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 --showcpuutilization::
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, 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 - srcline: filename and line number executed at the time of sample. The
83 DWARF debugging info must be provided.
84 - weight: Event specific weight, e.g. memory latency or transaction
85 abort cost. This is the global weight.
86 - local_weight: Local weight version of the weight above.
87 - transaction: Transaction abort flags.
88 - overhead: Overhead percentage of sample
89 - overhead_sys: Overhead percentage of sample running in system mode
90 - overhead_us: Overhead percentage of sample running in user mode
91 - overhead_guest_sys: Overhead percentage of sample running in system mode
93 - overhead_guest_us: Overhead percentage of sample running in user mode on
95 - sample: Number of sample
96 - period: Raw number of event count of sample
98 By default, comm, dso and symbol keys are used.
99 (i.e. --sort comm,dso,symbol)
101 If --branch-stack option is used, following sort keys are also
103 dso_from, dso_to, symbol_from, symbol_to, mispredict.
105 - dso_from: name of library or module branched from
106 - dso_to: name of library or module branched to
107 - symbol_from: name of function branched from
108 - symbol_to: name of function branched to
109 - mispredict: "N" for predicted branch, "Y" for mispredicted branch
110 - in_tx: branch in TSX transaction
111 - abort: TSX transaction abort.
112 - cycles: Cycles in basic block
114 And default sort keys are changed to comm, dso_from, symbol_from, dso_to
115 and symbol_to, see '--branch-stack'.
119 Specify output field - multiple keys can be specified in CSV format.
120 Following fields are available:
121 overhead, overhead_sys, overhead_us, overhead_children, sample and period.
122 Also it can contain any sort key(s).
124 By default, every sort keys not specified in -F will be appended
127 If --mem-mode option is used, following sort keys are also available
128 (incompatible with --branch-stack):
129 symbol_daddr, dso_daddr, locked, tlb, mem, snoop, dcacheline.
131 - symbol_daddr: name of data symbol being executed on at the time of sample
132 - dso_daddr: name of library or module containing the data being executed
133 on at the time of sample
134 - locked: whether the bus was locked at the time of sample
135 - tlb: type of tlb access for the data at the time of sample
136 - mem: type of memory access for the data at the time of sample
137 - snoop: type of snoop (if any) for the data at the time of sample
138 - dcacheline: the cacheline the data address is on at the time of sample
140 And default sort keys are changed to local_weight, mem, sym, dso,
141 symbol_daddr, dso_daddr, snoop, tlb, locked, see '--mem-mode'.
145 A regex filter to identify parent. The parent is a caller of this
146 function and searched through the callchain, thus it requires callchain
147 information recorded. The pattern is in the exteneded regex format and
148 defaults to "\^sys_|^do_page_fault", see '--sort parent'.
152 Only display entries with parent-match.
155 --column-widths=<width[,width...]>::
156 Force each column width to the provided list, for large terminal
157 readability. 0 means no limit (default behavior).
161 Use a special separator character and don't pad with spaces, replacing
162 all occurrences of this separator in symbol names (and other output)
163 with a '.' character, that thus it's the only non valid separator.
167 Dump raw trace in ASCII.
169 -g [type,min[,limit],order[,key][,branch]]::
171 Display call chains using type, min percent threshold, optional print
174 - flat: single column, linear exposure of call chains.
175 - graph: use a graph tree, displaying absolute overhead rates.
176 - fractal: like graph, but displays relative rates. Each branch of
177 the tree is considered as a new profiled object. +
180 - callee: callee based call graph.
181 - caller: inverted caller based call graph.
184 - function: compare on functions
185 - address: compare on individual code addresses
188 - branch: include last branch information in callgraph
189 when available. Usually more convenient to use --branch-history
192 Default: fractal,0.5,callee,function.
195 Accumulate callchain of children to parent entry so that then can
196 show up in the output. The output will have a new "Children" column
197 and will be sorted on the data. It requires callchains are recorded.
