blob: 3a35a8523e6df03aa8a5e2a0777c540085403627 [file] [log] [blame]
Ingo Molnare33e0a42009-04-20 15:58:01 +02001perf-record(1)
Ingo Molnarc1c23652009-05-30 12:38:51 +02002==============
Ingo Molnare33e0a42009-04-20 15:58:01 +02003
4NAME
5----
Ingo Molnar23ac9cb2009-05-27 09:33:18 +02006perf-record - Run a command and record its profile into perf.data
Ingo Molnare33e0a42009-04-20 15:58:01 +02007
8SYNOPSIS
9--------
10[verse]
11'perf record' [-e <EVENT> | --event=EVENT] [-l] [-a] <command>
Mike Galbraith9e0967532009-05-28 16:25:34 +020012'perf record' [-e <EVENT> | --event=EVENT] [-l] [-a] -- <command> [<options>]
Ingo Molnare33e0a42009-04-20 15:58:01 +020013
14DESCRIPTION
15-----------
16This command runs a command and gathers a performance counter profile
Ingo Molnar23ac9cb2009-05-27 09:33:18 +020017from it, into perf.data - without displaying anything.
Ingo Molnare33e0a42009-04-20 15:58:01 +020018
19This file can then be inspected later on, using 'perf report'.
20
21
22OPTIONS
23-------
24<command>...::
25 Any command you can specify in a shell.
26
27-e::
28--event=::
Frederic Weisbecker1b290d62009-11-23 15:42:35 +010029 Select the PMU event. Selection can be:
Ingo Molnare33e0a42009-04-20 15:58:01 +020030
Frederic Weisbecker1b290d62009-11-23 15:42:35 +010031 - a symbolic event name (use 'perf list' to list all events)
32
33 - a raw PMU event (eventsel+umask) in the form of rNNN where NNN is a
34 hexadecimal event descriptor.
35
36 - a hardware breakpoint event in the form of '\mem:addr[:access]'
37 where addr is the address in memory you want to break in.
38 Access is the memory access type (read, write, execute) it can
39 be passed as follows: '\mem:addr[:[r][w][x]]'.
40 If you want to profile read-write accesses in 0x1000, just set
41 'mem:0x1000:rw'.
Shawn Bohrer08dbd7e2010-11-30 19:57:16 -060042
43--filter=<filter>::
44 Event filter.
45
Ingo Molnare33e0a42009-04-20 15:58:01 +020046-a::
Shawn Bohrer08dbd7e2010-11-30 19:57:16 -060047--all-cpus::
48 System-wide collection from all CPUs.
Ingo Molnare33e0a42009-04-20 15:58:01 +020049
50-l::
Arnaldo Carvalho de Melo386c0b72009-08-05 10:04:53 -030051 Scale counter values.
52
53-p::
54--pid=::
David Ahernb52956c2012-02-08 09:32:52 -070055 Record events on existing process ID (comma separated list).
Shawn Bohrer08dbd7e2010-11-30 19:57:16 -060056
57-t::
58--tid=::
David Ahernb52956c2012-02-08 09:32:52 -070059 Record events on existing thread ID (comma separated list).
Adrian Hunter69e7e5b2013-11-18 11:55:57 +020060 This option also disables inheritance by default. Enable it by adding
61 --inherit.
Arnaldo Carvalho de Melo386c0b72009-08-05 10:04:53 -030062
Arnaldo Carvalho de Melo0d37aa32012-01-19 14:08:15 -020063-u::
64--uid=::
65 Record events in threads owned by uid. Name or number.
66
Arnaldo Carvalho de Melo386c0b72009-08-05 10:04:53 -030067-r::
68--realtime=::
69 Collect data with this RT SCHED_FIFO priority.
Jiri Olsa563aecb2013-06-05 13:35:06 +020070
Arnaldo Carvalho de Melo509051e2014-01-14 17:52:14 -030071--no-buffering::
Kirill Smelkovacac03f2011-01-12 17:59:36 +030072 Collect data without buffering.
