1 perf-top(1)
2 ===========
3
4 NAME
5 ----
6 perf-top - System profiling tool.
7
8 SYNOPSIS
9 --------
10 [verse]
11 'perf top' [-e <EVENT> | --event=EVENT] [<opti
12
13 DESCRIPTION
14 -----------
15 This command generates and displays a performa
16
17
18 OPTIONS
19 -------
20 -a::
21 --all-cpus::
22 System-wide collection. (default)
23
24 -c <count>::
25 --count=<count>::
26 Event period to sample.
27
28 -C <cpu-list>::
29 --cpu=<cpu>::
30 Monitor only on the list of CPUs provided. Mul
31 comma-separated list with no space: 0,1. Range
32 Default is to monitor all CPUS.
33
34 -d <seconds>::
35 --delay=<seconds>::
36 Number of seconds to delay between ref
37
38 -e <event>::
39 --event=<event>::
40 Select the PMU event. Selection can be
41 (use 'perf list' to list all events) o
42 of rN where N is a hexadecimal value t
43 encoding with the layout of the event
44 by entries in /sys/bus/event_source/de
45
46 --filter=<filter>::
47 Event filter. This option should foll
48 syntax see linkperf:perf-record[1].
49
50 -E <entries>::
51 --entries=<entries>::
52 Display this many functions.
53
54 -f <count>::
55 --count-filter=<count>::
56 Only display functions with more event
57
58 --group-sort-idx::
59 Sort the output by the event at the in
60 sort by the first event. It can suppor
61 amount of events. WARNING: This should
62
63 -F <freq>::
64 --freq=<freq>::
65 Profile at this frequency. Use 'max' t
66 allowed frequency, i.e. the value in t
67 sysctl.
68
69 -i::
70 --inherit::
71 Child tasks do not inherit counters.
72
73 -k <path>::
74 --vmlinux=<path>::
75 Path to vmlinux. Required for annotat
76
77 --ignore-vmlinux::
78 Ignore vmlinux files.
79
80 --kallsyms=<file>::
81 kallsyms pathname
82
83 -m <pages>::
84 --mmap-pages=<pages>::
85 Number of mmap data pages (must be a p
86 specification in bytes with appended u
87 The size is rounded up to the nearest
88
89 -p <pid>::
90 --pid=<pid>::
91 Profile events on existing Process ID
92
93 -t <tid>::
94 --tid=<tid>::
95 Profile events on existing thread ID (
96
97 -u::
98 --uid=::
99 Record events in threads owned by uid.
100
101 -r <priority>::
102 --realtime=<priority>::
103 Collect data with this RT SCHED_FIFO p
104
105 --sym-annotate=<symbol>::
106 Annotate this symbol.
107
108 -K::
109 --hide_kernel_symbols::
110 Hide kernel symbols.
111
112 -U::
113 --hide_user_symbols::
114 Hide user symbols.
115
116 --demangle-kernel::
117 Demangle kernel symbols.
118
119 -D::
120 --dump-symtab::
121 Dump the symbol table used for profili
122
123 -v::
124 --verbose::
125 Be more verbose (show counter open err
126
127 -z::
128 --zero::
129 Zero history across display updates.
130
131 -s::
132 --sort::
133 Sort by key(s): pid, comm, dso, symbol
134 local_weight, abort, in_tx, transactio
135 Please see description of --sort in th
136
137 --fields=::
138 Specify output field - multiple keys c
139 Following fields are available:
140 overhead, overhead_sys, overhead_us, o
141 Also it can contain any sort key(s).
142
143 By default, every sort keys not specif
144 automatically.
