1PERF-REPORT(1)                    perf Manual                   PERF-REPORT(1)
2
3
4

NAME

6       perf-report - Read perf.data (created by perf record) and display the
7       profile
8

SYNOPSIS

10       perf report [-i <file> | --input=file]
11

DESCRIPTION

13       This command displays the performance counter profile information
14       recorded via perf record.
15

OPTIONS

17       -i, --input=
18           Input file name. (default: perf.data unless stdin is a fifo)
19
20       -v, --verbose
21           Be more verbose. (show symbol address, etc)
22
23       -q, --quiet
24           Do not show any message. (Suppress -v)
25
26       -n, --show-nr-samples
27           Show the number of samples for each symbol
28
29       --show-cpu-utilization
30           Show sample percentage for different cpu modes.
31
32       -T, --threads
33           Show per-thread event counters. The input data file should be
34           recorded with -s option.
35
36       -c, --comms=
37           Only consider symbols in these comms. CSV that understands
38           file://filename entries. This option will affect the percentage of
39           the overhead column. See --percentage for more info.
40
41       --pid=
42           Only show events for given process ID (comma separated list).
43
44       --tid=
45           Only show events for given thread ID (comma separated list).
46
47       -d, --dsos=
48           Only consider symbols in these dsos. CSV that understands
49           file://filename entries. This option will affect the percentage of
50           the overhead column. See --percentage for more info.
51
52       -S, --symbols=
53           Only consider these symbols. CSV that understands file://filename
54           entries. This option will affect the percentage of the overhead
55           column. See --percentage for more info.
56
57       --symbol-filter=
58           Only show symbols that match (partially) with this filter.
59
60       -U, --hide-unresolved
61           Only display entries resolved to a symbol.
62
63       -s, --sort=
64           Sort histogram entries by given key(s) - multiple keys can be
65           specified in CSV format. Following sort keys are available: pid,
66           comm, dso, symbol, parent, cpu, socket, srcline, weight,
67           local_weight, cgroup_id.
68
69               Each key has following meaning:
70
71           ·   comm: command (name) of the task which can be read via
72               /proc/<pid>/comm
73
74           ·   pid: command and tid of the task
75
76           ·   dso: name of library or module executed at the time of sample
77
78           ·   dso_size: size of library or module executed at the time of
79               sample
80
81           ·   symbol: name of function executed at the time of sample
82
83           ·   symbol_size: size of function executed at the time of sample
84
85           ·   parent: name of function matched to the parent regex filter.
86               Unmatched entries are displayed as "[other]".
87
88           ·   cpu: cpu number the task ran at the time of sample
89
90           ·   socket: processor socket number the task ran at the time of
91               sample
92
93           ·   srcline: filename and line number executed at the time of
94               sample. The DWARF debugging info must be provided.
95
96           ·   srcfile: file name of the source file of the samples. Requires
97               dwarf information.
98
99           ·   weight: Event specific weight, e.g. memory latency or
100               transaction abort cost. This is the global weight.
101
102           ·   local_weight: Local weight version of the weight above.
103
104           ·   cgroup_id: ID derived from cgroup namespace device and inode
105               numbers.
106
107           ·   transaction: Transaction abort flags.
108
109           ·   overhead: Overhead percentage of sample
110
111           ·   overhead_sys: Overhead percentage of sample running in system
112               mode
113
114           ·   overhead_us: Overhead percentage of sample running in user mode
115
116           ·   overhead_guest_sys: Overhead percentage of sample running in
117               system mode on guest machine
118
119           ·   overhead_guest_us: Overhead percentage of sample running in
120               user mode on guest machine
121
122           ·   sample: Number of sample
123
124           ·   period: Raw number of event count of sample
125
126           ·   time: Separate the samples by time stamp with the resolution
127               specified by --time-quantum (default 100ms). Specify with
128               overhead and before it.
129
130                   By default, comm, dso and symbol keys are used.
131                   (i.e. --sort comm,dso,symbol)
132
133                   If --branch-stack option is used, following sort keys are also
134                   available:
135
136           ·   dso_from: name of library or module branched from
137
138           ·   dso_to: name of library or module branched to
139
140           ·   symbol_from: name of function branched from
141
142           ·   symbol_to: name of function branched to
143
144           ·   srcline_from: source file and line branched from
145
146           ·   srcline_to: source file and line branched to
147
148           ·   mispredict: "N" for predicted branch, "Y" for mispredicted
149               branch
150
151           ·   in_tx: branch in TSX transaction
152
153           ·   abort: TSX transaction abort.
154
155           ·   cycles: Cycles in basic block
156
157                   And default sort keys are changed to comm, dso_from, symbol_from, dso_to
158                   and symbol_to, see '--branch-stack'.
159
160                   When the sort key symbol is specified, columns "IPC" and "IPC Coverage"
161                   are enabled automatically. Column "IPC" reports the average IPC per function
162                   and column "IPC coverage" reports the percentage of instructions with
163                   sampled IPC in this function. IPC means Instruction Per Cycle. If it's low,
164                   it indicates there may be a performance bottleneck when the function is
165                   executed, such as a memory access bottleneck. If a function has high overhead
166                   and low IPC, it's worth further analyzing it to optimize its performance.
167
