anna-Maria Behnsen,
ying.huang, feng.tang, fengwei.yin, oliver.sang
Hello,
kernel test robot noticed a 2.7% improvement of netperf.Throughput_Mbps on:
commit: 61f7fdf8fd00ce33d30ca3fae8d643c0850ce945 ("timers/migration: Fix ignored event due to missing CPU update")
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master
testcase: netperf
test machine: 128 threads 2 sockets Intel(R) Xeon(R) Gold 6338 CPU @ 2.00GHz (Ice Lake) with 256G memory
parameters:
ip: ipv4
runtime: 300s
nr_threads: 50%
cluster: cs-localhost
test: SCTP_STREAM
cpufreq_governor: performance
In addition to that, the commit also has significant impact on the following tests:
+------------------+--------------------------------------------------------------------------------------------+
| testcase: change | netperf: netperf.Throughput_Mbps 1.1% improvement |
| test machine | 128 threads 2 sockets Intel(R) Xeon(R) Gold 6338 CPU @ 2.00GHz (Ice Lake) with 256G memory |
| test parameters | cluster=cs-localhost |
| | cpufreq_governor=performance |
| | ip=ipv4 |
| | nr_threads=200% |
| | runtime=300s |
| | test=SCTP_STREAM |
+------------------+--------------------------------------------------------------------------------------------+
Details are as below:
-------------------------------------------------------------------------------------------------->
The kernel config and materials to reproduce are available at:
https://download.01.org/0day-ci/archive/20240515/202405151553.6423572f-oliver.sang@intel.com
=========================================================================================
cluster/compiler/cpufreq_governor/ip/kconfig/nr_threads/rootfs/runtime/tbox_group/test/testcase:
cs-localhost/gcc-13/performance/ipv4/x86_64-rhel-8.3/50%/debian-12-x86_64-20240206.cgz/300s/lkp-icl-2sp2/SCTP_STREAM/netperf
commit:
cffaefd15a ("vdso: Use CONFIG_PAGE_SHIFT in vdso/datapage.h")
61f7fdf8fd ("timers/migration: Fix ignored event due to missing CPU update")
cffaefd15a8f423c 61f7fdf8fd00ce33d30ca3fae8d
---------------- ---------------------------
%stddev %change %stddev
\ | \
205059 ± 27% -57.6% 87009 ±102% numa-meminfo.node1.AnonHugePages
100.03 ± 27% -57.6% 42.42 ±102% numa-vmstat.node1.nr_anon_transparent_hugepages
0.01 +0.0 0.01 mpstat.cpu.all.soft%
0.03 ± 3% -0.0 0.03 ± 2% mpstat.cpu.all.sys%
1319042 +0.9% 1330457 proc-vmstat.numa_local
15112424 +2.3% 15464529 proc-vmstat.pgalloc_normal
15078290 +2.2% 15416220 proc-vmstat.pgfree
3.99 +2.7% 4.10 netperf.ThroughputBoth_Mbps
255.41 +2.7% 262.40 netperf.ThroughputBoth_total_Mbps
3.99 +2.7% 4.10 netperf.Throughput_Mbps
255.41 +2.7% 262.40 netperf.Throughput_total_Mbps
45617 +2.5% 46775 netperf.time.voluntary_context_switches
45152 +2.4% 46223 netperf.workload
3.72 ± 4% +65.0% 6.14 ± 5% perf-stat.i.MPKI
9.45 ± 3% +5.2 14.60 ± 4% perf-stat.i.cache-miss-rate%
771280 ± 3% +47.1% 1134581 ± 3% perf-stat.i.cache-misses
1.25 +2.4% 1.28 perf-stat.i.cpi
547.23 ± 4% -31.1% 376.90 ± 4% perf-stat.i.cycles-between-cache-misses
0.86 -2.2% 0.84 perf-stat.i.ipc
1.07 ± 3% +46.2% 1.56 ± 3% perf-stat.overall.MPKI
10.15 ± 3% +4.7 14.81 ± 3% perf-stat.overall.cache-miss-rate%
785.35 ± 3% -31.3% 539.92 ± 3% perf-stat.overall.cycles-between-cache-misses
769379 ± 3% +47.0% 1131341 ± 3% perf-stat.ps.cache-misses
0.99 ±180% +799.5% 8.94 ± 54% sched_debug.cfs_rq:/.left_deadline.avg
83.53 ±163% +1106.3% 1007 ± 58% sched_debug.cfs_rq:/.left_deadline.max
8.56 ±169% +962.5% 90.94 ± 56% sched_debug.cfs_rq:/.left_deadline.stddev
0.99 ±180% +792.4% 8.85 ± 55% sched_debug.cfs_rq:/.left_vruntime.avg
83.35 ±163% +1095.6% 996.57 ± 58% sched_debug.cfs_rq:/.left_vruntime.max
8.54 ±169% +953.5% 89.97 ± 56% sched_debug.cfs_rq:/.left_vruntime.stddev
0.99 ±180% +795.3% 8.88 ± 55% sched_debug.cfs_rq:/.right_vruntime.avg
83.35 ±163% +1095.6% 996.57 ± 58% sched_debug.cfs_rq:/.right_vruntime.max
8.54 ±169% +954.2% 90.03 ± 56% sched_debug.cfs_rq:/.right_vruntime.stddev
