1PERF-TIMECHART(1) perf Manual PERF-TIMECHART(1)
2
3
4
6 perf-timechart - Tool to visualize total system behavior during a
7 workload
8
11 There are two variants of perf timechart:
12
13 'perf timechart record <command>' to record the system level events
14 of an arbitrary workload. By default timechart records only scheduler
15 and CPU events (task switches, running times, CPU power states, etc),
16 but it's possible to record IO (disk, network) activity using -I argument.
17
18 'perf timechart' to turn a trace into a Scalable Vector Graphics file,
19 that can be viewed with popular SVG viewers such as 'Inkscape'. Depending
20 on the events in the perf.data file, timechart will contain scheduler/cpu
21 events or IO events.
22
23 In IO mode, every bar has two charts: upper and lower.
24 Upper bar shows incoming events (disk reads, ingress network packets).
25 Lower bar shows outgoing events (disk writes, egress network packets).
26 There are also poll bars which show how much time application spent
27 in poll/epoll/select syscalls.
28
30 -o, --output=
31 Select the output file (default: output.svg)
32
33 -i, --input=
34 Select the input file (default: perf.data unless stdin is a fifo)
35
36 -w, --width=
37 Select the width of the SVG file (default: 1000)
38
39 -P, --power-only
40 Only output the CPU power section of the diagram
41
42 -T, --tasks-only
43 Don’t output processor state transitions
44
45 -p, --process
46 Select the processes to display, by name or PID
47
48 -f, --force
49 Don’t complain, do it.
50
51 --symfs=<directory>
52 Look for files with symbols relative to this directory.
53
54 -n, --proc-num
55 Print task info for at least given number of tasks.
56
57 -t, --topology
58 Sort CPUs according to topology.
59
60 --highlight=<duration_nsecs|task_name>
61 Highlight tasks (using different color) that run more than given
62 duration or tasks with given name. If number is given it’s
63 interpreted as number of nanoseconds. If non-numeric string is
64 given it’s interpreted as task name.
65
66 --io-skip-eagain
67 Don’t draw EAGAIN IO events.
68
69 --io-min-time=<nsecs>
70 Draw small events as if they lasted min-time. Useful when you need
71 to see very small and fast IO. It’s possible to specify ms or us
72 suffix to specify time in milliseconds or microseconds. Default
73 value is 1ms.
74
75 --io-merge-dist=<nsecs>
76 Merge events that are merge-dist nanoseconds apart. Reduces number
77 of figures on the SVG and makes it more render-friendly. It’s
78 possible to specify ms or us suffix to specify time in milliseconds
79 or microseconds. Default value is 1us.
80
82 -P, --power-only
83 Record only power-related events
84
85 -T, --tasks-only
86 Record only tasks-related events
87
88 -I, --io-only
89 Record only io-related events
90
91 -g, --callchain
92 Do call-graph (stack chain/backtrace) recording
93
95 $ perf timechart record git pull
96
97 [ perf record: Woken up 13 times to write data ]
98 [ perf record: Captured and wrote 4.253 MB perf.data (~185801 samples) ]
99
100 $ perf timechart
101
102 Written 10.2 seconds of trace to output.svg.
103
104 Record system-wide timechart:
105
106 $ perf timechart record
107
108 then generate timechart and highlight 'gcc' tasks:
109
110 $ perf timechart --highlight gcc
111
112 Record system-wide IO events:
113
114 $ perf timechart record -I
115
116 then generate timechart:
117
118 $ perf timechart
119
121 perf-record(1)
122
123
124
125perf 06/03/2021 PERF-TIMECHART(1)