1 perf-sched(1) 1 perf-sched(1)
2 ============= !! 2 ==============
3 3
4 NAME 4 NAME
5 ---- 5 ----
6 perf-sched - Tool to trace/measure scheduler p 6 perf-sched - Tool to trace/measure scheduler properties (latencies)
7 7
8 SYNOPSIS 8 SYNOPSIS
9 -------- 9 --------
10 [verse] 10 [verse]
11 'perf sched' {record|latency|map|replay|script !! 11 'perf sched' {record|latency|replay|trace}
12 12
13 DESCRIPTION 13 DESCRIPTION
14 ----------- 14 -----------
15 There are several variants of 'perf sched': !! 15 There's four variants of perf sched:
16 16
17 'perf sched record <command>' to record the 17 'perf sched record <command>' to record the scheduling events
18 of an arbitrary workload. 18 of an arbitrary workload.
19 19
20 'perf sched latency' to report the per task 20 'perf sched latency' to report the per task scheduling latencies
21 and other scheduling properties of the workl 21 and other scheduling properties of the workload.
22 22
23 Example usage: !! 23 'perf sched trace' to see a detailed trace of the workload that
24 perf sched record -- sleep 1 !! 24 was recorded.
25 perf sched latency <<
26 <<
27 -------------------------------------------- <<
28 Task | Runtime ms | Cou <<
29 -------------------------------------------- <<
30 perf:(2) | 2.804 ms | <<
31 NetworkManager:1343 | 0.372 ms | <<
32 kworker/1:2-xfs:4649 | 0.012 ms | <<
33 kworker/3:1-xfs:388 | 0.011 ms | <<
34 sleep:147736 | 0.938 ms | <<
35 <<
36 It shows Runtime(time that a task spent actu <<
37 Count(number of times a delay was calculated <<
38 task was ready to run but was kept waiting). <<
39 <<
40 Tasks with the same command name are merged <<
41 given within (), However if -p option is use <<
42 <<
43 'perf sched script' to see a detailed trace <<
44 was recorded (aliased to 'perf script' for <<
45 25
46 'perf sched replay' to simulate the workload 26 'perf sched replay' to simulate the workload that was recorded
47 via perf sched record. (this is done by star 27 via perf sched record. (this is done by starting up mockup threads
48 that mimic the workload based on the events 28 that mimic the workload based on the events in the trace. These
49 threads can then replay the timings (CPU run 29 threads can then replay the timings (CPU runtime and sleep patterns)
50 of the workload as it occurred when it was r !! 30 of the workload as it occured when it was recorded - and can repeat
51 it a number of times, measuring its performa 31 it a number of times, measuring its performance.)
52 32
53 'perf sched map' to print a textual context- <<
54 workload captured via perf sched record. Co <<
55 individual CPUs, and the two-letter shortcut <<
56 are running on a CPU. A '*' denotes the CPU <<
57 a dot signals an idle CPU. <<
58 <<
59 'perf sched timehist' provides an analysis o <<
60 <<
61 Example usage: <<
62 perf sched record -- sleep 1 <<
63 perf sched timehist <<
64 <<
65 By default it shows the individual schedule <<
66 time (time between sched-out and next sched <<
67 task scheduling delay (time between runnabl <<
68 run time for the task: <<
69 <<
70 time cpu task name <<
71 [tid/pid] <<
72 -------------- ------ ----------------- <<
73 79371.874569 [0011] gcc[31949] <<
74 79371.874591 [0010] gcc[31951] <<
75 79371.874603 [0010] migration/10[59] <<
76 79371.874604 [0011] <idle> <<
77 79371.874723 [0005] <idle> <<
78 79371.874746 [0005] gcc[31949] <<
79 ... <<
80 <<
81 Times are in msec.usec. <<
82 <<
83 OPTIONS 33 OPTIONS
84 ------- 34 -------
85 -i:: <<
86 --input=<file>:: <<
87 Input file name. (default: perf.data u <<
88 <<
89 -v:: <<
90 --verbose:: <<
91 Be more verbose. (show symbol address, <<
92 <<
93 -D:: 35 -D::
94 --dump-raw-trace=:: 36 --dump-raw-trace=::
