6 perf-top - System profiling tool.
11 'perf top' [-e <EVENT> | --event=EVENT] [<options>]
15 This command generates and displays a performance counter profile in real time.
22 System-wide collection. (default)
26 Event period to sample.
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. Ranges of CPUs are specified with -: 0-2.
32 Default is to monitor all CPUS.
36 Number of seconds to delay between refreshes.
40 Select the PMU event. Selection can be a symbolic event name
41 (use 'perf list' to list all events) or a raw PMU event in the form
42 of rN where N is a hexadecimal value that represents the raw register
43 encoding with the layout of the event control registers as described
44 by entries in /sys/bus/event_source/devices/cpu/format/*.
48 Display this many functions.
51 --count-filter=<count>::
52 Only display functions with more events than this.
55 Sort the output by the event at the index n in group. If n is invalid,
56 sort by the first event. It can support multiple groups with different
57 amount of events. WARNING: This should be used on grouped events.
61 Profile at this frequency. Use 'max' to use the currently maximum
62 allowed frequency, i.e. the value in the kernel.perf_event_max_sample_rate
67 Child tasks do not inherit counters.
71 Path to vmlinux. Required for annotation functionality.
80 --mmap-pages=<pages>::
81 Number of mmap data pages (must be a power of two) or size
82 specification with appended unit character - B/K/M/G. The
83 size is rounded up to have nearest pages power of two value.
87 Profile events on existing Process ID (comma separated list).
91 Profile events on existing thread ID (comma separated list).
95 Record events in threads owned by uid. Name or number.
98 --realtime=<priority>::
99 Collect data with this RT SCHED_FIFO priority.
101 --sym-annotate=<symbol>::
102 Annotate this symbol.
105 --hide_kernel_symbols::
109 --hide_user_symbols::
113 Demangle kernel symbols.
117 Dump the symbol table used for profiling.
121 Be more verbose (show counter open errors, etc).
125 Zero history across display updates.
129 Sort by key(s): pid, comm, dso, symbol, parent, srcline, weight,
130 local_weight, abort, in_tx, transaction, overhead, sample, period.
131 Please see description of --sort in the perf-report man page.
134 Specify output field - multiple keys can be specified in CSV format.
135 Following fields are available:
136 overhead, overhead_sys, overhead_us, overhead_children, sample and period.
137 Also it can contain any sort key(s).
139 By default, every sort keys not specified in --field will be appended
144 Show a column with the number of samples.
146 --show-total-period::
147 Show a column with the sum of periods.
150 Only consider symbols in these dsos. This option will affect the
151 percentage of the overhead column. See --percentage for more info.
154 Only consider symbols in these comms. This option will affect the
155 percentage of the overhead column. See --percentage for more info.
158 Only consider these symbols. This option will affect the
159 percentage of the overhead column. See --percentage for more info.
162 --disassembler-style=:: Set disassembler style for objdump.
166 Remove first N entries from source file path names in executables
167 and add PREFIX. This allows to display source code compiled on systems
168 with different file system layout.
171 Interleave source code with assembly code. Enabled by default,
172 disable with --no-source.
175 Show raw instruction encoding of assembly instructions.
178 Enables call-graph (stack chain/backtrace) recording.
180 --call-graph [mode,type,min[,limit],order[,key][,branch]]::
181 Setup and enable call-graph (stack chain/backtrace) recording,
182 implies -g. See `--call-graph` section in perf-record and
183 perf-report man pages for details.
186 Accumulate callchain of children to parent entry so that then can
187 show up in the output. The output will have a new "Children" column
188 and will be sorted on the data. It requires -g/--call-graph option
189 enabled. See the `overhead calculation' section for more details.
190 Enabled by default, disable with --no-children.
193 Set the stack depth limit when parsing the callchain, anything
194 beyond the specified depth will be ignored. This is a trade-off
195 between information loss and faster processing especially for
196 workloads that can have a very long callchain stack.
198 Default: /proc/sys/kernel/perf_event_max_stack when present, 127 otherwise.
200 --ignore-callees=<regex>::
201 Ignore callees of the function(s) matching the given regex.
202 This has the effect of collecting the callers of each such
203 function into one place in the call-graph tree.
206 Do not show entries which have an overhead under that percent.
210 Determine how to display the overhead percentage of filtered entries.
211 Filters can be applied by --comms, --dsos and/or --symbols options and
212 Zoom operations on the TUI (thread, dso, etc).
214 "relative" means it's relative to filtered entries only so that the
215 sum of shown entries will be always 100%. "absolute" means it retains
216 the original value before and after the filter is applied.
219 --column-widths=<width[,width...]>::
220 Force each column width to the provided list, for large terminal
221 readability. 0 means no limit (default behavior).
224 When processing pre-existing threads /proc/XXX/mmap, it may take
225 a long time, because the file may be huge. A time out is needed
227 This option sets the time out limit. The default value is 500 ms.
232 Enable taken branch stack sampling. Any type of taken branch may be sampled.
233 This is a shortcut for --branch-filter any. See --branch-filter for more infos.
237 Enable taken branch stack sampling. Each sample captures a series of consecutive
238 taken branches. The number of branches captured with each sample depends on the
239 underlying hardware, the type of branches of interest, and the executed code.
