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 Show the number of samples for each symbol
32 --show-cpu-utilization::
33 Show sample percentage for different cpu modes.
37 Show per-thread event counters. The input data file should be recorded
41 Only consider symbols in these comms. CSV that understands
42 file://filename entries. This option will affect the percentage of
43 the overhead column. See --percentage for more info.
45 Only show events for given process ID (comma separated list).
48 Only show events for given thread ID (comma separated list).
51 Only consider symbols in these dsos. CSV that understands
52 file://filename entries. This option will affect the percentage of
53 the overhead column. See --percentage for more info.
56 Only consider these symbols. CSV that understands
57 file://filename entries. This option will affect the percentage of
58 the overhead column. See --percentage for more info.
61 Only show symbols that match (partially) with this filter.
65 Only display entries resolved to a symbol.
69 Sort histogram entries by given key(s) - multiple keys can be specified
70 in CSV format. Following sort keys are available:
71 pid, comm, dso, symbol, parent, cpu, socket, srcline, weight, local_weight.
73 Each key has following meaning:
75 - comm: command (name) of the task which can be read via /proc/<pid>/comm
76 - pid: command and tid of the task
77 - dso: name of library or module executed at the time of sample
78 - symbol: name of function executed at the time of sample
79 - parent: name of function matched to the parent regex filter. Unmatched
80 entries are displayed as "[other]".
81 - cpu: cpu number the task ran at the time of sample
82 - socket: processor socket number the task ran at the time of sample
83 - srcline: filename and line number executed at the time of sample. The
84 DWARF debugging info must be provided.
85 - srcfile: file name of the source file of the same. Requires dwarf
87 - weight: Event specific weight, e.g. memory latency or transaction
88 abort cost. This is the global weight.
89 - local_weight: Local weight version of the weight above.
90 - transaction: Transaction abort flags.
91 - overhead: Overhead percentage of sample
92 - overhead_sys: Overhead percentage of sample running in system mode
93 - overhead_us: Overhead percentage of sample running in user mode
94 - overhead_guest_sys: Overhead percentage of sample running in system mode
96 - overhead_guest_us: Overhead percentage of sample running in user mode on
98 - sample: Number of sample
99 - period: Raw number of event count of sample
101 By default, comm, dso and symbol keys are used.
102 (i.e. --sort comm,dso,symbol)
104 If --branch-stack option is used, following sort keys are also
106 dso_from, dso_to, symbol_from, symbol_to, mispredict.
108 - dso_from: name of library or module branched from
109 - dso_to: name of library or module branched to
110 - symbol_from: name of function branched from
111 - symbol_to: name of function branched to
112 - mispredict: "N" for predicted branch, "Y" for mispredicted branch
113 - in_tx: branch in TSX transaction
114 - abort: TSX transaction abort.
115 - cycles: Cycles in basic block
117 And default sort keys are changed to comm, dso_from, symbol_from, dso_to
118 and symbol_to, see '--branch-stack'.
120 If the data file has tracepoint event(s), following (dynamic) sort keys
122 trace, trace_fields, [<event>.]<field>[/raw]
124 - trace: pretty printed trace output in a single column
125 - trace_fields: fields in tracepoints in separate columns
126 - <field name>: optional event and field name for a specific field
128 The last form consists of event and field names. If event name is
129 omitted, it searches all events for matching field name. The matched
130 field will be shown only for the event has the field. The event name
131 supports substring match so user doesn't need to specify full subsystem
132 and event name everytime. For example, 'sched:sched_switch' event can
133 be shortened to 'switch' as long as it's not ambiguous. Also event can
134 be specified by its index (starting from 1) preceded by the '%'.
135 So '%1' is the first event, '%2' is the second, and so on.
137 The field name can have '/raw' suffix which disables pretty printing
138 and shows raw field value like hex numbers. The --raw-trace option
139 has the same effect for all dynamic sort keys.