4.83 ± 8% +24.3% 6.01 ± 9% sched_debug.cpu.nr_uninterruptible.stddev
0.82 ± 12% -0.4 0.43 ± 71% perf-profile.calltrace.cycles-pp.__do_sys_clone.do_syscall_64.entry_SYSCALL_64_after_hwframe._Fork
0.82 ± 12% -0.4 0.43 ± 71% perf-profile.calltrace.cycles-pp.do_syscall_64.entry_SYSCALL_64_after_hwframe._Fork
0.82 ± 12% -0.4 0.43 ± 71% perf-profile.calltrace.cycles-pp.entry_SYSCALL_64_after_hwframe._Fork
0.82 ± 12% -0.4 0.43 ± 71% perf-profile.calltrace.cycles-pp.kernel_clone.__do_sys_clone.do_syscall_64.entry_SYSCALL_64_after_hwframe._Fork
0.86 ± 17% -0.3 0.60 ± 46% perf-profile.calltrace.cycles-pp.__kmalloc_large_node.__kmalloc_node_track_caller.kmalloc_reserve.__alloc_skb._sctp_make_chunk
0.98 ± 15% -0.3 0.73 ± 14% perf-profile.calltrace.cycles-pp._Fork
0.92 ± 23% +0.4 1.33 ± 15% perf-profile.calltrace.cycles-pp.__memcpy.skb_copy_bits.skb_copy.sctp_make_reassembled_event.sctp_ulpq_partial_delivery
0.92 ± 23% +0.4 1.33 ± 15% perf-profile.calltrace.cycles-pp.skb_copy_bits.skb_copy.sctp_make_reassembled_event.sctp_ulpq_partial_delivery.sctp_cmd_interpreter
13.20 ± 7% +2.1 15.35 ± 3% perf-profile.calltrace.cycles-pp.irq_exit_rcu.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt.acpi_safe_halt.acpi_idle_enter
13.12 ± 7% +2.2 15.28 ± 4% perf-profile.calltrace.cycles-pp.__do_softirq.irq_exit_rcu.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt.acpi_safe_halt
8.72 ± 14% +2.5 11.20 ± 8% perf-profile.calltrace.cycles-pp.__napi_poll.net_rx_action.__do_softirq.irq_exit_rcu.sysvec_apic_timer_interrupt
8.71 ± 14% +2.5 11.20 ± 8% perf-profile.calltrace.cycles-pp.process_backlog.__napi_poll.net_rx_action.__do_softirq.irq_exit_rcu
8.72 ± 14% +2.5 11.21 ± 8% perf-profile.calltrace.cycles-pp.net_rx_action.__do_softirq.irq_exit_rcu.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt
0.82 ± 12% -0.3 0.56 ± 22% perf-profile.children.cycles-pp.__do_sys_clone
1.00 ± 13% -0.2 0.75 ± 16% perf-profile.children.cycles-pp._Fork
0.69 ± 17% -0.2 0.45 ± 22% perf-profile.children.cycles-pp.mas_store_prealloc
0.54 ± 8% -0.2 0.30 ± 45% perf-profile.children.cycles-pp.mas_wr_store_entry
0.45 ± 15% -0.2 0.24 ± 49% perf-profile.children.cycles-pp.dup_mmap
0.40 ± 28% -0.2 0.23 ± 38% perf-profile.children.cycles-pp.create_elf_tables
0.29 ± 31% -0.2 0.13 ± 58% perf-profile.children.cycles-pp.folio_batch_move_lru
0.23 ± 27% -0.1 0.08 ± 70% perf-profile.children.cycles-pp.lru_add_drain_cpu
0.22 ± 33% -0.1 0.07 ± 91% perf-profile.children.cycles-pp.lru_add_drain
0.22 ± 27% -0.1 0.10 ± 58% perf-profile.children.cycles-pp.aa_sk_perm
0.23 ± 38% +0.2 0.40 ± 20% perf-profile.children.cycles-pp.quiet_vmstat
0.20 ± 45% +0.2 0.40 ± 20% perf-profile.children.cycles-pp.need_update
16.11 ± 6% +2.0 18.08 ± 5% perf-profile.children.cycles-pp.irq_exit_rcu
0.18 ± 41% -0.1 0.05 ± 44% perf-profile.self.cycles-pp.entry_SYSCALL_64
0.03 ±100% +0.1 0.12 ± 6% perf-profile.self.cycles-pp.perf_iterate_sb
0.17 ± 63% +0.2 0.36 ± 25% perf-profile.self.cycles-pp.__schedule
0.29 ± 42% +0.2 0.53 ± 28% perf-profile.self.cycles-pp.__free_pages_ok
***************************************************************************************************
lkp-icl-2sp2: 128 threads 2 sockets Intel(R) Xeon(R) Gold 6338 CPU @ 2.00GHz (Ice Lake) with 256G memory
=========================================================================================
cluster/compiler/cpufreq_governor/ip/kconfig/nr_threads/rootfs/runtime/tbox_group/test/testcase:
cs-localhost/gcc-13/performance/ipv4/x86_64-rhel-8.3/200%/debian-12-x86_64-20240206.cgz/300s/lkp-icl-2sp2/SCTP_STREAM/netperf
commit:
cffaefd15a ("vdso: Use CONFIG_PAGE_SHIFT in vdso/datapage.h")
61f7fdf8fd ("timers/migration: Fix ignored event due to missing CPU update")
cffaefd15a8f423c 61f7fdf8fd00ce33d30ca3fae8d
---------------- ---------------------------
%stddev %change %stddev
\ | \