240 It is possible to select the types of branches captured by enabling filters.
241 For a full list of modifiers please see the perf record manpage.
243 The option requires at least one branch type among any, any_call, any_ret, ind_call, cond.
244 The privilege levels may be omitted, in which case, the privilege levels of the associated
245 event are applied to the branch filter. Both kernel (k) and hypervisor (hv) privilege
246 levels are subject to permissions. When sampling on multiple events, branch stack sampling
247 is enabled for all the sampling events. The sampled branch type is the same for all events.
248 The various filters must be specified as a comma separated list: --branch-filter any_ret,u,k
249 Note that this feature may not be available on all processors.
252 When displaying traceevent output, do not use print fmt or plugins.
255 Enable hierarchy output.
258 Enable this to use just the most recent records, which helps in high core count
259 machines such as Knights Landing/Mill, but right now is disabled by default as
260 the pausing used in this technique is leading to loss of metadata events such
261 as PERF_RECORD_MMAP which makes 'perf top' unable to resolve samples, leading
262 to lots of unknown samples appearing on the UI. Enable this if you are in such
263 machines and profiling a workload that doesn't creates short lived threads and/or
264 doesn't uses many executable mmap operations. Work is being planed to solve
265 this situation, till then, this will remain disabled by default.
268 Don't do ownership validation.
270 --num-thread-synthesize::
271 The number of threads to run when synthesizing events for existing processes.
272 By default, the number of threads equals to the number of online CPUs.
275 Record events of type PERF_RECORD_NAMESPACES and display it with the
276 'cgroup_id' sort key.
280 monitor only in the container (cgroup) called "name". This option is available only
281 in per-cpu mode. The cgroup filesystem must be mounted. All threads belonging to
282 container "name" are monitored when they run on the monitored CPUs. Multiple cgroups
283 can be provided. Each cgroup is applied to the corresponding event, i.e., first cgroup
284 to first event, second cgroup to second event and so on. It is possible to provide
285 an empty cgroup (monitor all the time) using, e.g., -G foo,,bar. Cgroups must have
286 corresponding events, i.e., they always refer to events defined earlier on the command
287 line. If the user wants to track multiple events for a specific cgroup, the user can
288 use '-e e1 -e e2 -G foo,foo' or just use '-e e1 -e e2 -G foo'.
291 Record events of type PERF_RECORD_CGROUP and display it with the
294 --switch-on EVENT_NAME::
295 Only consider events after this event is found.
299 Find out where broadcast packets are handled
301 perf probe -L icmp_rcv
303 Insert a probe there:
305 perf probe icmp_rcv:59
307 Start perf top and ask it to only consider the cycles events when a
308 broadcast packet arrives This will show a menu with two entries and
309 will start counting when a broadcast packet arrives:
311 perf top -e cycles,probe:icmp_rcv --switch-on=probe:icmp_rcv
313 Alternatively one can ask for a group and then two overhead columns
314 will appear, the first for cycles and the second for the switch-on event.
316 perf top -e '{cycles,probe:icmp_rcv}' --switch-on=probe:icmp_rcv
318 This may be interesting to measure a workload only after some initialization
319 phase is over, i.e. insert a perf probe at that point and use the above
320 examples replacing probe:icmp_rcv with the just-after-init probe.
322 --switch-off EVENT_NAME::
323 Stop considering events after this event is found.
325 --show-on-off-events::
326 Show the --switch-on/off events too. This has no effect in 'perf top' now
327 but probably we'll make the default not to show the switch-on/off events
328 on the --group mode and if there is only one event besides the off/on ones,
329 go straight to the histogram browser, just like 'perf top' with no events
330 explicitly specified does.
333 Show callgraph with stitched LBRs, which may have more complete
334 callgraph. The option must be used with --call-graph lbr recording.
335 Disabled by default. In common cases with call stack overflows,
336 it can recreate better call stacks than the default lbr call stack
337 output. But this approach is not foolproof. There can be cases
338 where it creates incorrect call stacks from incorrect matches.
339 The known limitations include exception handing such as
340 setjmp/longjmp will have calls/returns not match.
343 --pfm-events events::
344 Select a PMU event using libpfm4 syntax (see http://perfmon2.sf.net)
345 including support for event filters. For example '--pfm-events
346 inst_retired:any_p:u:c=1:i'. More than one event can be passed to the
347 option using the comma separator. Hardware events and generic hardware
348 events cannot be mixed together. The latter must be used with the -e
349 option. The -e option and this one can be mixed and matched. Events
350 can be grouped using the {} notation.
353 INTERACTIVE PROMPTING KEYS
354 --------------------------
357 Display refresh delay.
360 Number of entries to display.
363 Event to display when multiple counters are active.
366 Profile display filter (>= hit count).
369 Annotation display filter (>= % of total).
375 Stop annotation, return to full profile display.
384 Toggle event count zeroing across display updates.
389 Pressing any unmapped key displays a menu, and prompts for input.
391 include::callchain-overhead-calculation.txt[]
395 linkperf:perf-stat[1], linkperf:perf-list[1], linkperf:perf-report[1]