145
146 -n::
147 --show-nr-samples::
148 Show a column with the number of sampl
149
150 --show-total-period::
151 Show a column with the sum of periods.
152
153 --dsos::
154 Only consider symbols in these dsos.
155 percentage of the overhead column. Se
156
157 --comms::
158 Only consider symbols in these comms.
159 percentage of the overhead column. Se
160
161 --symbols::
162 Only consider these symbols. This opt
163 percentage of the overhead column. Se
164
165 -M::
166 --disassembler-style=:: Set disassembler style
167
168 --addr2line=<path>::
169 Path to addr2line binary.
170
171 --objdump=<path>::
172 Path to objdump binary.
173
174 --prefix=PREFIX::
175 --prefix-strip=N::
176 Remove first N entries from source fil
177 and add PREFIX. This allows to display
178 with different file system layout.
179
180 --source::
181 Interleave source code with assembly c
182 disable with --no-source.
183
184 --asm-raw::
185 Show raw instruction encoding of assem
186
187 -g::
188 Enables call-graph (stack chain/backtr
189
190 --call-graph [mode,type,min[,limit],order[,key
191 Setup and enable call-graph (stack cha
192 implies -g. See `--call-graph` sectio
193 perf-report man pages for details.
194
195 --children::
196 Accumulate callchain of children to pa
197 show up in the output. The output wil
198 and will be sorted on the data. It re
199 enabled. See the `overhead calculatio
200 Enabled by default, disable with --no-
201
202 --max-stack::
203 Set the stack depth limit when parsing
204 beyond the specified depth will be ign
205 between information loss and faster pr
206 workloads that can have a very long ca
207
208 Default: /proc/sys/kernel/perf_event_m
209
210 --ignore-callees=<regex>::
211 Ignore callees of the function(s) matc
212 This has the effect of collecting the
213 function into one place in the call-gr
214
215 --percent-limit::
216 Do not show entries which have an over
217 (Default: 0).
218
219 --percentage::
220 Determine how to display the overhead
221 Filters can be applied by --comms, --d
222 Zoom operations on the TUI (thread, ds
223
224 "relative" means it's relative to filt
225 sum of shown entries will be always 10
226 the original value before and after th
227
228 -w::
229 --column-widths=<width[,width...]>::
230 Force each column width to the provide
231 readability. 0 means no limit (defaul
232
233 --proc-map-timeout::
234 When processing pre-existing threads /
235 a long time, because the file may be h
236 in such cases.
237 This option sets the time out limit. T
238
239
240 -b::
241 --branch-any::
242 Enable taken branch stack sampling. An
243 This is a shortcut for --branch-filter
244
245 -j::
246 --branch-filter::
247 Enable taken branch stack sampling. Ea
248 taken branches. The number of branches
249 underlying hardware, the type of branc
250 It is possible to select the types of
251 For a full list of modifiers please se
252
253 The option requires at least one branc
254 The privilege levels may be omitted, i
255 event are applied to the branch filter
256 levels are subject to permissions. Wh
257 is enabled for all the sampling events
258 The various filters must be specified
259 Note that this feature may not be avai
260
261 --branch-history::
262 Add the addresses of sampled taken bra
263 This allows to examine the path the pr
264
265 --raw-trace::
266 When displaying traceevent output, do
267
268 -H::
269 --hierarchy::
270 Enable hierarchical output. In the hi
271 samples based on the criteria and then
272 level sort key.
273
274 For example, in normal output:
275
276 perf report -s dso,sym
277 #
278 # Overhead Shared Object Symbo
279 # ........ ................. .....
280 50.00% [kernel.kallsyms] [k] k
281 20.00% perf [.] f
282 15.00% [kernel.kallsyms] [k] k
283 10.00% perf [.] b
284 5.00% libc.so [.] l
285
286 In hierarchy output:
287
288 perf report -s dso,sym --hierarchy
289 #
290 # Overhead Shared Object / Symbol
291 # .......... ......................
