1 perf-top(1) 1 perf-top(1)
2 =========== 2 ===========
3 3
4 NAME 4 NAME
5 ---- 5 ----
6 perf-top - System profiling tool. 6 perf-top - System profiling tool.
7 7
8 SYNOPSIS 8 SYNOPSIS
9 -------- 9 --------
10 [verse] 10 [verse]
11 'perf top' [-e <EVENT> | --event=EVENT] [<opti 11 'perf top' [-e <EVENT> | --event=EVENT] [<options>]
12 12
13 DESCRIPTION 13 DESCRIPTION
14 ----------- 14 -----------
15 This command generates and displays a performa 15 This command generates and displays a performance counter profile in real time.
16 16
17 17
18 OPTIONS 18 OPTIONS
19 ------- 19 -------
20 -a:: 20 -a::
21 --all-cpus:: 21 --all-cpus::
22 System-wide collection. (default) 22 System-wide collection. (default)
23 23
24 -c <count>:: 24 -c <count>::
25 --count=<count>:: 25 --count=<count>::
26 Event period to sample. 26 Event period to sample.
27 27
28 -C <cpu-list>:: 28 -C <cpu-list>::
29 --cpu=<cpu>:: 29 --cpu=<cpu>::
30 Monitor only on the list of CPUs provided. Mul 30 Monitor only on the list of CPUs provided. Multiple CPUs can be provided as a
31 comma-separated list with no space: 0,1. Range 31 comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2.
32 Default is to monitor all CPUS. 32 Default is to monitor all CPUS.
33 33
34 -d <seconds>:: 34 -d <seconds>::
35 --delay=<seconds>:: 35 --delay=<seconds>::
36 Number of seconds to delay between ref 36 Number of seconds to delay between refreshes.
37 37
38 -e <event>:: 38 -e <event>::
39 --event=<event>:: 39 --event=<event>::
40 Select the PMU event. Selection can be 40 Select the PMU event. Selection can be a symbolic event name
41 (use 'perf list' to list all events) o !! 41 (use 'perf list' to list all events) or a raw PMU
42 of rN where N is a hexadecimal value t !! 42 event (eventsel+umask) in the form of rNNN where NNN is a
43 encoding with the layout of the event !! 43 hexadecimal event descriptor.
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 44
50 -E <entries>:: 45 -E <entries>::
51 --entries=<entries>:: 46 --entries=<entries>::
52 Display this many functions. 47 Display this many functions.
53 48
54 -f <count>:: 49 -f <count>::
55 --count-filter=<count>:: 50 --count-filter=<count>::
56 Only display functions with more event 51 Only display functions with more events than this.
57 52
58 --group-sort-idx:: !! 53 --group::
59 Sort the output by the event at the in !! 54 Put the counters into a counter group.
60 sort by the first event. It can suppor <<
61 amount of events. WARNING: This should <<
62 55
63 -F <freq>:: 56 -F <freq>::
64 --freq=<freq>:: 57 --freq=<freq>::
65 Profile at this frequency. Use 'max' t 58 Profile at this frequency. Use 'max' to use the currently maximum
66 allowed frequency, i.e. the value in t 59 allowed frequency, i.e. the value in the kernel.perf_event_max_sample_rate
67 sysctl. 60 sysctl.
68 61
69 -i:: 62 -i::
70 --inherit:: 63 --inherit::
71 Child tasks do not inherit counters. 64 Child tasks do not inherit counters.
72 65
73 -k <path>:: 66 -k <path>::
74 --vmlinux=<path>:: 67 --vmlinux=<path>::
75 Path to vmlinux. Required for annotat 68 Path to vmlinux. Required for annotation functionality.
76 69
77 --ignore-vmlinux:: 70 --ignore-vmlinux::
78 Ignore vmlinux files. 71 Ignore vmlinux files.
79 72
80 --kallsyms=<file>:: 73 --kallsyms=<file>::
81 kallsyms pathname 74 kallsyms pathname
82 75
83 -m <pages>:: 76 -m <pages>::
84 --mmap-pages=<pages>:: 77 --mmap-pages=<pages>::
85 Number of mmap data pages (must be a p 78 Number of mmap data pages (must be a power of two) or size
86 specification in bytes with appended u !! 79 specification with appended unit character - B/K/M/G. The