141 The default sort keys are changed to 'trace' if all events in the data
146 Specify output field - multiple keys can be specified in CSV format.
147 Following fields are available:
148 overhead, overhead_sys, overhead_us, overhead_children, sample and period.
149 Also it can contain any sort key(s).
151 By default, every sort keys not specified in -F will be appended
154 If --mem-mode option is used, following sort keys are also available
155 (incompatible with --branch-stack):
156 symbol_daddr, dso_daddr, locked, tlb, mem, snoop, dcacheline.
158 - symbol_daddr: name of data symbol being executed on at the time of sample
159 - dso_daddr: name of library or module containing the data being executed
160 on at the time of sample
161 - locked: whether the bus was locked at the time of sample
162 - tlb: type of tlb access for the data at the time of sample
163 - mem: type of memory access for the data at the time of sample
164 - snoop: type of snoop (if any) for the data at the time of sample
165 - dcacheline: the cacheline the data address is on at the time of sample
167 And default sort keys are changed to local_weight, mem, sym, dso,
168 symbol_daddr, dso_daddr, snoop, tlb, locked, see '--mem-mode'.
172 A regex filter to identify parent. The parent is a caller of this
173 function and searched through the callchain, thus it requires callchain
174 information recorded. The pattern is in the exteneded regex format and
175 defaults to "\^sys_|^do_page_fault", see '--sort parent'.
179 Only display entries with parent-match.
182 --column-widths=<width[,width...]>::
183 Force each column width to the provided list, for large terminal
184 readability. 0 means no limit (default behavior).
188 Use a special separator character and don't pad with spaces, replacing
189 all occurrences of this separator in symbol names (and other output)
190 with a '.' character, that thus it's the only non valid separator.
194 Dump raw trace in ASCII.
197 --call-graph=<print_type,threshold[,print_limit],order,sort_key[,branch],value>::
198 Display call chains using type, min percent threshold, print limit,
199 call order, sort key, optional branch and value. Note that ordering of
200 parameters is not fixed so any parement can be given in an arbitraty order.
201 One exception is the print_limit which should be preceded by threshold.
203 print_type can be either:
204 - flat: single column, linear exposure of call chains.
205 - graph: use a graph tree, displaying absolute overhead rates. (default)
206 - fractal: like graph, but displays relative rates. Each branch of
207 the tree is considered as a new profiled object.
208 - folded: call chains are displayed in a line, separated by semicolons
209 - none: disable call chain display.
211 threshold is a percentage value which specifies a minimum percent to be
212 included in the output call graph. Default is 0.5 (%).
214 print_limit is only applied when stdio interface is used. It's to limit
215 number of call graph entries in a single hist entry. Note that it needs
216 to be given after threshold (but not necessarily consecutive).
217 Default is 0 (unlimited).
220 - callee: callee based call graph.
221 - caller: inverted caller based call graph.
222 Default is 'caller' when --children is used, otherwise 'callee'.
225 - function: compare on functions (default)
226 - address: compare on individual code addresses
229 - branch: include last branch information in callgraph when available.
230 Usually more convenient to use --branch-history for this.
233 - percent: diplay overhead percent (default)
234 - period: display event period
235 - count: display event count
238 Accumulate callchain of children to parent entry so that then can
239 show up in the output. The output will have a new "Children" column
240 and will be sorted on the data. It requires callchains are recorded.
241 See the `overhead calculation' section for more details.
244 Set the stack depth limit when parsing the callchain, anything
245 beyond the specified depth will be ignored. This is a trade-off
246 between information loss and faster processing especially for
247 workloads that can have a very long callchain stack.
248 Note that when using the --itrace option the synthesized callchain size
249 will override this value if the synthesized callchain size is bigger.
255 alias for inverted caller based call graph.
257 --ignore-callees=<regex>::
258 Ignore callees of the function(s) matching the given regex.
259 This has the effect of collecting the callers of each such
260 function into one place in the call-graph tree.