0.08 ± 2% -0.0 0.05 ± 6% mpstat.cpu.all.sys%
107327 -1.4% 105800 proc-vmstat.nr_slab_unreclaimable
136.33 ± 8% +47.3% 200.83 ± 6% perf-c2c.DRAM.remote
15.67 ± 14% +210.6% 48.67 ± 10% perf-c2c.HITM.remote
143276 ± 5% -23.3% 109924 ± 6% sched_debug.cpu.avg_idle.stddev
3791 ± 9% -61.3% 1467 ± 40% sched_debug.cpu.nr_switches.min
4.05 +1.1% 4.10 netperf.ThroughputBoth_Mbps
1037 +1.1% 1049 netperf.ThroughputBoth_total_Mbps
4.05 +1.1% 4.10 netperf.Throughput_Mbps
1037 +1.1% 1049 netperf.Throughput_total_Mbps
1000 ± 3% -79.5% 205.17 ± 25% netperf.time.involuntary_context_switches
10.06 -58.5% 4.18 ± 14% netperf.time.system_time
18.49 +13.8% 21.04 perf-stat.i.MPKI
1.736e+08 -1.4% 1.711e+08 perf-stat.i.branch-instructions
21.87 +3.0 24.84 perf-stat.i.cache-miss-rate%
4862812 +10.9% 5394912 perf-stat.i.cache-misses
22488461 -2.3% 21975605 perf-stat.i.cache-references
9.721e+08 -2.5% 9.482e+08 perf-stat.i.cpu-cycles
464.99 ± 2% -44.1% 260.00 ± 5% perf-stat.i.cpu-migrations
181.26 ± 2% -14.3% 155.42 perf-stat.i.cycles-between-cache-misses
8.602e+08 -1.5% 8.47e+08 perf-stat.i.instructions
5.64 +12.7% 6.36 perf-stat.overall.MPKI
21.61 +2.9 24.54 perf-stat.overall.cache-miss-rate%
200.48 -12.1% 176.18 perf-stat.overall.cycles-between-cache-misses
1423522 -2.7% 1385059 perf-stat.overall.path-length
1.734e+08 -1.4% 1.709e+08 perf-stat.ps.branch-instructions
4847041 +11.0% 5377964 perf-stat.ps.cache-misses
22424436 -2.3% 21912803 perf-stat.ps.cache-references
9.717e+08 -2.5% 9.475e+08 perf-stat.ps.cpu-cycles
463.38 ± 2% -44.1% 259.19 ± 5% perf-stat.ps.cpu-migrations
8.594e+08 -1.5% 8.462e+08 perf-stat.ps.instructions
2.611e+11 -1.9% 2.561e+11 perf-stat.total.instructions
12.30 ± 4% -7.1 5.15 ± 13% perf-profile.calltrace.cycles-pp.kthread.ret_from_fork.ret_from_fork_asm
12.30 ± 4% -7.1 5.15 ± 13% perf-profile.calltrace.cycles-pp.ret_from_fork.ret_from_fork_asm
12.30 ± 4% -7.1 5.15 ± 13% perf-profile.calltrace.cycles-pp.ret_from_fork_asm
10.50 ± 7% -6.9 3.63 ± 15% perf-profile.calltrace.cycles-pp.smpboot_thread_fn.kthread.ret_from_fork.ret_from_fork_asm
9.62 ± 7% -6.8 2.79 ± 20% perf-profile.calltrace.cycles-pp.run_ksoftirqd.smpboot_thread_fn.kthread.ret_from_fork.ret_from_fork_asm
9.62 ± 7% -6.8 2.79 ± 20% perf-profile.calltrace.cycles-pp.__do_softirq.run_ksoftirqd.smpboot_thread_fn.kthread.ret_from_fork
9.56 ± 7% -6.8 2.76 ± 20% perf-profile.calltrace.cycles-pp.net_rx_action.__do_softirq.run_ksoftirqd.smpboot_thread_fn.kthread
9.55 ± 7% -6.8 2.76 ± 20% perf-profile.calltrace.cycles-pp.process_backlog.__napi_poll.net_rx_action.__do_softirq.run_ksoftirqd
9.55 ± 7% -6.8 2.76 ± 20% perf-profile.calltrace.cycles-pp.__napi_poll.net_rx_action.__do_softirq.run_ksoftirqd.smpboot_thread_fn
3.63 ± 11% -1.2 2.46 ± 9% perf-profile.calltrace.cycles-pp.sctp_skb_recv_datagram.sctp_recvmsg.inet_recvmsg.sock_recvmsg.____sys_recvmsg
2.78 ± 15% -1.1 1.70 ± 14% perf-profile.calltrace.cycles-pp.cpu_startup_entry.rest_init.start_kernel.x86_64_start_reservations.x86_64_start_kernel
2.78 ± 15% -1.1 1.70 ± 14% perf-profile.calltrace.cycles-pp.do_idle.cpu_startup_entry.rest_init.start_kernel.x86_64_start_reservations
2.78 ± 15% -1.1 1.70 ± 14% perf-profile.calltrace.cycles-pp.rest_init.start_kernel.x86_64_start_reservations.x86_64_start_kernel.common_startup_64
2.78 ± 15% -1.1 1.70 ± 14% perf-profile.calltrace.cycles-pp.start_kernel.x86_64_start_reservations.x86_64_start_kernel.common_startup_64
2.78 ± 15% -1.1 1.70 ± 14% perf-profile.calltrace.cycles-pp.x86_64_start_kernel.common_startup_64
2.78 ± 15% -1.1 1.70 ± 14% perf-profile.calltrace.cycles-pp.x86_64_start_reservations.x86_64_start_kernel.common_startup_64
2.74 ± 15% -1.1 1.66 ± 15% perf-profile.calltrace.cycles-pp.cpuidle_idle_call.do_idle.cpu_startup_entry.rest_init.start_kernel
2.69 ± 15% -1.1 1.64 ± 15% perf-profile.calltrace.cycles-pp.cpuidle_enter.cpuidle_idle_call.do_idle.cpu_startup_entry.rest_init