198 See the `overhead calculation' section for more details.
201 Set the stack depth limit when parsing the callchain, anything
202 beyond the specified depth will be ignored. This is a trade-off
203 between information loss and faster processing especially for
204 workloads that can have a very long callchain stack.
210 alias for inverted caller based call graph.
212 --ignore-callees=<regex>::
213 Ignore callees of the function(s) matching the given regex.
214 This has the effect of collecting the callers of each such
215 function into one place in the call-graph tree.
218 Pretty printing style. key: normal, raw
220 --stdio:: Use the stdio interface.
222 --tui:: Use the TUI interface, that is integrated with annotate and allows
223 zooming into DSOs or threads, among other features. Use of --tui
224 requires a tty, if one is not present, as when piping to other
225 commands, the stdio interface is used.
227 --gtk:: Use the GTK2 interface.
238 Load module symbols. WARNING: This should only be used with -k and
243 Don't complain, do it.
245 --symfs=<directory>::
246 Look for files with symbols relative to this directory.
249 --cpu:: Only report samples for the list of CPUs provided. Multiple CPUs can
250 be provided as a comma-separated list with no space: 0,1. Ranges of
251 CPUs are specified with -: 0-2. Default is to report samples on all
255 --disassembler-style=:: Set disassembler style for objdump.
258 Interleave source code with assembly code. Enabled by default,
259 disable with --no-source.
262 Show raw instruction encoding of assembly instructions.
264 --show-total-period:: Show a column with the sum of periods.
268 Display extended information about the perf.data file. This adds
269 information which may be very large and thus may clutter the display.
270 It currently includes: cpu and numa topology of the host system.
274 Use the addresses of sampled taken branches instead of the instruction
275 address to build the histograms. To generate meaningful output, the
276 perf.data file must have been obtained using perf record -b or
277 perf record --branch-filter xxx where xxx is a branch filter option.
278 perf report is able to auto-detect whether a perf.data file contains
279 branch stacks and it will automatically switch to the branch view mode,
280 unless --no-branch-stack is used.
283 Add the addresses of sampled taken branches to the callstack.
284 This allows to examine the path the program took to each sample.
285 The data collection must have used -b (or -j) and -g.
288 Path to objdump binary.
291 Show event group information together.
294 Demangle symbol names to human readable form. It's enabled by default,
295 disable with --no-demangle.
298 Demangle kernel symbol names to human readable form (for C++ kernels).
301 Use the data addresses of samples in addition to instruction addresses
302 to build the histograms. To generate meaningful output, the perf.data
303 file must have been obtained using perf record -d -W and using a
304 special event -e cpu/mem-loads/ or -e cpu/mem-stores/. See
305 'perf mem' for simpler access.
308 Do not show entries which have an overhead under that percent.
312 Determine how to display the overhead percentage of filtered entries.
313 Filters can be applied by --comms, --dsos and/or --symbols options and
314 Zoom operations on the TUI (thread, dso, etc).
316 "relative" means it's relative to filtered entries only so that the
317 sum of shown entries will be always 100%. "absolute" means it retains
318 the original value before and after the filter is applied.
321 Show header information in the perf.data file. This includes
322 various information like hostname, OS and perf version, cpu/mem
323 info, perf command line, event list and so on. Currently only
324 --stdio output supports this feature.
327 Show only perf.data header (forces --stdio).
330 Options for decoding instruction tracing data. The options are:
332 i synthesize instructions events
333 b synthesize branches events
334 c synthesize branches events (calls only)
335 r synthesize branches events (returns only)
336 x synthesize transactions events
337 e synthesize error events
339 g synthesize a call chain (use with i or x)
341 The default is all events i.e. the same as --itrace=ibxe
343 In addition, the period (default 100000) for instructions events
344 can be specified in units of:
350 ns nanoseconds (default)
352 Also the call chain size (default 16, max. 1024) for instructions or
353 transactions events can be specified.
355 To disable decoding entirely, use --no-itrace.
358 include::callchain-overhead-calculation.txt[]
362 linkperf:perf-stat[1], linkperf:perf-annotate[1]