168                   If the --mem-mode option is used, the following sort keys are also available
169                   (incompatible with --branch-stack):
170                   symbol_daddr, dso_daddr, locked, tlb, mem, snoop, dcacheline.
171
172           ·   symbol_daddr: name of data symbol being executed on at the time
173               of sample
174
175           ·   dso_daddr: name of library or module containing the data being
176               executed on at the time of the sample
177
178           ·   locked: whether the bus was locked at the time of the sample
179
180           ·   tlb: type of tlb access for the data at the time of the sample
181
182           ·   mem: type of memory access for the data at the time of the
183               sample
184
185           ·   snoop: type of snoop (if any) for the data at the time of the
186               sample
187
188           ·   dcacheline: the cacheline the data address is on at the time of
189               the sample
190
191           ·   phys_daddr: physical address of data being executed on at the
192               time of sample
193
194                   And the default sort keys are changed to local_weight, mem, sym, dso,
195                   symbol_daddr, dso_daddr, snoop, tlb, locked, see '--mem-mode'.
196
197                   If the data file has tracepoint event(s), following (dynamic) sort keys
198                   are also available:
199                   trace, trace_fields, [<event>.]<field>[/raw]
200
201           ·   trace: pretty printed trace output in a single column
202
203           ·   trace_fields: fields in tracepoints in separate columns
204
205           ·   <field name>: optional event and field name for a specific
206               field
207
208                   The last form consists of event and field names.  If event name is
209                   omitted, it searches all events for matching field name.  The matched
210                   field will be shown only for the event has the field.  The event name
211                   supports substring match so user doesn't need to specify full subsystem
212                   and event name everytime.  For example, 'sched:sched_switch' event can
213                   be shortened to 'switch' as long as it's not ambiguous.  Also event can
214                   be specified by its index (starting from 1) preceded by the '%'.
215                   So '%1' is the first event, '%2' is the second, and so on.
216
217                   The field name can have '/raw' suffix which disables pretty printing
218                   and shows raw field value like hex numbers.  The --raw-trace option
219                   has the same effect for all dynamic sort keys.
220
221                   The default sort keys are changed to 'trace' if all events in the data
222                   file are tracepoint.
223
224       -F, --fields=
225           Specify output field - multiple keys can be specified in CSV
226           format. Following fields are available: overhead, overhead_sys,
227           overhead_us, overhead_children, sample and period. Also it can
228           contain any sort key(s).
229
230               By default, every sort keys not specified in -F will be appended
231               automatically.
232
233               If the keys starts with a prefix '+', then it will append the specified
234               field(s) to the default field order. For example: perf report -F +period,sample.
235
236       -p, --parent=<regex>
237           A regex filter to identify parent. The parent is a caller of this
238           function and searched through the callchain, thus it requires
239           callchain information recorded. The pattern is in the extended
240           regex format and defaults to "^sys_|^do_page_fault", see --sort
241           parent.
242
243       -x, --exclude-other
244           Only display entries with parent-match.
245
246       -w, --column-widths=<width[,width...]>
247           Force each column width to the provided list, for large terminal
248           readability. 0 means no limit (default behavior).
249
250       -t, --field-separator=
251           Use a special separator character and don’t pad with spaces,
252           replacing all occurrences of this separator in symbol names (and
253           other output) with a .  character, that thus it’s the only non
254           valid separator.
255
256       -D, --dump-raw-trace
257           Dump raw trace in ASCII.
258
259       -g,
260       --call-graph=<print_type,threshold[,print_limit],order,sort_key[,branch],value>
261           Display call chains using type, min percent threshold, print limit,
262           call order, sort key, optional branch and value. Note that ordering
263           is not fixed so any parameter can be given in an arbitrary order.
264           One exception is the print_limit which should be preceded by
265           threshold.
266
267               print_type can be either:
268               - flat: single column, linear exposure of call chains.
269               - graph: use a graph tree, displaying absolute overhead rates. (default)
270               - fractal: like graph, but displays relative rates. Each branch of
271                        the tree is considered as a new profiled object.
272               - folded: call chains are displayed in a line, separated by semicolons
273               - none: disable call chain display.
274
275               threshold is a percentage value which specifies a minimum percent to be
276               included in the output call graph.  Default is 0.5 (%).
277
278               print_limit is only applied when stdio interface is used.  It's to limit
279               number of call graph entries in a single hist entry.  Note that it needs
280               to be given after threshold (but not necessarily consecutive).
281               Default is 0 (unlimited).
282
283               order can be either:
284               - callee: callee based call graph.
285               - caller: inverted caller based call graph.
286               Default is 'caller' when --children is used, otherwise 'callee'.
287
288               sort_key can be:
289               - function: compare on functions (default)
290               - address: compare on individual code addresses
291               - srcline: compare on source filename and line number
292
293               branch can be:
294               - branch: include last branch information in callgraph when available.