6.25 ± 2% -1.0 5.28 ± 5% perf-profile.calltrace.cycles-pp.skb_copy.sctp_make_reassembled_event.sctp_ulpq_tail_data.sctp_cmd_interpreter.sctp_do_sm
6.40 ± 2% -1.0 5.42 ± 5% perf-profile.calltrace.cycles-pp.sctp_make_reassembled_event.sctp_ulpq_tail_data.sctp_cmd_interpreter.sctp_do_sm.sctp_assoc_bh_rcv
2.19 ± 14% -0.9 1.27 ± 25% perf-profile.calltrace.cycles-pp._nohz_idle_balance.__do_softirq.irq_exit_rcu.sysvec_call_function_single.asm_sysvec_call_function_single
2.19 ± 14% -0.9 1.28 ± 25% perf-profile.calltrace.cycles-pp.__do_softirq.irq_exit_rcu.sysvec_call_function_single.asm_sysvec_call_function_single.acpi_safe_halt
2.19 ± 14% -0.9 1.28 ± 25% perf-profile.calltrace.cycles-pp.irq_exit_rcu.sysvec_call_function_single.asm_sysvec_call_function_single.acpi_safe_halt.acpi_idle_enter
6.96 ± 3% -0.9 6.05 ± 5% perf-profile.calltrace.cycles-pp.sctp_ulpq_tail_data.sctp_cmd_interpreter.sctp_do_sm.sctp_assoc_bh_rcv.sctp_rcv
4.27 ± 6% -0.8 3.43 ± 13% perf-profile.calltrace.cycles-pp.sysvec_call_function_single.asm_sysvec_call_function_single.acpi_safe_halt.acpi_idle_enter.cpuidle_enter_state
5.07 ± 4% -0.7 4.36 ± 3% perf-profile.calltrace.cycles-pp.skb_copy_bits.skb_copy.sctp_make_reassembled_event.sctp_ulpq_tail_data.sctp_cmd_interpreter
5.05 ± 4% -0.7 4.36 ± 3% perf-profile.calltrace.cycles-pp.__memcpy.skb_copy_bits.skb_copy.sctp_make_reassembled_event.sctp_ulpq_tail_data
3.52 ± 8% -0.6 2.94 ± 7% perf-profile.calltrace.cycles-pp.sctp_packet_transmit_chunk.sctp_outq_flush_data.sctp_outq_flush.sctp_cmd_interpreter.sctp_do_sm
3.48 ± 9% -0.6 2.92 ± 8% perf-profile.calltrace.cycles-pp.sctp_packet_transmit.sctp_packet_transmit_chunk.sctp_outq_flush_data.sctp_outq_flush.sctp_cmd_interpreter
1.59 ± 7% -0.6 1.04 ± 11% perf-profile.calltrace.cycles-pp.native_queued_spin_lock_slowpath._raw_spin_lock_bh.lock_sock_nested.sctp_skb_recv_datagram.sctp_recvmsg
1.63 ± 8% -0.5 1.11 ± 13% perf-profile.calltrace.cycles-pp._raw_spin_lock_bh.lock_sock_nested.sctp_skb_recv_datagram.sctp_recvmsg.inet_recvmsg
1.65 ± 8% -0.5 1.12 ± 13% perf-profile.calltrace.cycles-pp.lock_sock_nested.sctp_skb_recv_datagram.sctp_recvmsg.inet_recvmsg.sock_recvmsg
1.50 ± 12% -0.4 1.05 ± 11% perf-profile.calltrace.cycles-pp.schedule_timeout.sctp_skb_recv_datagram.sctp_recvmsg.inet_recvmsg.sock_recvmsg
1.47 ± 11% -0.4 1.03 ± 11% perf-profile.calltrace.cycles-pp.schedule.schedule_timeout.sctp_skb_recv_datagram.sctp_recvmsg.inet_recvmsg
1.46 ± 12% -0.4 1.03 ± 12% perf-profile.calltrace.cycles-pp.__schedule.schedule.schedule_timeout.sctp_skb_recv_datagram.sctp_recvmsg
1.79 ± 14% -0.4 1.40 ± 12% perf-profile.calltrace.cycles-pp.consume_skb.sctp_chunk_put.sctp_datamsg_put.sctp_chunk_free.sctp_outq_sack
1.01 ± 19% -0.4 0.62 ± 10% perf-profile.calltrace.cycles-pp.__free_pages_ok.skb_release_data.consume_skb.sctp_chunk_put.sctp_ulpevent_free
1.30 ± 18% -0.4 0.92 ± 23% perf-profile.calltrace.cycles-pp.get_page_from_freelist.__alloc_pages.__kmalloc_large_node.__kmalloc_node_track_caller.kmalloc_reserve
3.96 ± 6% -0.4 3.58 ± 6% perf-profile.calltrace.cycles-pp.sctp_packet_transmit.sctp_outq_flush.sctp_cmd_interpreter.sctp_do_sm.sctp_primitive_SEND
0.95 ± 8% -0.2 0.74 ± 10% perf-profile.calltrace.cycles-pp.skb_release_data.kfree_skb_reason.sctp_recvmsg.inet_recvmsg.sock_recvmsg
1.16 ± 5% -0.2 1.00 ± 10% perf-profile.calltrace.cycles-pp.kfree_skb_reason.sctp_recvmsg.inet_recvmsg.sock_recvmsg.____sys_recvmsg
0.66 ± 10% +0.2 0.86 ± 18% perf-profile.calltrace.cycles-pp.exit_mmap.__mmput.exit_mm.do_exit.do_group_exit
0.67 ± 10% +0.2 0.88 ± 18% perf-profile.calltrace.cycles-pp.__mmput.exit_mm.do_exit.do_group_exit.__x64_sys_exit_group
0.67 ± 10% +0.2 0.88 ± 20% perf-profile.calltrace.cycles-pp.exit_mm.do_exit.do_group_exit.__x64_sys_exit_group.do_syscall_64
0.74 ± 16% +0.4 1.11 ± 13% perf-profile.calltrace.cycles-pp.sctp_generate_timeout_event.call_timer_fn.__run_timers.timer_expire_remote.tmigr_handle_remote_cpu