87 The size is rounded up to the nearest !! 80 size is rounded up to have nearest pages power of two value.
88 81
89 -p <pid>:: 82 -p <pid>::
90 --pid=<pid>:: 83 --pid=<pid>::
91 Profile events on existing Process ID 84 Profile events on existing Process ID (comma separated list).
92 85
93 -t <tid>:: 86 -t <tid>::
94 --tid=<tid>:: 87 --tid=<tid>::
95 Profile events on existing thread ID ( 88 Profile events on existing thread ID (comma separated list).
96 89
97 -u:: 90 -u::
98 --uid=:: 91 --uid=::
99 Record events in threads owned by uid. 92 Record events in threads owned by uid. Name or number.
100 93
101 -r <priority>:: 94 -r <priority>::
102 --realtime=<priority>:: 95 --realtime=<priority>::
103 Collect data with this RT SCHED_FIFO p 96 Collect data with this RT SCHED_FIFO priority.
104 97
105 --sym-annotate=<symbol>:: 98 --sym-annotate=<symbol>::
106 Annotate this symbol. 99 Annotate this symbol.
107 100
108 -K:: 101 -K::
109 --hide_kernel_symbols:: 102 --hide_kernel_symbols::
110 Hide kernel symbols. 103 Hide kernel symbols.
111 104
112 -U:: 105 -U::
113 --hide_user_symbols:: 106 --hide_user_symbols::
114 Hide user symbols. 107 Hide user symbols.
115 108
116 --demangle-kernel:: 109 --demangle-kernel::
117 Demangle kernel symbols. 110 Demangle kernel symbols.
118 111
119 -D:: 112 -D::
120 --dump-symtab:: 113 --dump-symtab::
121 Dump the symbol table used for profili 114 Dump the symbol table used for profiling.
122 115
123 -v:: 116 -v::
124 --verbose:: 117 --verbose::
125 Be more verbose (show counter open err 118 Be more verbose (show counter open errors, etc).
126 119
127 -z:: 120 -z::
128 --zero:: 121 --zero::
129 Zero history across display updates. 122 Zero history across display updates.
130 123
131 -s:: 124 -s::
132 --sort:: 125 --sort::
133 Sort by key(s): pid, comm, dso, symbol 126 Sort by key(s): pid, comm, dso, symbol, parent, srcline, weight,
134 local_weight, abort, in_tx, transactio 127 local_weight, abort, in_tx, transaction, overhead, sample, period.
135 Please see description of --sort in th 128 Please see description of --sort in the perf-report man page.
136 129
137 --fields=:: 130 --fields=::
138 Specify output field - multiple keys c 131 Specify output field - multiple keys can be specified in CSV format.
139 Following fields are available: 132 Following fields are available:
140 overhead, overhead_sys, overhead_us, o 133 overhead, overhead_sys, overhead_us, overhead_children, sample and period.
141 Also it can contain any sort key(s). 134 Also it can contain any sort key(s).
142 135
143 By default, every sort keys not specif 136 By default, every sort keys not specified in --field will be appended
144 automatically. 137 automatically.
145 138
146 -n:: 139 -n::
147 --show-nr-samples:: 140 --show-nr-samples::
148 Show a column with the number of sampl 141 Show a column with the number of samples.
149 142
150 --show-total-period:: 143 --show-total-period::
151 Show a column with the sum of periods. 144 Show a column with the sum of periods.
152 145
153 --dsos:: 146 --dsos::
154 Only consider symbols in these dsos. 147 Only consider symbols in these dsos. This option will affect the
155 percentage of the overhead column. Se 148 percentage of the overhead column. See --percentage for more info.
156 149
157 --comms:: 150 --comms::
158 Only consider symbols in these comms. 151 Only consider symbols in these comms. This option will affect the
159 percentage of the overhead column. Se 152 percentage of the overhead column. See --percentage for more info.
160 153
161 --symbols:: 154 --symbols::
162 Only consider these symbols. This opt 155 Only consider these symbols. This option will affect the
163 percentage of the overhead column. Se 156 percentage of the overhead column. See --percentage for more info.