295                         Usually more convenient to use --branch-history for this.
296
297               value can be:
298               - percent: display overhead percent (default)
299               - period: display event period
300               - count: display event count
301
302       --children
303           Accumulate callchain of children to parent entry so that then can
304           show up in the output. The output will have a new "Children" column
305           and will be sorted on the data. It requires callchains are
306           recorded. See the ‘overhead calculation’ section for more details.
307           Enabled by default, disable with --no-children.
308
309       --max-stack
310           Set the stack depth limit when parsing the callchain, anything
311           beyond the specified depth will be ignored. This is a trade-off
312           between information loss and faster processing especially for
313           workloads that can have a very long callchain stack. Note that when
314           using the --itrace option the synthesized callchain size will
315           override this value if the synthesized callchain size is bigger.
316
317               Default: 127
318
319       -G, --inverted
320           alias for inverted caller based call graph.
321
322       --ignore-callees=<regex>
323           Ignore callees of the function(s) matching the given regex. This
324           has the effect of collecting the callers of each such function into
325           one place in the call-graph tree.
326
327       --pretty=<key>
328           Pretty printing style. key: normal, raw
329
330       --stdio
331           Use the stdio interface.
332
333       --stdio-color
334           always, never or auto, allowing configuring color output via the
335           command line, in addition to via "color.ui" .perfconfig. Use
336           --stdio-color always to generate color even when redirecting to a
337           pipe or file. Using just --stdio-color is equivalent to using
338           always.
339
340       --tui
341           Use the TUI interface, that is integrated with annotate and allows
342           zooming into DSOs or threads, among other features. Use of --tui
343           requires a tty, if one is not present, as when piping to other
344           commands, the stdio interface is used.
345
346       --gtk
347           Use the GTK2 interface.
348
349       -k, --vmlinux=<file>
350           vmlinux pathname
351
352       --ignore-vmlinux
353           Ignore vmlinux files.
354
355       --kallsyms=<file>
356           kallsyms pathname
357
358       -m, --modules
359           Load module symbols. WARNING: This should only be used with -k and
360           a LIVE kernel.
361
362       -f, --force
363           Don’t do ownership validation.
364
365       --symfs=<directory>
366           Look for files with symbols relative to this directory.
367
368       -C, --cpu
369           Only report samples for the list of CPUs provided. Multiple CPUs
370           can be provided as a comma-separated list with no space: 0,1.
371           Ranges of CPUs are specified with -: 0-2. Default is to report
372           samples on all CPUs.
373
374       -M, --disassembler-style=
375           Set disassembler style for objdump.
376
377       --source
378           Interleave source code with assembly code. Enabled by default,
379           disable with --no-source.
380
381       --asm-raw
382           Show raw instruction encoding of assembly instructions.
383
384       --show-total-period
385           Show a column with the sum of periods.
386
387       -I, --show-info
388           Display extended information about the perf.data file. This adds
389           information which may be very large and thus may clutter the
390           display. It currently includes: cpu and numa topology of the host
391           system.
392
393       -b, --branch-stack
394           Use the addresses of sampled taken branches instead of the
395           instruction address to build the histograms. To generate meaningful
396           output, the perf.data file must have been obtained using perf
397           record -b or perf record --branch-filter xxx where xxx is a branch
398           filter option. perf report is able to auto-detect whether a
399           perf.data file contains branch stacks and it will automatically
400           switch to the branch view mode, unless --no-branch-stack is used.
401
402       --branch-history
403           Add the addresses of sampled taken branches to the callstack. This
404           allows to examine the path the program took to each sample. The
405           data collection must have used -b (or -j) and -g.
406
407       --objdump=<path>
408           Path to objdump binary.
409
410       --group
411           Show event group information together. It forces group output also
412           if there are no groups defined in data file.
413
414       --demangle
415           Demangle symbol names to human readable form. It’s enabled by
416           default, disable with --no-demangle.
417
418       --demangle-kernel
419           Demangle kernel symbol names to human readable form (for C++
420           kernels).
421
422       --mem-mode
423           Use the data addresses of samples in addition to instruction
424           addresses to build the histograms. To generate meaningful output,
425           the perf.data file must have been obtained using perf record -d -W
426           and using a special event -e cpu/mem-loads/p or -e
427           cpu/mem-stores/p. See perf mem for simpler access.
428
429       --percent-limit
430           Do not show entries which have an overhead under that percent.
431           (Default: 0). Note that this option also sets the percent limit
432           (threshold) of callchains. However the default value of callchain
433           threshold is different than the default value of hist entries.
434           Please see the --call-graph option for details.
435
436       --percentage
437           Determine how to display the overhead percentage of filtered
438           entries. Filters can be applied by --comms, --dsos and/or --symbols
439           options and Zoom operations on the TUI (thread, dso, etc).
440
441               "relative" means it's relative to filtered entries only so that the
442               sum of shown entries will be always 100%.  "absolute" means it retains