0.55 ± 47% +0.4 0.96 ± 13% perf-profile.calltrace.cycles-pp.sctp_cmd_interpreter.sctp_do_sm.sctp_generate_timeout_event.call_timer_fn.__run_timers
0.85 ± 15% +0.4 1.27 ± 13% perf-profile.calltrace.cycles-pp.__run_timers.timer_expire_remote.tmigr_handle_remote_cpu.tmigr_handle_remote_up.tmigr_handle_remote
0.85 ± 16% +0.4 1.28 ± 13% perf-profile.calltrace.cycles-pp.timer_expire_remote.tmigr_handle_remote_cpu.tmigr_handle_remote_up.tmigr_handle_remote.__do_softirq
1.49 ± 11% +0.5 1.98 ± 11% perf-profile.calltrace.cycles-pp.tmigr_handle_remote_up.tmigr_handle_remote.__do_softirq.irq_exit_rcu.sysvec_apic_timer_interrupt
1.51 ± 11% +0.5 2.01 ± 12% perf-profile.calltrace.cycles-pp.tmigr_handle_remote.__do_softirq.irq_exit_rcu.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt
1.27 ± 10% +0.5 1.77 ± 11% perf-profile.calltrace.cycles-pp.tmigr_handle_remote_cpu.tmigr_handle_remote_up.tmigr_handle_remote.__do_softirq.irq_exit_rcu
0.00 +0.6 0.63 ± 13% perf-profile.calltrace.cycles-pp._raw_spin_lock.raw_spin_rq_lock_nested._nohz_idle_balance.do_idle.cpu_startup_entry
0.00 +0.6 0.63 ± 14% perf-profile.calltrace.cycles-pp.raw_spin_rq_lock_nested._nohz_idle_balance.do_idle.cpu_startup_entry.start_secondary
0.00 +0.8 0.75 ± 17% perf-profile.calltrace.cycles-pp.update_blocked_averages._nohz_idle_balance.do_idle.cpu_startup_entry.start_secondary
8.46 ± 4% +0.9 9.41 ± 3% perf-profile.calltrace.cycles-pp._copy_to_iter.__skb_datagram_iter.skb_copy_datagram_iter.sctp_recvmsg.inet_recvmsg
8.71 ± 4% +1.0 9.67 ± 2% perf-profile.calltrace.cycles-pp.skb_copy_datagram_iter.sctp_recvmsg.inet_recvmsg.sock_recvmsg.____sys_recvmsg
8.70 ± 4% +1.0 9.67 ± 2% perf-profile.calltrace.cycles-pp.__skb_datagram_iter.skb_copy_datagram_iter.sctp_recvmsg.inet_recvmsg.sock_recvmsg
0.80 ± 21% +1.2 2.02 ± 4% perf-profile.calltrace.cycles-pp._nohz_idle_balance.do_idle.cpu_startup_entry.start_secondary.common_startup_64
12.15 ± 3% +6.4 18.54 ± 5% perf-profile.calltrace.cycles-pp.__napi_poll.net_rx_action.__do_softirq.irq_exit_rcu.sysvec_apic_timer_interrupt
12.15 ± 3% +6.4 18.55 ± 5% perf-profile.calltrace.cycles-pp.net_rx_action.__do_softirq.irq_exit_rcu.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt
12.14 ± 3% +6.4 18.54 ± 5% perf-profile.calltrace.cycles-pp.process_backlog.__napi_poll.net_rx_action.__do_softirq.irq_exit_rcu
21.77 +6.5 28.25 ± 3% perf-profile.calltrace.cycles-pp.acpi_safe_halt.acpi_idle_enter.cpuidle_enter_state.cpuidle_enter.cpuidle_idle_call
14.95 ± 3% +6.8 21.71 ± 4% perf-profile.calltrace.cycles-pp.__do_softirq.irq_exit_rcu.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt.acpi_safe_halt
14.98 ± 3% +6.8 21.74 ± 4% perf-profile.calltrace.cycles-pp.irq_exit_rcu.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt.acpi_safe_halt.acpi_idle_enter
26.76 ± 2% +6.8 33.55 ± 2% perf-profile.calltrace.cycles-pp.cpuidle_enter_state.cpuidle_enter.cpuidle_idle_call.do_idle.cpu_startup_entry
26.47 ± 2% +6.8 33.31 ± 2% perf-profile.calltrace.cycles-pp.acpi_idle_enter.cpuidle_enter_state.cpuidle_enter.cpuidle_idle_call.do_idle
16.26 ± 4% +7.1 23.40 ± 4% perf-profile.calltrace.cycles-pp.sysvec_apic_timer_interrupt.asm_sysvec_apic_timer_interrupt.acpi_safe_halt.acpi_idle_enter.cpuidle_enter_state
19.50 ± 3% +7.6 27.06 ± 3% perf-profile.calltrace.cycles-pp.asm_sysvec_apic_timer_interrupt.acpi_safe_halt.acpi_idle_enter.cpuidle_enter_state.cpuidle_enter
24.14 ± 2% +7.8 31.96 ± 2% perf-profile.calltrace.cycles-pp.cpuidle_enter.cpuidle_idle_call.do_idle.cpu_startup_entry.start_secondary
26.39 ± 2% +7.9 34.26 ± 2% perf-profile.calltrace.cycles-pp.cpuidle_idle_call.do_idle.cpu_startup_entry.start_secondary.common_startup_64
31.94 ± 2% +8.0 39.97 perf-profile.calltrace.cycles-pp.common_startup_64
29.16 ± 2% +9.1 38.26 ± 2% perf-profile.calltrace.cycles-pp.cpu_startup_entry.start_secondary.common_startup_64