164 157
165 -M:: 158 -M::
166 --disassembler-style=:: Set disassembler style 159 --disassembler-style=:: Set disassembler style for objdump.
167 160
168 --addr2line=<path>:: <<
169 Path to addr2line binary. <<
170 <<
171 --objdump=<path>:: <<
172 Path to objdump binary. <<
173 <<
174 --prefix=PREFIX:: 161 --prefix=PREFIX::
175 --prefix-strip=N:: 162 --prefix-strip=N::
176 Remove first N entries from source fil 163 Remove first N entries from source file path names in executables
177 and add PREFIX. This allows to display 164 and add PREFIX. This allows to display source code compiled on systems
178 with different file system layout. 165 with different file system layout.
179 166
180 --source:: 167 --source::
181 Interleave source code with assembly c 168 Interleave source code with assembly code. Enabled by default,
182 disable with --no-source. 169 disable with --no-source.
183 170
184 --asm-raw:: 171 --asm-raw::
185 Show raw instruction encoding of assem 172 Show raw instruction encoding of assembly instructions.
186 173
187 -g:: 174 -g::
188 Enables call-graph (stack chain/backtr 175 Enables call-graph (stack chain/backtrace) recording.
189 176
190 --call-graph [mode,type,min[,limit],order[,key 177 --call-graph [mode,type,min[,limit],order[,key][,branch]]::
191 Setup and enable call-graph (stack cha 178 Setup and enable call-graph (stack chain/backtrace) recording,
192 implies -g. See `--call-graph` sectio 179 implies -g. See `--call-graph` section in perf-record and
193 perf-report man pages for details. 180 perf-report man pages for details.
194 181
195 --children:: 182 --children::
196 Accumulate callchain of children to pa 183 Accumulate callchain of children to parent entry so that then can
197 show up in the output. The output wil 184 show up in the output. The output will have a new "Children" column
198 and will be sorted on the data. It re 185 and will be sorted on the data. It requires -g/--call-graph option
199 enabled. See the `overhead calculatio 186 enabled. See the `overhead calculation' section for more details.
200 Enabled by default, disable with --no- 187 Enabled by default, disable with --no-children.
201 188
202 --max-stack:: 189 --max-stack::
203 Set the stack depth limit when parsing 190 Set the stack depth limit when parsing the callchain, anything
204 beyond the specified depth will be ign 191 beyond the specified depth will be ignored. This is a trade-off
205 between information loss and faster pr 192 between information loss and faster processing especially for
206 workloads that can have a very long ca 193 workloads that can have a very long callchain stack.
207 194
208 Default: /proc/sys/kernel/perf_event_m 195 Default: /proc/sys/kernel/perf_event_max_stack when present, 127 otherwise.
209 196
210 --ignore-callees=<regex>:: 197 --ignore-callees=<regex>::
211 Ignore callees of the function(s) matc 198 Ignore callees of the function(s) matching the given regex.
212 This has the effect of collecting the 199 This has the effect of collecting the callers of each such
213 function into one place in the call-gr 200 function into one place in the call-graph tree.
214 201
215 --percent-limit:: 202 --percent-limit::
216 Do not show entries which have an over 203 Do not show entries which have an overhead under that percent.
217 (Default: 0). 204 (Default: 0).
218 205
219 --percentage:: 206 --percentage::
220 Determine how to display the overhead 207 Determine how to display the overhead percentage of filtered entries.
221 Filters can be applied by --comms, --d 208 Filters can be applied by --comms, --dsos and/or --symbols options and
222 Zoom operations on the TUI (thread, ds 209 Zoom operations on the TUI (thread, dso, etc).
223 210
224 "relative" means it's relative to filt 211 "relative" means it's relative to filtered entries only so that the
225 sum of shown entries will be always 10 212 sum of shown entries will be always 100%. "absolute" means it retains
226 the original value before and after th 213 the original value before and after the filter is applied.
227 214
228 -w:: 215 -w::
229 --column-widths=<width[,width...]>:: 216 --column-widths=<width[,width...]>::
230 Force each column width to the provide 217 Force each column width to the provided list, for large terminal
231 readability. 0 means no limit (defaul 218 readability. 0 means no limit (default behavior).
