blob: 63f938b887dd135d31cc0e8928f650e23bbd3c0b [file] [log] [blame]
Ingo Molnar0a02ad92009-09-11 12:12:54 +02001perf-sched(1)
Takashi Iwaiea665362018-03-07 11:54:41 +01002=============
Ingo Molnar0a02ad92009-09-11 12:12:54 +02003
4NAME
5----
Ingo Molnarc13f0d32009-09-13 16:51:04 +02006perf-sched - Tool to trace/measure scheduler properties (latencies)
Ingo Molnar0a02ad92009-09-11 12:12:54 +02007
8SYNOPSIS
9--------
10[verse]
David Ahern49394a22016-11-16 15:06:29 +090011'perf sched' {record|latency|map|replay|script|timehist}
Ingo Molnar0a02ad92009-09-11 12:12:54 +020012
13DESCRIPTION
14-----------
David Ahern49394a22016-11-16 15:06:29 +090015There are several variants of 'perf sched':
Ingo Molnarc13f0d32009-09-13 16:51:04 +020016
17 'perf sched record <command>' to record the scheduling events
18 of an arbitrary workload.
19
20 'perf sched latency' to report the per task scheduling latencies
21 and other scheduling properties of the workload.
22
Jiri Olsae78cb362011-09-27 11:16:35 +020023 'perf sched script' to see a detailed trace of the workload that
24 was recorded (aliased to 'perf script' for now).
Ingo Molnarc13f0d32009-09-13 16:51:04 +020025
26 'perf sched replay' to simulate the workload that was recorded
27 via perf sched record. (this is done by starting up mockup threads
28 that mimic the workload based on the events in the trace. These
29 threads can then replay the timings (CPU runtime and sleep patterns)
Randy Dunlap854c5542010-03-31 11:31:00 -070030 of the workload as it occurred when it was recorded - and can repeat
Ingo Molnarc13f0d32009-09-13 16:51:04 +020031 it a number of times, measuring its performance.)
Ingo Molnar0a02ad92009-09-11 12:12:54 +020032
Shawn Bohrer1eacc942010-11-30 19:57:18 -060033 'perf sched map' to print a textual context-switching outline of
34 workload captured via perf sched record. Columns stand for
35 individual CPUs, and the two-letter shortcuts stand for tasks that
36 are running on a CPU. A '*' denotes the CPU that had the event, and
37 a dot signals an idle CPU.
38
David Ahern49394a22016-11-16 15:06:29 +090039 'perf sched timehist' provides an analysis of scheduling events.
40
41 Example usage:
42 perf sched record -- sleep 1
43 perf sched timehist
44
45 By default it shows the individual schedule events, including the wait
46 time (time between sched-out and next sched-in events for the task), the
47 task scheduling delay (time between wakeup and actually running) and run
48 time for the task:
49
50 time cpu task name wait time sch delay run time
51 [tid/pid] (msec) (msec) (msec)
52 -------------- ------ -------------------- --------- --------- ---------
53 79371.874569 [0011] gcc[31949] 0.014 0.000 1.148
54 79371.874591 [0010] gcc[31951] 0.000 0.000 0.024
55 79371.874603 [0010] migration/10[59] 3.350 0.004 0.011
56 79371.874604 [0011] <idle> 1.148 0.000 0.035
57 79371.874723 [0005] <idle> 0.016 0.000 1.383
58 79371.874746 [0005] gcc[31949] 0.153 0.078 0.022
59 ...
60
61 Times are in msec.usec.
62
Ingo Molnar0a02ad92009-09-11 12:12:54 +020063OPTIONS
64-------
Shawn Bohrer1eacc942010-11-30 19:57:18 -060065-i::
66--input=<file>::
Robert Richterefad1412011-12-07 10:02:54 +010067 Input file name. (default: perf.data unless stdin is a fifo)
Shawn Bohrer1eacc942010-11-30 19:57:18 -060068
69-v::
70--verbose::
71 Be more verbose. (show symbol address, etc)
72
Ingo Molnar0a02ad92009-09-11 12:12:54 +020073-D::
74--dump-raw-trace=::
75 Display verbose dump of the sched data.
76
Sihyeon Jange9b61e52017-11-16 23:26:05 +090077-f::
78--force::
79 Don't complain, do it.
80
Jiri Olsa99623c62016-04-12 15:29:26 +020081OPTIONS for 'perf sched map'
82----------------------------
83
84--compact::
85 Show only CPUs with activity. Helps visualizing on high core
86 count systems.
87
Jiri Olsa73643bb2016-04-12 15:29:31 +020088--cpus::
89 Show just entries with activities for the given CPUs.
90
Jiri Olsacf294f22016-04-12 15:29:30 +020091--color-cpus::
92 Highlight the given cpus.
93
Jiri Olsaa151a372016-04-12 15:29:29 +020094--color-pids::
95 Highlight the given pids.
96
David Ahern49394a22016-11-16 15:06:29 +090097OPTIONS for 'perf sched timehist'
98---------------------------------
99-k::
100--vmlinux=<file>::
101 vmlinux pathname
102
103--kallsyms=<file>::
104 kallsyms pathname
105
David Ahern6c973c92016-11-16 15:06:32 +0900106-g::
Takuya Yamamotoe8103e42018-04-10 23:35:39 +0900107--call-graph::
108 Display call chains if present (default on).
David Ahern6c973c92016-11-16 15:06:32 +0900109
110--max-stack::
111 Maximum number of functions to display in backtrace, default 5.
112
David Ahern0f59d7a2017-09-01 10:49:12 -0700113-p=::
114--pid=::
115 Only show events for given process ID (comma separated list).
116
117-t=::
118--tid=::
119 Only show events for given thread ID (comma separated list).
120
David Ahern49394a22016-11-16 15:06:29 +0900121-s::
122--summary::
123 Show only a summary of scheduling by thread with min, max, and average
124 run times (in sec) and relative stddev.
125
126-S::
127--with-summary::
128 Show all scheduling events followed by a summary by thread with min,
129 max, and average run times (in sec) and relative stddev.
130
131--symfs=<directory>::
132 Look for files with symbols relative to this directory.
133
David Aherna407b062016-11-16 15:06:33 +0900134-V::
135--cpu-visual::
136 Show visual aid for sched switches by CPU: 'i' marks idle time,
137 's' are scheduler events.
138
David Ahernfc1469f2016-11-16 15:06:31 +0900139-w::
140--wakeups::
141 Show wakeup events.
142
David Ahern350f54f2016-11-25 09:28:41 -0700143-M::
144--migrations::
145 Show migration events.
146
Brendan Gregg292c4a82017-03-14 01:56:29 +0000147-n::
148--next::
149 Show next task.
150
Namhyung Kim07235f82016-12-08 23:47:54 +0900151-I::
152--idle-hist::
153 Show idle-related events only.
154
David Ahern853b7402016-11-29 10:15:44 -0700155--time::
156 Only analyze samples within given time window: <start>,<stop>. Times
157 have the format seconds.microseconds. If start is not given (i.e., time
158 string is ',x.y') then analysis starts at the beginning of the file. If
159 stop time is not given (i.e, time string is 'x.y,') then analysis goes
160 to end of file.
161
Namhyung Kim414e0502017-01-13 19:45:22 +0900162--state::
163 Show task state when it switched out.
David Ahern853b7402016-11-29 10:15:44 -0700164
Ingo Molnar0a02ad92009-09-11 12:12:54 +0200165SEE ALSO
166--------
167linkperf:perf-record[1]