29.13 ± 2% +9.1 38.24 ± 2% perf-profile.calltrace.cycles-pp.do_idle.cpu_startup_entry.start_secondary.common_startup_64
29.16 ± 2% +9.1 38.27 ± 2% perf-profile.calltrace.cycles-pp.start_secondary.common_startup_64
12.40 ± 4% -7.2 5.22 ± 12% perf-profile.children.cycles-pp.ret_from_fork_asm
12.30 ± 4% -7.1 5.15 ± 13% perf-profile.children.cycles-pp.kthread
12.36 ± 4% -7.1 5.21 ± 12% perf-profile.children.cycles-pp.ret_from_fork
10.50 ± 7% -6.9 3.63 ± 15% perf-profile.children.cycles-pp.smpboot_thread_fn
9.62 ± 7% -6.8 2.79 ± 20% perf-profile.children.cycles-pp.run_ksoftirqd
31.94 -1.5 30.44 perf-profile.children.cycles-pp.__do_softirq
5.16 ± 10% -1.4 3.72 ± 11% perf-profile.children.cycles-pp.native_queued_spin_lock_slowpath
3.68 ± 11% -1.3 2.36 ± 9% perf-profile.children.cycles-pp._raw_spin_lock_irqsave
3.33 ± 8% -1.2 2.12 ± 15% perf-profile.children.cycles-pp.load_balance
17.33 ± 2% -1.2 16.12 ± 3% perf-profile.children.cycles-pp.__memcpy
2.88 ± 9% -1.2 1.70 ± 8% perf-profile.children.cycles-pp.find_busiest_group
3.63 ± 11% -1.2 2.46 ± 9% perf-profile.children.cycles-pp.sctp_skb_recv_datagram
2.85 ± 9% -1.2 1.68 ± 8% perf-profile.children.cycles-pp.update_sd_lb_stats
2.78 ± 15% -1.1 1.70 ± 14% perf-profile.children.cycles-pp.rest_init
2.78 ± 15% -1.1 1.70 ± 14% perf-profile.children.cycles-pp.start_kernel
2.78 ± 15% -1.1 1.70 ± 14% perf-profile.children.cycles-pp.x86_64_start_kernel
2.78 ± 15% -1.1 1.70 ± 14% perf-profile.children.cycles-pp.x86_64_start_reservations
2.61 ± 10% -1.1 1.53 ± 8% perf-profile.children.cycles-pp.update_sg_lb_stats
7.41 ± 2% -1.0 6.42 ± 4% perf-profile.children.cycles-pp.sctp_ulpq_tail_data
16.46 ± 2% -1.0 15.46 ± 3% perf-profile.children.cycles-pp.sctp_packet_transmit
3.51 ± 5% -0.9 2.58 ± 9% perf-profile.children.cycles-pp.kmalloc_reserve
3.42 ± 5% -0.9 2.49 ± 10% perf-profile.children.cycles-pp.__kmalloc_node_track_caller
50.46 -0.9 49.53 perf-profile.children.cycles-pp.do_syscall_64
50.52 -0.9 49.60 perf-profile.children.cycles-pp.entry_SYSCALL_64_after_hwframe
2.19 ± 5% -0.9 1.27 ± 17% perf-profile.children.cycles-pp.rmqueue
3.37 ± 5% -0.9 2.46 ± 10% perf-profile.children.cycles-pp.__kmalloc_large_node
3.96 ± 3% -0.9 3.06 ± 8% perf-profile.children.cycles-pp.__alloc_skb
4.38 ± 6% -0.9 3.48 ± 4% perf-profile.children.cycles-pp.__schedule
3.21 ± 4% -0.9 2.34 ± 13% perf-profile.children.cycles-pp.__alloc_pages
3.04 ± 4% -0.9 2.17 ± 14% perf-profile.children.cycles-pp.get_page_from_freelist
4.32 ± 6% -0.9 3.46 ± 13% perf-profile.children.cycles-pp.sysvec_call_function_single
2.59 ± 17% -0.8 1.77 ± 5% perf-profile.children.cycles-pp.__free_pages_ok
9.60 ± 2% -0.8 8.78 ± 4% perf-profile.children.cycles-pp.sctp_make_reassembled_event
9.34 ± 2% -0.8 8.54 ± 4% perf-profile.children.cycles-pp.skb_copy
1.69 ± 20% -0.8 0.90 ± 14% perf-profile.children.cycles-pp.pick_next_task_fair
3.32 ± 9% -0.8 2.54 ± 5% perf-profile.children.cycles-pp.schedule
2.05 ± 15% -0.8 1.28 ± 10% perf-profile.children.cycles-pp.newidle_balance
1.47 ± 21% -0.8 0.71 ± 10% perf-profile.children.cycles-pp.free_one_page
3.71 ± 14% -0.6 3.07 ± 6% perf-profile.children.cycles-pp.skb_release_data
2.37 ± 9% -0.6 1.73 ± 6% perf-profile.children.cycles-pp.schedule_timeout
2.04 ± 9% -0.5 1.50 ± 12% perf-profile.children.cycles-pp.lock_sock_nested
2.06 ± 8% -0.5 1.52 ± 13% perf-profile.children.cycles-pp._raw_spin_lock_bh
0.50 ± 47% -0.2 0.26 ± 18% perf-profile.children.cycles-pp.cpu_util
0.27 ± 23% -0.2 0.10 ± 40% perf-profile.children.cycles-pp.task_work_run
0.23 ± 19% -0.2 0.06 ± 52% perf-profile.children.cycles-pp.task_mm_cid_work
0.26 ± 24% -0.1 0.14 ± 17% perf-profile.children.cycles-pp._find_next_and_bit
0.16 ± 27% -0.1 0.08 ± 35% perf-profile.children.cycles-pp.rw_verify_area
0.15 ± 9% +0.0 0.18 ± 14% perf-profile.children.cycles-pp.__cmd_record
0.15 ± 9% +0.0 0.18 ± 14% perf-profile.children.cycles-pp.cmd_record