95 Display verbose dump of the sched data 37 Display verbose dump of the sched data.
96 <<
97 -f:: <<
98 --force:: <<
99 Don't complain, do it. <<
100 <<
101 OPTIONS for 'perf sched latency' <<
102 ------------------------------- <<
103 <<
104 -C:: <<
105 --CPU <n>:: <<
106 CPU to profile on. <<
107 <<
108 -p:: <<
109 --pids:: <<
110 latency stats per pid instead of per c <<
111 <<
112 -s:: <<
113 --sort <key[,key2...]>:: <<
114 sort by key(s): runtime, switch, avg, <<
115 by default it's sorted by "avg ,max ,s <<
116 <<
117 OPTIONS for 'perf sched map' <<
118 ---------------------------- <<
119 <<
120 --compact:: <<
121 Show only CPUs with activity. Helps vi <<
122 count systems. <<
123 <<
124 --cpus:: <<
125 Show just entries with activities for <<
126 <<
127 --color-cpus:: <<
128 Highlight the given cpus. <<
129 <<
130 --color-pids:: <<
131 Highlight the given pids. <<
132 <<
133 --task-name <task>:: <<
134 Map output only for the given task nam <<
135 task names with a comma (without white <<
136 time is printed and is represented by <<
137 task name(s). <<
138 ('-' indicates other tasks while '.' i <<
139 <<
140 --fuzzy-name:: <<
141 Given task name(s) can be partially ma <<
142 <<
143 OPTIONS for 'perf sched timehist' <<
144 --------------------------------- <<
145 -k:: <<
146 --vmlinux=<file>:: <<
147 vmlinux pathname <<
148 <<
149 --kallsyms=<file>:: <<
150 kallsyms pathname <<
151 <<
152 -g:: <<
153 --call-graph:: <<
154 Display call chains if present (defaul <<
155 <<
156 --max-stack:: <<
157 Maximum number of functions to display <<
158 <<
159 -C=:: <<
160 --cpu=:: <<
161 Only show events for the given CPU(s) <<
162 <<
163 -p=:: <<
164 --pid=:: <<
165 Only show events for given process ID <<
166 <<
167 -t=:: <<
168 --tid=:: <<
169 Only show events for given thread ID ( <<
170 <<
171 -s:: <<
172 --summary:: <<
173 Show only a summary of scheduling by threa <<
174 run times (in sec) and relative stddev. <<
175 <<
176 -S:: <<
177 --with-summary:: <<
178 Show all scheduling events followed by a s <<
179 max, and average run times (in sec) and re <<
180 <<
181 --symfs=<directory>:: <<
182 Look for files with symbols relative to th <<
183 <<
184 -V:: <<
185 --cpu-visual:: <<
186 Show visual aid for sched switches by <<
187 's' are scheduler events. <<
188 <<
189 -w:: <<
190 --wakeups:: <<
191 Show wakeup events. <<
192 <<
193 -M:: <<
194 --migrations:: <<
195 Show migration events. <<
196 <<
197 -n:: <<
198 --next:: <<
199 Show next task. <<
200 <<
201 -I:: <<
202 --idle-hist:: <<
203 Show idle-related events only. <<
204 <<
205 --time:: <<
206 Only analyze samples within given time <<
207 have the format seconds.microseconds. <<
208 string is ',x.y') then analysis starts <<
209 stop time is not given (i.e, time stri <<
210 to end of file. <<
211 <<
212 --state:: <<
213 Show task state when it switched out. <<
214 <<
215 --show-prio:: <<
216 Show task priority. <<
217 <<
218 --prio:: <<
219 Only show events for given task priori <<
220 provided as a comma-separated list wit <<
221 priorities are specified with -: 120-1 <<
222 provided: 0,120-129. <<
223 <<
224 OPTIONS for 'perf sched replay' <<
225 ------------------------------ <<
226 <<
227 -r:: <<
228 --repeat <n>:: <<
229 repeat the workload n times (0: infini <<
230 38
231 SEE ALSO 39 SEE ALSO
232 -------- 40 --------
233 linkperf:perf-record[1] 41 linkperf:perf-record[1]
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.