443               the original value before and after the filter is applied.
444
445       --header
446           Show header information in the perf.data file. This includes
447           various information like hostname, OS and perf version, cpu/mem
448           info, perf command line, event list and so on. Currently only
449           --stdio output supports this feature.
450
451       --header-only
452           Show only perf.data header (forces --stdio).
453
454       --time
455           Only analyze samples within given time window: <start>,<stop>.
456           Times have the format seconds.nanoseconds. If start is not given
457           (i.e. time string is ,x.y) then analysis starts at the beginning of
458           the file. If stop time is not given (i.e. time string is x.y,) then
459           analysis goes to end of file. Multiple ranges can be separated by
460           spaces, which requires the argument to be quoted e.g. --time
461           "1234.567,1234.789 1235,"
462
463               Also support time percent with multiple time ranges. Time string is
464               'a%/n,b%/m,...' or 'a%-b%,c%-%d,...'.
465
466               For example:
467               Select the second 10% time slice:
468
469               perf report --time 10%/2
470
471               Select from 0% to 10% time slice:
472
473               perf report --time 0%-10%
474
475               Select the first and second 10% time slices:
476
477               perf report --time 10%/1,10%/2
478
479               Select from 0% to 10% and 30% to 40% slices:
480
481               perf report --time 0%-10%,30%-40%
482
483       --itrace
484           Options for decoding instruction tracing data. The options are:
485
486               i       synthesize instructions events
487               b       synthesize branches events
488               c       synthesize branches events (calls only)
489               r       synthesize branches events (returns only)
490               x       synthesize transactions events
491               w       synthesize ptwrite events
492               p       synthesize power events
493               e       synthesize error events
494               d       create a debug log
495               g       synthesize a call chain (use with i or x)
496               l       synthesize last branch entries (use with i or x)
497               s       skip initial number of events
498
499               The default is all events i.e. the same as --itrace=ibxwpe,
500               except for perf script where it is --itrace=ce
501
502               In addition, the period (default 100000, except for perf script where it is 1)
503               for instructions events can be specified in units of:
504
505               i       instructions
506               t       ticks
507               ms      milliseconds
508               us      microseconds
509               ns      nanoseconds (default)
510
511               Also the call chain size (default 16, max. 1024) for instructions or
512               transactions events can be specified.
513
514               Also the number of last branch entries (default 64, max. 1024) for
515               instructions or transactions events can be specified.
516
517               It is also possible to skip events generated (instructions, branches, transactions,
518               ptwrite, power) at the beginning. This is useful to ignore initialization code.
519
520               --itrace=i0nss1000000
521
522               skips the first million instructions.
523
524               To disable decoding entirely, use --no-itrace.
525
526       --full-source-path
527           Show the full path for source files for srcline output.
528
529       --show-ref-call-graph
530           When multiple events are sampled, it may not be needed to collect
531           callgraphs for all of them. The sample sites are usually nearby,
532           and it’s enough to collect the callgraphs on a reference event. So
533           user can use "call-graph=no" event modifier to disable callgraph
534           for other events to reduce the overhead. However, perf report
535           cannot show callgraphs for the event which disable the callgraph.
536           This option extends the perf report to show reference callgraphs,
537           which collected by reference event, in no callgraph event.
538
539       --socket-filter
540           Only report the samples on the processor socket that match with
541           this filter
542
543       --samples=N
544           Save N individual samples for each histogram entry to show context
545           in perf report tui browser.
546
547       --raw-trace
548           When displaying traceevent output, do not use print fmt or plugins.
549
550       --hierarchy
551           Enable hierarchical output.
552
553       --inline
554           If a callgraph address belongs to an inlined function, the inline
555           stack will be printed. Each entry is function name or file/line.
556           Enabled by default, disable with --no-inline.
557
558       --mmaps
559           Show --tasks output plus mmap information in a format similar to
560           /proc/<PID>/maps.
561
562               Please note that not all mmaps are stored, options affecting which ones
563               are include 'perf record --data', for instance.
564
565       --ns
566           Show time stamps in nanoseconds.
567
568       --stats
569           Display overall events statistics without any further processing.
570           (like the one at the end of the perf report -D command)
571
572       --tasks
573           Display monitored tasks stored in perf data. Displaying
574           pid/tid/ppid plus the command string aligned to distinguish parent
575           and child tasks.
576
577       --percent-type
578           Set annotation percent type from following choices: global-period,
579           local-period, global-hits, local-hits
580
581               The local/global keywords set if the percentage is computed
582               in the scope of the function (local) or the whole data (global).
583               The period/hits keywords set the base the percentage is computed
584               on - the samples period or the number of samples (hits).
585
586       --time-quantum
587           Configure time quantum for time sort key. Default 100ms. Accepts s,
588           us, ms, ns units.
589

