6 perf-report - Read perf.data (created by perf record) and display the profile
11 'perf report' [-i <file> | --input=file]
15 This command displays the performance counter profile information recorded
22 Input file name. (default: perf.data unless stdin is a fifo)
26 Be more verbose. (show symbol address, etc)
30 Do not show any message. (Suppress -v)
34 Show the number of samples for each symbol
36 --show-cpu-utilization::
37 Show sample percentage for different cpu modes.
41 Show per-thread event counters. The input data file should be recorded
45 Only consider symbols in these comms. CSV that understands
46 file://filename entries. This option will affect the percentage of
47 the overhead column. See --percentage for more info.
49 Only show events for given process ID (comma separated list).
52 Only show events for given thread ID (comma separated list).
55 Only consider symbols in these dsos. CSV that understands
56 file://filename entries. This option will affect the percentage of
57 the overhead column. See --percentage for more info.
60 Only consider these symbols. CSV that understands
61 file://filename entries. This option will affect the percentage of
62 the overhead column. See --percentage for more info.
65 Only show symbols that match (partially) with this filter.
69 Only display entries resolved to a symbol.
73 Sort histogram entries by given key(s) - multiple keys can be specified
74 in CSV format. Following sort keys are available:
75 pid, comm, dso, symbol, parent, cpu, socket, srcline, weight,
76 local_weight, cgroup_id.
78 Each key has following meaning:
80 - comm: command (name) of the task which can be read via /proc/<pid>/comm
81 - pid: command and tid of the task
82 - dso: name of library or module executed at the time of sample
83 - dso_size: size of library or module executed at the time of sample
84 - symbol: name of function executed at the time of sample
85 - symbol_size: size of function executed at the time of sample
86 - parent: name of function matched to the parent regex filter. Unmatched
87 entries are displayed as "[other]".
88 - cpu: cpu number the task ran at the time of sample
89 - socket: processor socket number the task ran at the time of sample
90 - srcline: filename and line number executed at the time of sample. The
91 DWARF debugging info must be provided.
92 - srcfile: file name of the source file of the same. Requires dwarf
94 - weight: Event specific weight, e.g. memory latency or transaction
95 abort cost. This is the global weight.
96 - local_weight: Local weight version of the weight above.
97 - cgroup_id: ID derived from cgroup namespace device and inode numbers.
98 - transaction: Transaction abort flags.
99 - overhead: Overhead percentage of sample
100 - overhead_sys: Overhead percentage of sample running in system mode
101 - overhead_us: Overhead percentage of sample running in user mode
102 - overhead_guest_sys: Overhead percentage of sample running in system mode
104 - overhead_guest_us: Overhead percentage of sample running in user mode on
106 - sample: Number of sample
107 - period: Raw number of event count of sample
109 By default, comm, dso and symbol keys are used.
110 (i.e. --sort comm,dso,symbol)
112 If --branch-stack option is used, following sort keys are also
115 - dso_from: name of library or module branched from
116 - dso_to: name of library or module branched to
117 - symbol_from: name of function branched from
118 - symbol_to: name of function branched to
119 - srcline_from: source file and line branched from
120 - srcline_to: source file and line branched to
121 - mispredict: "N" for predicted branch, "Y" for mispredicted branch
122 - in_tx: branch in TSX transaction
123 - abort: TSX transaction abort.
124 - cycles: Cycles in basic block
126 And default sort keys are changed to comm, dso_from, symbol_from, dso_to
127 and symbol_to, see '--branch-stack'.
129 If the --mem-mode option is used, the following sort keys are also available
130 (incompatible with --branch-stack):
131 symbol_daddr, dso_daddr, locked, tlb, mem, snoop, dcacheline.
133 - symbol_daddr: name of data symbol being executed on at the time of sample
134 - dso_daddr: name of library or module containing the data being executed
135 on at the time of the sample
136 - locked: whether the bus was locked at the time of the sample
137 - tlb: type of tlb access for the data at the time of the sample
138 - mem: type of memory access for the data at the time of the sample
139 - snoop: type of snoop (if any) for the data at the time of the sample
140 - dcacheline: the cacheline the data address is on at the time of the sample
141 - phys_daddr: physical address of data being executed on at the time of sample
143 And the default sort keys are changed to local_weight, mem, sym, dso,
144 symbol_daddr, dso_daddr, snoop, tlb, locked, see '--mem-mode'.
146 If the data file has tracepoint event(s), following (dynamic) sort keys
148 trace, trace_fields, [<event>.]<field>[/raw]
150 - trace: pretty printed trace output in a single column
151 - trace_fields: fields in tracepoints in separate columns
152 - <field name>: optional event and field name for a specific field
154 The last form consists of event and field names. If event name is
155 omitted, it searches all events for matching field name. The matched
156 field will be shown only for the event has the field. The event name
157 supports substring match so user doesn't need to specify full subsystem
158 and event name everytime. For example, 'sched:sched_switch' event can
159 be shortened to 'switch' as long as it's not ambiguous. Also event can
160 be specified by its index (starting from 1) preceded by the '%'.