292 65.00% [kernel.kallsyms]
293 50.00% [k] kfunc1
294 15.00% [k] kfunc2
295 30.00% perf
296 20.00% [.] foo
297 10.00% [.] bar
298 5.00% libc.so
299 5.00% [.] libcall
300
301 --overwrite::
302 Enable this to use just the most recen
303 machines such as Knights Landing/Mill,
304 the pausing used in this technique is
305 as PERF_RECORD_MMAP which makes 'perf
306 to lots of unknown samples appearing o
307 machines and profiling a workload that
308 doesn't uses many executable mmap oper
309 this situation, till then, this will r
310
311 --force::
312 Don't do ownership validation.
313
314 --num-thread-synthesize::
315 The number of threads to run when synt
316 By default, the number of threads equa
317
318 --namespaces::
319 Record events of type PERF_RECORD_NAME
320 'cgroup_id' sort key.
321
322 -G name::
323 --cgroup name::
324 monitor only in the container (cgroup) called
325 in per-cpu mode. The cgroup filesystem must be
326 container "name" are monitored when they run o
327 can be provided. Each cgroup is applied to the
328 to first event, second cgroup to second event
329 an empty cgroup (monitor all the time) using,
330 corresponding events, i.e., they always refer
331 line. If the user wants to track multiple even
332 use '-e e1 -e e2 -G foo,foo' or just use '-e e
333
334 --all-cgroups::
335 Record events of type PERF_RECORD_CGRO
336 'cgroup' sort key.
337
338 --switch-on EVENT_NAME::
339 Only consider events after this event
340
341 E.g.:
342
343 Find out where broadcast packets ar
344
345 perf probe -L icmp_rcv
346
347 Insert a probe there:
348
349 perf probe icmp_rcv:59
350
351 Start perf top and ask it to only c
352 broadcast packet arrives This will
353 will start counting when a broadcas
354
355 perf top -e cycles,probe:icmp_
356
357 Alternatively one can ask for a gro
358 will appear, the first for cycles a
359
360 perf top -e '{cycles,probe:icm
361
362 This may be interesting to measure a w
363 phase is over, i.e. insert a perf prob
364 examples replacing probe:icmp_rcv with
365
366 --switch-off EVENT_NAME::
367 Stop considering events after this eve
368
369 --show-on-off-events::
370 Show the --switch-on/off events too. T
371 but probably we'll make the default no
372 on the --group mode and if there is on
373 go straight to the histogram browser,
374 explicitly specified does.
375
376 --stitch-lbr::
377 Show callgraph with stitched LBRs, whi
378 callgraph. The option must be used wit
379 Disabled by default. In common cases w
380 it can recreate better call stacks tha
381 output. But this approach is not foolp
382 where it creates incorrect call stacks
383 The known limitations include exceptio
384 setjmp/longjmp will have calls/returns
385
386 ifdef::HAVE_LIBPFM[]
387 --pfm-events events::
388 Select a PMU event using libpfm4 syntax (see h
389 including support for event filters. For examp
390 inst_retired:any_p:u:c=1:i'. More than one eve
391 option using the comma separator. Hardware eve
392 events cannot be mixed together. The latter mu
393 option. The -e option and this one can be mixe
394 can be grouped using the {} notation.
395 endif::HAVE_LIBPFM[]
396
397 INTERACTIVE PROMPTING KEYS
398 --------------------------
399
400 [d]::
401 Display refresh delay.
402
403 [e]::
404 Number of entries to display.
405
406 [E]::
407 Event to display when multiple counter
408
409 [f]::
410 Profile display filter (>= hit count).
411
412 [F]::
413 Annotation display filter (>= % of tot
414
415 [s]::
416 Annotate symbol.
417
418 [S]::
419 Stop annotation, return to full profil
420
421 [K]::
422 Hide kernel symbols.
423
424 [U]::
425 Hide user symbols.
426
427 [z]::
428 Toggle event count zeroing across disp
429
430 [qQ]::
431 Quit.
432
433 Pressing any unmapped key displays a menu, and
434
435 include::callchain-overhead-calculation.txt[]
436
437 SEE ALSO
438 --------
439 linkperf:perf-stat[1], linkperf:perf-list[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.