[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20140314062043.GA28336@localhost>
Date: Fri, 14 Mar 2014 14:20:43 +0800
From: Fengguang Wu <fengguang.wu@...el.com>
To: Alex Shi <alex.shi@...aro.org>
Cc: LKML <linux-kernel@...r.kernel.org>, lkp@...org
Subject: [sched] 373e9ec30ad: -11.2% qperf.udp_bw.send_bw.MB_sec
Alex, we noticed the below changes on
https://github.com/alexshi/power-scheduling.git single-balance
commit 373e9ec30ad1efee2c4ba6b58fc317626e6482d0 ("sched: only do load balance on tick_do_timer_cpu")
test case: lkp-a06/micro/qperf/600s
v3.14-rc6 373e9ec30ad1efee2c4ba6b58
--------------- -------------------------
273 ~ 0% -11.2% 242 ~ 1% TOTAL qperf.udp_bw.send_bw.MB_sec
268 ~ 0% -9.8% 242 ~ 0% TOTAL qperf.tcp_bw.bw.MB_sec
266 ~ 0% -9.8% 240 ~ 1% TOTAL qperf.udp_bw.recv_bw.MB_sec
48281 ~ 2% -81.3% 9035 ~ 1% TOTAL softirqs.SCHED
1359067 ~ 4% -30.7% 942510 ~ 2% TOTAL interrupts.RES
10201776 ~14% -30.1% 7133067 ~10% TOTAL cpuidle.C3.time
52277488 ~ 4% -26.6% 38366608 ~ 6% TOTAL cpuidle.C2.time
28 ~14% -25.3% 21 ~18% TOTAL cpuidle.POLL.usage
1516 ~16% -27.0% 1107 ~22% TOTAL cpuidle.POLL.time
3955 ~ 7% -24.8% 2972 ~14% TOTAL cpuidle.C3.usage
1.10 ~ 9% -17.0% 0.91 ~16% TOTAL perf-profile.cpu-cycles.menu_select.cpuidle_idle_call.arch_cpu_idle.cpu_startup_entry.start_secondary
9066611 ~ 0% -8.3% 8313872 ~ 1% TOTAL proc-vmstat.pgalloc_dma32
24350865 ~ 0% -8.3% 22328921 ~ 1% TOTAL proc-vmstat.pgfree
15285179 ~ 0% -8.3% 14016032 ~ 1% TOTAL proc-vmstat.pgalloc_normal
2961 ~ 3% -19.7% 2378 ~ 3% TOTAL vmstat.system.in
3887389 ~20% +9.7% 4262549 ~ 0% TOTAL time.voluntary_context_switches
14152 ~ 1% +2.2% 14464 ~ 0% TOTAL vmstat.system.cs
Legend:
~XX% - stddev percent
[+-]XX% - change percent
qperf.udp_bw.send_bw.MB_sec
280 ++-------------------------------------------------------------------+
| .*..... |
275 *+...*.....*.... ...*....*.... .. * |
270 ++ *....*.. *. |
| |
265 ++ |
260 ++ |
| |
255 ++ |
250 ++ O |
| |
245 ++ O |
240 O+ O O O O O O O
| O O O |
235 ++----------------------------------------O--------------------------+
[*] bisect-good sample
[O] bisect-bad sample
Thanks,
Fengguang
View attachment "reproduce" of type "text/plain" (90 bytes)
Powered by blists - more mailing lists