161 So '%1' is the first event, '%2' is the second, and so on.
163 The field name can have '/raw' suffix which disables pretty printing
164 and shows raw field value like hex numbers. The --raw-trace option
165 has the same effect for all dynamic sort keys.
167 The default sort keys are changed to 'trace' if all events in the data
172 Specify output field - multiple keys can be specified in CSV format.
173 Following fields are available:
174 overhead, overhead_sys, overhead_us, overhead_children, sample and period.
175 Also it can contain any sort key(s).
177 By default, every sort keys not specified in -F will be appended
180 If the keys starts with a prefix '+', then it will append the specified
181 field(s) to the default field order. For example: perf report -F +period,sample.
185 A regex filter to identify parent. The parent is a caller of this
186 function and searched through the callchain, thus it requires callchain
187 information recorded. The pattern is in the extended regex format and
188 defaults to "\^sys_|^do_page_fault", see '--sort parent'.
192 Only display entries with parent-match.
195 --column-widths=<width[,width...]>::
196 Force each column width to the provided list, for large terminal
197 readability. 0 means no limit (default behavior).
201 Use a special separator character and don't pad with spaces, replacing
202 all occurrences of this separator in symbol names (and other output)
203 with a '.' character, that thus it's the only non valid separator.
207 Dump raw trace in ASCII.
210 --call-graph=<print_type,threshold[,print_limit],order,sort_key[,branch],value>::
211 Display call chains using type, min percent threshold, print limit,
212 call order, sort key, optional branch and value. Note that ordering
213 is not fixed so any parameter can be given in an arbitrary order.
214 One exception is the print_limit which should be preceded by threshold.
216 print_type can be either:
217 - flat: single column, linear exposure of call chains.
218 - graph: use a graph tree, displaying absolute overhead rates. (default)
219 - fractal: like graph, but displays relative rates. Each branch of
220 the tree is considered as a new profiled object.
221 - folded: call chains are displayed in a line, separated by semicolons
222 - none: disable call chain display.
224 threshold is a percentage value which specifies a minimum percent to be
225 included in the output call graph. Default is 0.5 (%).
227 print_limit is only applied when stdio interface is used. It's to limit
228 number of call graph entries in a single hist entry. Note that it needs
229 to be given after threshold (but not necessarily consecutive).
230 Default is 0 (unlimited).
233 - callee: callee based call graph.
234 - caller: inverted caller based call graph.
235 Default is 'caller' when --children is used, otherwise 'callee'.
238 - function: compare on functions (default)
239 - address: compare on individual code addresses
240 - srcline: compare on source filename and line number
243 - branch: include last branch information in callgraph when available.
244 Usually more convenient to use --branch-history for this.
247 - percent: diplay overhead percent (default)
248 - period: display event period
249 - count: display event count
252 Accumulate callchain of children to parent entry so that then can
253 show up in the output. The output will have a new "Children" column
254 and will be sorted on the data. It requires callchains are recorded.
255 See the `overhead calculation' section for more details. Enabled by
256 default, disable with --no-children.
259 Set the stack depth limit when parsing the callchain, anything
260 beyond the specified depth will be ignored. This is a trade-off
261 between information loss and faster processing especially for
262 workloads that can have a very long callchain stack.
263 Note that when using the --itrace option the synthesized callchain size
264 will override this value if the synthesized callchain size is bigger.
270 alias for inverted caller based call graph.
272 --ignore-callees=<regex>::
273 Ignore callees of the function(s) matching the given regex.
274 This has the effect of collecting the callers of each such
275 function into one place in the call-graph tree.
278 Pretty printing style. key: normal, raw
280 --stdio:: Use the stdio interface.
283 'always', 'never' or 'auto', allowing configuring color output
284 via the command line, in addition to via "color.ui" .perfconfig.
285 Use '--stdio-color always' to generate color even when redirecting
286 to a pipe or file. Using just '--stdio-color' is equivalent to
289 --tui:: Use the TUI interface, that is integrated with annotate and allows
290 zooming into DSOs or threads, among other features. Use of --tui
291 requires a tty, if one is not present, as when piping to other
292 commands, the stdio interface is used.
294 --gtk:: Use the GTK2 interface.
301 Ignore vmlinux files.
308 Load module symbols. WARNING: This should only be used with -k and
313 Don't do ownership validation.
315 --symfs=<directory>::
316 Look for files with symbols relative to this directory.
319 --cpu:: Only report samples for the list of CPUs provided. Multiple CPUs can
320 be provided as a comma-separated list with no space: 0,1. Ranges of
321 CPUs are specified with -: 0-2. Default is to report samples on all
325 --disassembler-style=:: Set disassembler style for objdump.