0.05 ± 75% +0.1 0.14 ± 19% perf-profile.children.cycles-pp.call_cpuidle
0.31 ± 16% +0.1 0.41 ± 13% perf-profile.children.cycles-pp.write
0.12 ± 35% +0.1 0.24 ± 34% perf-profile.children.cycles-pp.tick_do_update_jiffies64
0.08 ± 19% +0.1 0.19 ± 21% perf-profile.children.cycles-pp.sctp_backlog_rcv
0.22 ± 33% +0.1 0.35 ± 13% perf-profile.children.cycles-pp.perf_read
0.08 ± 19% +0.1 0.22 ± 25% perf-profile.children.cycles-pp.__release_sock
0.14 ± 19% +0.2 0.30 ± 14% perf-profile.children.cycles-pp.release_sock
0.31 ± 11% +0.2 0.48 ± 37% perf-profile.children.cycles-pp.zap_pte_range
0.34 ± 19% +0.2 0.52 ± 32% perf-profile.children.cycles-pp.unmap_page_range
0.31 ± 14% +0.2 0.50 ± 35% perf-profile.children.cycles-pp.zap_pmd_range
0.44 ± 15% +0.2 0.64 ± 15% perf-profile.children.cycles-pp.__rhashtable_lookup
0.68 ± 9% +0.2 0.90 ± 19% perf-profile.children.cycles-pp.exit_mm
0.86 ± 4% +0.2 1.10 ± 18% perf-profile.children.cycles-pp.exit_mmap
0.86 ± 4% +0.3 1.12 ± 17% perf-profile.children.cycles-pp.__mmput
1.48 ± 11% +0.3 1.77 ± 9% perf-profile.children.cycles-pp.__run_timers
0.64 ± 15% +0.3 0.99 ± 34% perf-profile.children.cycles-pp.__sctp_rcv_lookup
0.63 ± 15% +0.4 0.98 ± 34% perf-profile.children.cycles-pp.sctp_addrs_lookup_transport
0.88 ± 15% +0.4 1.32 ± 13% perf-profile.children.cycles-pp.timer_expire_remote
1.53 ± 10% +0.5 2.03 ± 11% perf-profile.children.cycles-pp.tmigr_handle_remote_up
1.55 ± 10% +0.5 2.07 ± 11% perf-profile.children.cycles-pp.tmigr_handle_remote
1.31 ± 9% +0.5 1.83 ± 11% perf-profile.children.cycles-pp.tmigr_handle_remote_cpu
0.50 ± 14% +0.5 1.03 ± 9% perf-profile.children.cycles-pp.raw_spin_rq_lock_nested
8.71 ± 4% +1.0 9.67 ± 2% perf-profile.children.cycles-pp.skb_copy_datagram_iter
8.52 ± 4% +1.0 9.48 ± 2% perf-profile.children.cycles-pp._copy_to_iter
8.70 ± 4% +1.0 9.67 ± 2% perf-profile.children.cycles-pp.__skb_datagram_iter
1.79 ± 12% +1.0 2.80 ± 4% perf-profile.children.cycles-pp._raw_spin_lock
17.97 ± 3% +5.6 23.57 ± 4% perf-profile.children.cycles-pp.irq_exit_rcu
26.84 ± 2% +6.8 33.60 ± 2% perf-profile.children.cycles-pp.cpuidle_enter
29.13 ± 2% +6.8 35.92 ± 2% perf-profile.children.cycles-pp.cpuidle_idle_call
26.77 ± 2% +6.8 33.56 ± 2% perf-profile.children.cycles-pp.cpuidle_enter_state
26.45 ± 2% +6.8 33.28 ± 2% perf-profile.children.cycles-pp.acpi_safe_halt
17.04 ± 3% +6.8 23.88 ± 4% perf-profile.children.cycles-pp.sysvec_apic_timer_interrupt
26.47 ± 2% +6.8 33.31 ± 2% perf-profile.children.cycles-pp.acpi_idle_enter
18.83 ± 3% +7.0 25.78 ± 3% perf-profile.children.cycles-pp.asm_sysvec_apic_timer_interrupt
31.93 ± 2% +8.0 39.96 perf-profile.children.cycles-pp.do_idle
31.94 ± 2% +8.0 39.97 perf-profile.children.cycles-pp.common_startup_64
31.94 ± 2% +8.0 39.97 perf-profile.children.cycles-pp.cpu_startup_entry
29.16 ± 2% +9.1 38.27 ± 2% perf-profile.children.cycles-pp.start_secondary
5.14 ± 10% -1.4 3.71 ± 11% perf-profile.self.cycles-pp.native_queued_spin_lock_slowpath
16.85 ± 2% -0.9 15.98 ± 2% perf-profile.self.cycles-pp.__memcpy
1.83 ± 10% -0.7 1.09 ± 9% perf-profile.self.cycles-pp.update_sg_lb_stats
0.20 ± 24% -0.2 0.05 ± 74% perf-profile.self.cycles-pp.task_mm_cid_work
0.35 ± 25% -0.1 0.21 ± 16% perf-profile.self.cycles-pp.cpu_util
0.23 ± 24% -0.1 0.13 ± 21% perf-profile.self.cycles-pp._find_next_and_bit
0.02 ±141% +0.1 0.08 ± 22% perf-profile.self.cycles-pp.folio_remove_rmap_ptes
0.02 ±141% +0.1 0.08 ± 38% perf-profile.self.cycles-pp.__run_timers
0.05 ± 75% +0.1 0.13 ± 18% perf-profile.self.cycles-pp.call_cpuidle
0.03 ±150% +0.1 0.12 ± 35% perf-profile.self.cycles-pp.update_blocked_averages
0.18 ± 33% +0.1 0.32 ± 22% perf-profile.self.cycles-pp.sctp_packet_config
1.29 ± 11% +0.4 1.66 ± 7% perf-profile.self.cycles-pp._raw_spin_lock
8.42 ± 4% +1.0 9.45 ± 2% perf-profile.self.cycles-pp._copy_to_iter
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.