232 219
233 --proc-map-timeout:: 220 --proc-map-timeout::
234 When processing pre-existing threads / 221 When processing pre-existing threads /proc/XXX/mmap, it may take
235 a long time, because the file may be h 222 a long time, because the file may be huge. A time out is needed
236 in such cases. 223 in such cases.
237 This option sets the time out limit. T 224 This option sets the time out limit. The default value is 500 ms.
238 225
239 226
240 -b:: 227 -b::
241 --branch-any:: 228 --branch-any::
242 Enable taken branch stack sampling. An 229 Enable taken branch stack sampling. Any type of taken branch may be sampled.
243 This is a shortcut for --branch-filter 230 This is a shortcut for --branch-filter any. See --branch-filter for more infos.
244 231
245 -j:: 232 -j::
246 --branch-filter:: 233 --branch-filter::
247 Enable taken branch stack sampling. Ea 234 Enable taken branch stack sampling. Each sample captures a series of consecutive
248 taken branches. The number of branches 235 taken branches. The number of branches captured with each sample depends on the
249 underlying hardware, the type of branc 236 underlying hardware, the type of branches of interest, and the executed code.
250 It is possible to select the types of 237 It is possible to select the types of branches captured by enabling filters.
251 For a full list of modifiers please se 238 For a full list of modifiers please see the perf record manpage.
252 239
253 The option requires at least one branc 240 The option requires at least one branch type among any, any_call, any_ret, ind_call, cond.
254 The privilege levels may be omitted, i 241 The privilege levels may be omitted, in which case, the privilege levels of the associated
255 event are applied to the branch filter 242 event are applied to the branch filter. Both kernel (k) and hypervisor (hv) privilege
256 levels are subject to permissions. Wh 243 levels are subject to permissions. When sampling on multiple events, branch stack sampling
257 is enabled for all the sampling events 244 is enabled for all the sampling events. The sampled branch type is the same for all events.
258 The various filters must be specified 245 The various filters must be specified as a comma separated list: --branch-filter any_ret,u,k
259 Note that this feature may not be avai 246 Note that this feature may not be available on all processors.
260 247
261 --branch-history:: <<
262 Add the addresses of sampled taken bra <<
263 This allows to examine the path the pr <<
264 <<
265 --raw-trace:: 248 --raw-trace::
266 When displaying traceevent output, do 249 When displaying traceevent output, do not use print fmt or plugins.
267 250
268 -H:: <<
269 --hierarchy:: 251 --hierarchy::
270 Enable hierarchical output. In the hi !! 252 Enable hierarchy output.
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 <<
300 253
301 --overwrite:: 254 --overwrite::
302 Enable this to use just the most recen 255 Enable this to use just the most recent records, which helps in high core count
303 machines such as Knights Landing/Mill, 256 machines such as Knights Landing/Mill, but right now is disabled by default as
304 the pausing used in this technique is 257 the pausing used in this technique is leading to loss of metadata events such
305 as PERF_RECORD_MMAP which makes 'perf 258 as PERF_RECORD_MMAP which makes 'perf top' unable to resolve samples, leading
306 to lots of unknown samples appearing o 259 to lots of unknown samples appearing on the UI. Enable this if you are in such
307 machines and profiling a workload that 260 machines and profiling a workload that doesn't creates short lived threads and/or
308 doesn't uses many executable mmap oper 261 doesn't uses many executable mmap operations. Work is being planed to solve
309 this situation, till then, this will r 262 this situation, till then, this will remain disabled by default.
310 263
311 --force:: 264 --force::
312 Don't do ownership validation. 265 Don't do ownership validation.
313 266
314 --num-thread-synthesize:: 267 --num-thread-synthesize::
315 The number of threads to run when synt 268 The number of threads to run when synthesizing events for existing processes.
316 By default, the number of threads equa 269 By default, the number of threads equals to the number of online CPUs.
317 270
318 --namespaces:: 271 --namespaces::
319 Record events of type PERF_RECORD_NAME 272 Record events of type PERF_RECORD_NAMESPACES and display it with the
320 'cgroup_id' sort key. 273 'cgroup_id' sort key.