328 Interleave source code with assembly code. Enabled by default,
329 disable with --no-source.
332 Show raw instruction encoding of assembly instructions.
334 --show-total-period:: Show a column with the sum of periods.
338 Display extended information about the perf.data file. This adds
339 information which may be very large and thus may clutter the display.
340 It currently includes: cpu and numa topology of the host system.
344 Use the addresses of sampled taken branches instead of the instruction
345 address to build the histograms. To generate meaningful output, the
346 perf.data file must have been obtained using perf record -b or
347 perf record --branch-filter xxx where xxx is a branch filter option.
348 perf report is able to auto-detect whether a perf.data file contains
349 branch stacks and it will automatically switch to the branch view mode,
350 unless --no-branch-stack is used.
353 Add the addresses of sampled taken branches to the callstack.
354 This allows to examine the path the program took to each sample.
355 The data collection must have used -b (or -j) and -g.
358 Path to objdump binary.
361 Show event group information together. It forces group output also
362 if there are no groups defined in data file.
365 Demangle symbol names to human readable form. It's enabled by default,
366 disable with --no-demangle.
369 Demangle kernel symbol names to human readable form (for C++ kernels).
372 Use the data addresses of samples in addition to instruction addresses
373 to build the histograms. To generate meaningful output, the perf.data
374 file must have been obtained using perf record -d -W and using a
375 special event -e cpu/mem-loads/p or -e cpu/mem-stores/p. See
376 'perf mem' for simpler access.
379 Do not show entries which have an overhead under that percent.
380 (Default: 0). Note that this option also sets the percent limit (threshold)
381 of callchains. However the default value of callchain threshold is
382 different than the default value of hist entries. Please see the
383 --call-graph option for details.
386 Determine how to display the overhead percentage of filtered entries.
387 Filters can be applied by --comms, --dsos and/or --symbols options and
388 Zoom operations on the TUI (thread, dso, etc).
390 "relative" means it's relative to filtered entries only so that the
391 sum of shown entries will be always 100%. "absolute" means it retains
392 the original value before and after the filter is applied.
395 Show header information in the perf.data file. This includes
396 various information like hostname, OS and perf version, cpu/mem
397 info, perf command line, event list and so on. Currently only
398 --stdio output supports this feature.
401 Show only perf.data header (forces --stdio).
404 Only analyze samples within given time window: <start>,<stop>. Times
405 have the format seconds.microseconds. If start is not given (i.e., time
406 string is ',x.y') then analysis starts at the beginning of the file. If
407 stop time is not given (i.e, time string is 'x.y,') then analysis goes
410 Also support time percent with multiple time range. Time string is
411 'a%/n,b%/m,...' or 'a%-b%,c%-%d,...'.
414 Select the second 10% time slice:
416 perf report --time 10%/2
418 Select from 0% to 10% time slice:
420 perf report --time 0%-10%
422 Select the first and second 10% time slices:
424 perf report --time 10%/1,10%/2
426 Select from 0% to 10% and 30% to 40% slices:
428 perf report --time 0%-10%,30%-40%
431 Options for decoding instruction tracing data. The options are:
433 include::itrace.txt[]
435 To disable decoding entirely, use --no-itrace.
438 Show the full path for source files for srcline output.
440 --show-ref-call-graph::
441 When multiple events are sampled, it may not be needed to collect
442 callgraphs for all of them. The sample sites are usually nearby,
443 and it's enough to collect the callgraphs on a reference event.
444 So user can use "call-graph=no" event modifier to disable callgraph
445 for other events to reduce the overhead.
446 However, perf report cannot show callgraphs for the event which
447 disable the callgraph.
448 This option extends the perf report to show reference callgraphs,
449 which collected by reference event, in no callgraph event.
452 Only report the samples on the processor socket that match with this filter
455 When displaying traceevent output, do not use print fmt or plugins.
458 Enable hierarchical output.
461 If a callgraph address belongs to an inlined function, the inline stack
462 will be printed. Each entry is function name or file/line. Enabled by
463 default, disable with --no-inline.
466 Show --tasks output plus mmap information in a format similar to
469 Please note that not all mmaps are stored, options affecting which ones
470 are include 'perf record --data', for instance.
473 Display overall events statistics without any further processing.
474 (like the one at the end of the perf report -D command)
477 Display monitored tasks stored in perf data. Displaying pid/tid/ppid
478 plus the command string aligned to distinguish parent and child tasks.
480 include::callchain-overhead-calculation.txt[]
484 linkperf:perf-stat[1], linkperf:perf-annotate[1], linkperf:perf-record[1]