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [linus:master] [timers/migration] 61f7fdf8fd: netperf.Throughput_Mbps 2.7% improvement
2024-05-15 8:06 [linus:master] [timers/migration] 61f7fdf8fd: netperf.Throughput_Mbps 2.7% improvement kernel test robot
@ 2024-05-15 14:14 ` Frederic Weisbecker
2024-05-16 2:42 ` Oliver Sang
0 siblings, 1 reply; 3+ messages in thread
From: Frederic Weisbecker @ 2024-05-15 14:14 UTC (permalink / raw)
To: kernel test robot
Cc: oe-lkp, lkp, linux-kernel, Thomas Gleixner, Anna-Maria Behnsen,
ying.huang, feng.tang, fengwei.yin
Le Wed, May 15, 2024 at 04:06:50PM +0800, kernel test robot a écrit :
>
>
> Hello,
>
> kernel test robot noticed a 2.7% improvement of netperf.Throughput_Mbps on:
>
>
> commit: 61f7fdf8fd00ce33d30ca3fae8d643c0850ce945 ("timers/migration: Fix ignored event due to missing CPU update")
> https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master
Nice, so this means the previously reported regression with netperf due to the
new timer migration code is gone, right?
For reference: https://lore.kernel.org/all/202403011511.24defbbd-oliver.sang@intel.com/
Thanks.
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [linus:master] [timers/migration] 61f7fdf8fd: netperf.Throughput_Mbps 2.7% improvement
2024-05-15 14:14 ` Frederic Weisbecker
@ 2024-05-16 2:42 ` Oliver Sang
0 siblings, 0 replies; 3+ messages in thread
From: Oliver Sang @ 2024-05-16 2:42 UTC (permalink / raw)
To: Frederic Weisbecker
Cc: oe-lkp, lkp, linux-kernel, Thomas Gleixner, Anna-Maria Behnsen,
ying.huang, feng.tang, fengwei.yin, oliver.sang
hi, Frederic Weisbecker,
On Wed, May 15, 2024 at 04:14:14PM +0200, Frederic Weisbecker wrote:
> Le Wed, May 15, 2024 at 04:06:50PM +0800, kernel test robot a écrit :
> >
> >
> > Hello,
> >
> > kernel test robot noticed a 2.7% improvement of netperf.Throughput_Mbps on:
> >
> >
> > commit: 61f7fdf8fd00ce33d30ca3fae8d643c0850ce945 ("timers/migration: Fix ignored event due to missing CPU update")
> > https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git master
>
> Nice, so this means the previously reported regression with netperf due to the
> new timer migration code is gone, right?
>
> For reference: https://lore.kernel.org/all/202403011511.24defbbd-oliver.sang@intel.com/
I believe so.
previous report is for 7ee9887703, the scores are:
=========================================================================================
cluster/compiler/cpufreq_governor/ip/kconfig/nr_threads/rootfs/runtime/tbox_group/test/testcase:
cs-localhost/gcc-12/performance/ipv4/x86_64-rhel-8.3/200%/debian-12-x86_64-20240206.cgz/300s/lkp-icl-2sp2/SCTP_STREAM/netperf
<---- (1)
57e95a5c4117dc6a 7ee988770326fca440472200c3e
---------------- ---------------------------
%stddev %change %stddev
\ | \
4.10 -1.2% 4.05 netperf.ThroughputBoth_Mbps
1049 -1.2% 1037 netperf.ThroughputBoth_total_Mbps
4.10 -1.2% 4.05 netperf.Throughput_Mbps
1049 -1.2% 1037 netperf.Throughput_total_Mbps
for this report, we just have same test's score as:
=========================================================================================
cluster/compiler/cpufreq_governor/ip/kconfig/nr_threads/rootfs/runtime/tbox_group/test/testcase:
cs-localhost/gcc-13/performance/ipv4/x86_64-rhel-8.3/200%/debian-12-x86_64-20240206.cgz/300s/lkp-icl-2sp2/SCTP_STREAM/netperf
<---- (2)
cffaefd15a8f423c 61f7fdf8fd00ce33d30ca3fae8d
---------------- ---------------------------
%stddev %change %stddev
\ | \
4.05 +1.1% 4.10 netperf.ThroughputBoth_Mbps
1037 +1.1% 1049 netperf.ThroughputBoth_total_Mbps
4.05 +1.1% 4.10 netperf.Throughput_Mbps
1037 +1.1% 1049 netperf.Throughput_total_Mbps
from option lines (1)(2), only diff is compiler version which we updated to use
gcc-13 recently. but seems no impact in this case.
what we can see from 2 tables above are, 7ee9887703 just has same scores as
parent of 61f7fdf8fd (cffaefd15a), and 61f7fdf8fd just has same scores as parent
of 7ee9887703 (57e95a5c41). these mean the regression is just resolved.
>
> Thanks.
>
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2024-05-16 2:42 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2024-05-15 8:06 [linus:master] [timers/migration] 61f7fdf8fd: netperf.Throughput_Mbps 2.7% improvement kernel test robot
2024-05-15 14:14 ` Frederic Weisbecker
2024-05-16 2:42 ` Oliver Sang
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.