263 Pretty printing style. key: normal, raw
265 --stdio:: Use the stdio interface.
267 --tui:: Use the TUI interface, that is integrated with annotate and allows
268 zooming into DSOs or threads, among other features. Use of --tui
269 requires a tty, if one is not present, as when piping to other
270 commands, the stdio interface is used.
272 --gtk:: Use the GTK2 interface.
283 Load module symbols. WARNING: This should only be used with -k and
288 Don't complain, do it.
290 --symfs=<directory>::
291 Look for files with symbols relative to this directory.
294 --cpu:: Only report samples for the list of CPUs provided. Multiple CPUs can
295 be provided as a comma-separated list with no space: 0,1. Ranges of
296 CPUs are specified with -: 0-2. Default is to report samples on all
300 --disassembler-style=:: Set disassembler style for objdump.
303 Interleave source code with assembly code. Enabled by default,
304 disable with --no-source.
307 Show raw instruction encoding of assembly instructions.
309 --show-total-period:: Show a column with the sum of periods.
313 Display extended information about the perf.data file. This adds
314 information which may be very large and thus may clutter the display.
315 It currently includes: cpu and numa topology of the host system.
319 Use the addresses of sampled taken branches instead of the instruction
320 address to build the histograms. To generate meaningful output, the
321 perf.data file must have been obtained using perf record -b or
322 perf record --branch-filter xxx where xxx is a branch filter option.
323 perf report is able to auto-detect whether a perf.data file contains
324 branch stacks and it will automatically switch to the branch view mode,
325 unless --no-branch-stack is used.
328 Add the addresses of sampled taken branches to the callstack.
329 This allows to examine the path the program took to each sample.
330 The data collection must have used -b (or -j) and -g.
333 Path to objdump binary.
336 Show event group information together.
339 Demangle symbol names to human readable form. It's enabled by default,
340 disable with --no-demangle.
343 Demangle kernel symbol names to human readable form (for C++ kernels).
346 Use the data addresses of samples in addition to instruction addresses
347 to build the histograms. To generate meaningful output, the perf.data
348 file must have been obtained using perf record -d -W and using a
349 special event -e cpu/mem-loads/ or -e cpu/mem-stores/. See
350 'perf mem' for simpler access.
353 Do not show entries which have an overhead under that percent.
357 Determine how to display the overhead percentage of filtered entries.
358 Filters can be applied by --comms, --dsos and/or --symbols options and
359 Zoom operations on the TUI (thread, dso, etc).
361 "relative" means it's relative to filtered entries only so that the
362 sum of shown entries will be always 100%. "absolute" means it retains
363 the original value before and after the filter is applied.
366 Show header information in the perf.data file. This includes
367 various information like hostname, OS and perf version, cpu/mem
368 info, perf command line, event list and so on. Currently only
369 --stdio output supports this feature.
372 Show only perf.data header (forces --stdio).
375 Options for decoding instruction tracing data. The options are:
377 include::itrace.txt[]
379 To disable decoding entirely, use --no-itrace.
382 Show the full path for source files for srcline output.
384 --show-ref-call-graph::
385 When multiple events are sampled, it may not be needed to collect
386 callgraphs for all of them. The sample sites are usually nearby,
387 and it's enough to collect the callgraphs on a reference event.
388 So user can use "call-graph=no" event modifier to disable callgraph
389 for other events to reduce the overhead.
390 However, perf report cannot show callgraphs for the event which
391 disable the callgraph.
392 This option extends the perf report to show reference callgraphs,
393 which collected by reference event, in no callgraph event.
396 Only report the samples on the processor socket that match with this filter
399 When displaying traceevent output, do not use print fmt or plugins.
401 include::callchain-overhead-calculation.txt[]
405 linkperf:perf-stat[1], linkperf:perf-annotate[1]