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
42 event (eventsel+umask) in the form of rNNN where NNN is a
43 hexadecimal event descriptor.
47 Display this many functions.
50 --count-filter=<count>::
51 Only display functions with more events than this.
54 Put the counters into a counter group.
58 Profile at this frequency.
62 Child tasks do not inherit counters.
66 Path to vmlinux. Required for annotation functionality.
69 --mmap-pages=<pages>::
70 Number of mmap data pages (must be a power of two) or size
71 specification with appended unit character - B/K/M/G. The
72 size is rounded up to have nearest pages power of two value.
76 Profile events on existing Process ID (comma separated list).
80 Profile events on existing thread ID (comma separated list).
84 Record events in threads owned by uid. Name or number.
87 --realtime=<priority>::
88 Collect data with this RT SCHED_FIFO priority.
90 --sym-annotate=<symbol>::
94 --hide_kernel_symbols::
102 Demangle kernel symbols.
106 Dump the symbol table used for profiling.
110 Be more verbose (show counter open errors, etc).
114 Zero history across display updates.
118 Sort by key(s): pid, comm, dso, symbol, parent, srcline, weight,
119 local_weight, abort, in_tx, transaction, overhead, sample, period.
120 Please see description of --sort in the perf-report man page.
123 Specify output field - multiple keys can be specified in CSV format.
124 Following fields are available:
125 overhead, overhead_sys, overhead_us, overhead_children, sample and period.
126 Also it can contain any sort key(s).
128 By default, every sort keys not specified in --field will be appended
133 Show a column with the number of samples.
135 --show-total-period::
136 Show a column with the sum of periods.
139 Only consider symbols in these dsos. This option will affect the
140 percentage of the overhead column. See --percentage for more info.
143 Only consider symbols in these comms. This option will affect the
144 percentage of the overhead column. See --percentage for more info.
147 Only consider these symbols. This option will affect the
148 percentage of the overhead column. See --percentage for more info.
151 --disassembler-style=:: Set disassembler style for objdump.
154 Interleave source code with assembly code. Enabled by default,
155 disable with --no-source.
158 Show raw instruction encoding of assembly instructions.
161 Enables call-graph (stack chain/backtrace) recording.
163 --call-graph [mode,type,min[,limit],order[,key][,branch]]::
164 Setup and enable call-graph (stack chain/backtrace) recording,
165 implies -g. See `--call-graph` section in perf-record and
166 perf-report man pages for details.
169 Accumulate callchain of children to parent entry so that then can
170 show up in the output. The output will have a new "Children" column
171 and will be sorted on the data. It requires -g/--call-graph option
172 enabled. See the `overhead calculation' section for more details.
175 Set the stack depth limit when parsing the callchain, anything
176 beyond the specified depth will be ignored. This is a trade-off
177 between information loss and faster processing especially for
178 workloads that can have a very long callchain stack.
182 --ignore-callees=<regex>::
183 Ignore callees of the function(s) matching the given regex.
184 This has the effect of collecting the callers of each such
185 function into one place in the call-graph tree.
188 Do not show entries which have an overhead under that percent.
192 Determine how to display the overhead percentage of filtered entries.
193 Filters can be applied by --comms, --dsos and/or --symbols options and
194 Zoom operations on the TUI (thread, dso, etc).
196 "relative" means it's relative to filtered entries only so that the
197 sum of shown entries will be always 100%. "absolute" means it retains
198 the original value before and after the filter is applied.
201 --column-widths=<width[,width...]>::
202 Force each column width to the provided list, for large terminal
203 readability. 0 means no limit (default behavior).
206 When processing pre-existing threads /proc/XXX/mmap, it may take
207 a long time, because the file may be huge. A time out is needed
209 This option sets the time out limit. The default value is 500 ms.
214 Enable taken branch stack sampling. Any type of taken branch may be sampled.
215 This is a shortcut for --branch-filter any. See --branch-filter for more infos.
219 Enable taken branch stack sampling. Each sample captures a series of consecutive
220 taken branches. The number of branches captured with each sample depends on the
221 underlying hardware, the type of branches of interest, and the executed code.
222 It is possible to select the types of branches captured by enabling filters.
223 For a full list of modifiers please see the perf record manpage.
225 The option requires at least one branch type among any, any_call, any_ret, ind_call, cond.
226 The privilege levels may be omitted, in which case, the privilege levels of the associated
227 event are applied to the branch filter. Both kernel (k) and hypervisor (hv) privilege
228 levels are subject to permissions. When sampling on multiple events, branch stack sampling
229 is enabled for all the sampling events. The sampled branch type is the same for all events.
230 The various filters must be specified as a comma separated list: --branch-filter any_ret,u,k
231 Note that this feature may not be available on all processors.
234 When displaying traceevent output, do not use print fmt or plugins.
237 Enable hierarchy output.
239 INTERACTIVE PROMPTING KEYS
240 --------------------------
243 Display refresh delay.
246 Number of entries to display.
249 Event to display when multiple counters are active.
252 Profile display filter (>= hit count).
255 Annotation display filter (>= % of total).
261 Stop annotation, return to full profile display.
264 Toggle event count zeroing across display updates.
269 Pressing any unmapped key displays a menu, and prompts for input.
271 include::callchain-overhead-calculation.txt[]
275 linkperf:perf-stat[1], linkperf:perf-list[1], linkperf:perf-report[1]