321 274
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:: 275 --switch-on EVENT_NAME::
339 Only consider events after this event 276 Only consider events after this event is found.
340 277
341 E.g.: 278 E.g.:
342 279
343 Find out where broadcast packets ar 280 Find out where broadcast packets are handled
344 281
345 perf probe -L icmp_rcv 282 perf probe -L icmp_rcv
346 283
347 Insert a probe there: 284 Insert a probe there:
348 285
349 perf probe icmp_rcv:59 286 perf probe icmp_rcv:59
350 287
351 Start perf top and ask it to only c 288 Start perf top and ask it to only consider the cycles events when a
352 broadcast packet arrives This will 289 broadcast packet arrives This will show a menu with two entries and
353 will start counting when a broadcas 290 will start counting when a broadcast packet arrives:
354 291
355 perf top -e cycles,probe:icmp_ 292 perf top -e cycles,probe:icmp_rcv --switch-on=probe:icmp_rcv
356 293
357 Alternatively one can ask for a gro !! 294 Alternatively one can ask for --group and then two overhead columns
358 will appear, the first for cycles a 295 will appear, the first for cycles and the second for the switch-on event.
359 296
360 perf top -e '{cycles,probe:icm !! 297 perf top --group -e cycles,probe:icmp_rcv --switch-on=probe:icmp_rcv
361 298
362 This may be interesting to measure a w 299 This may be interesting to measure a workload only after some initialization
363 phase is over, i.e. insert a perf prob 300 phase is over, i.e. insert a perf probe at that point and use the above
364 examples replacing probe:icmp_rcv with 301 examples replacing probe:icmp_rcv with the just-after-init probe.
365 302
366 --switch-off EVENT_NAME:: 303 --switch-off EVENT_NAME::
367 Stop considering events after this eve 304 Stop considering events after this event is found.
368 305
369 --show-on-off-events:: 306 --show-on-off-events::
370 Show the --switch-on/off events too. T 307 Show the --switch-on/off events too. This has no effect in 'perf top' now
371 but probably we'll make the default no 308 but probably we'll make the default not to show the switch-on/off events
372 on the --group mode and if there is on 309 on the --group mode and if there is only one event besides the off/on ones,
373 go straight to the histogram browser, 310 go straight to the histogram browser, just like 'perf top' with no events
374 explicitly specified does. !! 311 explicitely specified does.
375 312
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[] <<
396 313
397 INTERACTIVE PROMPTING KEYS 314 INTERACTIVE PROMPTING KEYS
398 -------------------------- 315 --------------------------
399 316
400 [d]:: 317 [d]::
401 Display refresh delay. 318 Display refresh delay.
402 319
403 [e]:: 320 [e]::
404 Number of entries to display. 321 Number of entries to display.
405 322
406 [E]:: 323 [E]::
407 Event to display when multiple counter 324 Event to display when multiple counters are active.
408 325
409 [f]:: 326 [f]::
410 Profile display filter (>= hit count). 327 Profile display filter (>= hit count).
411 328
412 [F]:: 329 [F]::
413 Annotation display filter (>= % of tot 330 Annotation display filter (>= % of total).
414 331
415 [s]:: 332 [s]::
416 Annotate symbol. 333 Annotate symbol.
417 334
418 [S]:: 335 [S]::
419 Stop annotation, return to full profil 336 Stop annotation, return to full profile display.
420 337
421 [K]:: 338 [K]::
422 Hide kernel symbols. 339 Hide kernel symbols.
423 340
424 [U]:: 341 [U]::
425 Hide user symbols. 342 Hide user symbols.
426 343
427 [z]:: 344 [z]::
428 Toggle event count zeroing across disp 345 Toggle event count zeroing across display updates.
429 346
430 [qQ]:: 347 [qQ]::
431 Quit. 348 Quit.
432 349
433 Pressing any unmapped key displays a menu, and 350 Pressing any unmapped key displays a menu, and prompts for input.
434 351
435 include::callchain-overhead-calculation.txt[] 352 include::callchain-overhead-calculation.txt[]
436 353
437 SEE ALSO 354 SEE ALSO
438 -------- 355 --------
439 linkperf:perf-stat[1], linkperf:perf-list[1], 356 linkperf:perf-stat[1], linkperf:perf-list[1], linkperf:perf-report[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.