OVERHEAD CALCULATION

591       The overhead can be shown in two columns as Children and Self when perf
592       collects callchains. The self overhead is simply calculated by adding
593       all period values of the entry - usually a function (symbol). This is
594       the value that perf shows traditionally and sum of all the self
595       overhead values should be 100%.
596
597       The children overhead is calculated by adding all period values of the
598       child functions so that it can show the total overhead of the higher
599       level functions even if they don’t directly execute much. Children here
600       means functions that are called from another (parent) function.
601
602       It might be confusing that the sum of all the children overhead values
603       exceeds 100% since each of them is already an accumulation of self
604       overhead of its child functions. But with this enabled, users can find
605       which function has the most overhead even if samples are spread over
606       the children.
607
608       Consider the following example; there are three functions like below.
609
610
611           .ft C
612           void foo(void) {
613               /* do something */
614           }
615
616           void bar(void) {
617               /* do something */
618               foo();
619           }
620
621           int main(void) {
622               bar()
623               return 0;
624           }
625           .ft
626
627
628       In this case foo is a child of bar, and bar is an immediate child of
629       main so foo also is a child of main. In other words, main is a parent
630       of foo and bar, and bar is a parent of foo.
631
632       Suppose all samples are recorded in foo and bar only. When it’s
633       recorded with callchains the output will show something like below in
634       the usual (self-overhead-only) output of perf report:
635
636
637           .ft C
638           Overhead  Symbol
639           ........  .....................
640             60.00%  foo
641                     |
642                     --- foo
643                         bar
644                         main
645                         __libc_start_main
646
647             40.00%  bar
648                     |
649                     --- bar
650                         main
651                         __libc_start_main
652           .ft
653
654
655       When the --children option is enabled, the self overhead values of
656       child functions (i.e. foo and bar) are added to the parents to
657       calculate the children overhead. In this case the report could be
658       displayed as:
659
660
661           .ft C
662           Children      Self  Symbol
663           ........  ........  ....................
664            100.00%     0.00%  __libc_start_main
665                     |
666                     --- __libc_start_main
667
668            100.00%     0.00%  main
669                     |
670                     --- main
671                         __libc_start_main
672
673            100.00%    40.00%  bar
674                     |
675                     --- bar
676                         main
677                         __libc_start_main
678
679             60.00%    60.00%  foo
680                     |
681                     --- foo
682                         bar
683                         main
684                         __libc_start_main
685           .ft
686
687
688       In the above output, the self overhead of foo (60%) was add to the
689       children overhead of bar, main and __libc_start_main. Likewise, the
690       self overhead of bar (40%) was added to the children overhead of main
691       and \_\_libc_start_main.
692
693       So \_\_libc_start_main and main are shown first since they have same
694       (100%) children overhead (even though they have zero self overhead) and
695       they are the parents of foo and bar.
696
697       Since v3.16 the children overhead is shown by default and the output is
698       sorted by its values. The children overhead is disabled by specifying
699       --no-children option on the command line or by adding report.children =
700       false or top.children = false in the perf config file.
701

SEE ALSO

703       perf-stat(1), perf-annotate(1), perf-record(1)
704
705
706
707perf                              11/12/2019                    PERF-REPORT(1)
Impressum