Arnaldo Carvalho de Melo386c0b72009-08-05 10:04:53 -030073
Arnaldo Carvalho de Melo386c0b72009-08-05 10:04:53 -030074-c::
75--count=::
76 Event period to sample.
77
78-o::
79--output=::
80 Output file name.
81
82-i::
Stephane Eranian2e6cdf92010-05-12 10:40:01 +020083--no-inherit::
84 Child tasks do not inherit counters.
Arnaldo Carvalho de Melo386c0b72009-08-05 10:04:53 -030085-F::
86--freq=::
87 Profile at this frequency.
88
89-m::
90--mmap-pages=::
Jiri Olsa27050f52013-09-01 12:36:13 +020091 Number of mmap data pages (must be a power of two) or size
92 specification with appended unit character - B/K/M/G. The
93 size is rounded up to have nearest pages power of two value.
Arnaldo Carvalho de Melo386c0b72009-08-05 10:04:53 -030094
95-g::
Jiri Olsa09b0fd42013-10-26 16:25:33 +020096 Enables call-graph (stack chain/backtrace) recording.
97
Arnaldo Carvalho de Melo386c0b72009-08-05 10:04:53 -030098--call-graph::
Jiri Olsa09b0fd42013-10-26 16:25:33 +020099 Setup and enable call-graph (stack chain/backtrace) recording,
100 implies -g.
101
102 Allows specifying "fp" (frame pointer) or "dwarf"
103 (DWARF's CFI - Call Frame Information) as the method to collect
104 the information used to show the call graphs.
105
106 In some systems, where binaries are build with gcc
107 --fomit-frame-pointer, using the "fp" method will produce bogus
108 call graphs, using "dwarf", if available (perf tools linked to
109 the libunwind library) should be used instead.
Arnaldo Carvalho de Melo386c0b72009-08-05 10:04:53 -0300110
Arnaldo Carvalho de Melob44308f2010-10-26 15:20:09 -0200111-q::
112--quiet::
113 Don't print any message, useful for scripting.
114
Arnaldo Carvalho de Melo386c0b72009-08-05 10:04:53 -0300115-v::
116--verbose::
117 Be more verbose (show counter open errors, etc).
118
119-s::
120--stat::
121 Per thread counts.
122
123-d::
124--data::
125 Sample addresses.
126
Arnaldo Carvalho de Melo9c90a612010-12-02 10:25:28 -0200127-T::
128--timestamp::
129 Sample timestamps. Use it with 'perf report -D' to see the timestamps,
130 for instance.
131
Arnaldo Carvalho de Melo386c0b72009-08-05 10:04:53 -0300132-n::
133--no-samples::
134 Don't sample.
Ingo Molnare33e0a42009-04-20 15:58:01 +0200135
Frederic Weisbeckerec7ba4e2009-08-31 03:32:03 +0200136-R::
137--raw-samples::
Frederic Weisbeckerbdef3b02010-04-14 20:05:17 +0200138Collect raw sample records from all opened counters (default for tracepoint counters).
Frederic Weisbeckerec7ba4e2009-08-31 03:32:03 +0200139
Stephane Eranianc45c6ea2010-05-28 12:00:01 +0200140-C::
141--cpu::
Shawn Bohrer08dbd7e2010-11-30 19:57:16 -0600142Collect samples only on the list of CPUs provided. Multiple CPUs can be provided as a
143comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2.
Stephane Eranianc45c6ea2010-05-28 12:00:01 +0200144In per-thread mode with inheritance mode on (default), samples are captured only when
145the thread executes on the designated CPUs. Default is to monitor all CPUs.
146
Stephane Eraniana1ac1d32010-06-17 11:39:01 +0200147-N::
148--no-buildid-cache::
149Do not update the builid cache. This saves some overhead in situations
150where the information in the perf.data file (which includes buildids)
151is sufficient.
152
Stephane Eranian023695d2011-02-14 11:20:01 +0200153-G name,...::
154--cgroup name,...::
155monitor only in the container (cgroup) called "name". This option is available only
156in per-cpu mode. The cgroup filesystem must be mounted. All threads belonging to
157container "name" are monitored when they run on the monitored CPUs. Multiple cgroups
158can be provided. Each cgroup is applied to the corresponding event, i.e., first cgroup
159to first event, second cgroup to second event and so on. It is possible to provide
160an empty cgroup (monitor all the time) using, e.g., -G foo,,bar. Cgroups must have
161corresponding events, i.e., they always refer to events defined earlier on the command
162line.
163
Roberto Agostino Vitillobdfebd82012-02-09 23:21:02 +0100164-b::
Stephane Eraniana5aabda2012-03-08 23:47:45 +0100165--branch-any::
166Enable taken branch stack sampling. Any type of taken branch may be sampled.
167This is a shortcut for --branch-filter any. See --branch-filter for more infos.
168
169-j::
170--branch-filter::
Roberto Agostino Vitillobdfebd82012-02-09 23:21:02 +0100171Enable taken branch stack sampling. Each sample captures a series of consecutive
172taken branches. The number of branches captured with each sample depends on the
173underlying hardware, the type of branches of interest, and the executed code.
174It is possible to select the types of branches captured by enabling filters. The
175following filters are defined:
176
Stephane Eraniana5aabda2012-03-08 23:47:45 +0100177 - any: any type of branches
Roberto Agostino Vitillobdfebd82012-02-09 23:21:02 +0100178 - any_call: any function call or system call
179 - any_ret: any function return or system call return
Anshuman Khandual2e49a942012-05-18 14:16:50 +0530180 - ind_call: any indirect branch
Roberto Agostino Vitillobdfebd82012-02-09 23:21:02 +0100181 - u: only when the branch target is at the user level
182 - k: only when the branch target is in the kernel
183 - hv: only when the target is at the hypervisor level
Andi Kleen0126d492013-09-20 07:40:42 -0700184 - in_tx: only when the target is in a hardware transaction
185 - no_tx: only when the target is not in a hardware transaction
186 - abort_tx: only when the target is a hardware transaction abort
Roberto Agostino Vitillobdfebd82012-02-09 23:21:02 +0100187
188+
Stephane Eraniana5aabda2012-03-08 23:47:45 +0100189The option requires at least one branch type among any, any_call, any_ret, ind_call.
Masanari Iida9c768202012-11-30 14:10:25 +0900190The privilege levels may be omitted, in which case, the privilege levels of the associated
Stephane Eraniana5aabda2012-03-08 23:47:45 +0100191event are applied to the branch filter. Both kernel (k) and hypervisor (hv) privilege
192levels are subject to permissions. When sampling on multiple events, branch stack sampling
193is enabled for all the sampling events. The sampled branch type is the same for all events.
194The various filters must be specified as a comma separated list: --branch-filter any_ret,u,k
195Note that this feature may not be available on all processors.
Roberto Agostino Vitillobdfebd82012-02-09 23:21:02 +0100196
Andi Kleen05484292013-01-24 16:10:29 +0100197--weight::
198Enable weightened sampling. An additional weight is recorded per sample and can be
199displayed with the weight and local_weight sort keys. This currently works for TSX
200abort events and some memory events in precise mode on modern Intel CPUs.
201
Andi Kleen475eeab2013-09-20 07:40:43 -0700202--transaction::
203Record transaction flags for transaction related events.
204
Adrian Hunter3aa59392013-11-15 15:52:29 +0200205--per-thread::
206Use per-thread mmaps. By default per-cpu mmaps are created. This option
207overrides that and uses per-thread mmaps. A side-effect of that is that
208inheritance is automatically disabled. --per-thread is ignored with a warning
209if combined with -a or -C options.
Adrian Hunter539e6bb2013-11-01 15:51:34 +0200210
Andi Kleen6619a532014-01-11 13:38:27 -0800211--initial-delay msecs::
212After starting the program, wait msecs before measuring. This is useful to
213filter out the startup phase of the program, which is often very different.
214
Ingo Molnare33e0a42009-04-20 15:58:01 +0200215SEE ALSO
216--------
Thomas Gleixner386b05e2009-06-06 14:56:33 +0200217linkperf:perf-stat[1], linkperf:perf-list[1]