blob: 90c3558c2c12fd5b2420f7e5ee7b3129fbd77255 [file] [log] [blame]
Ingo Molnar1d8c8b22009-04-20 15:52:29 +02001perf-stat(1)
Ingo Molnar6e6b7542008-04-15 22:39:31 +02002============
Ingo Molnar1d8c8b22009-04-20 15:52:29 +02003
4NAME
5----
6perf-stat - Run a command and gather performance counter statistics
7
8SYNOPSIS
9--------
10[verse]
Shawn Bohrer8c207692010-11-30 19:57:19 -060011'perf stat' [-e <EVENT> | --event=EVENT] [-a] <command>
12'perf stat' [-e <EVENT> | --event=EVENT] [-a] -- <command> [<options>]
Ingo Molnar1d8c8b22009-04-20 15:52:29 +020013
14DESCRIPTION
15-----------
16This command runs a command and gathers performance counter statistics
17from it.
18
19
20OPTIONS
21-------
22<command>...::
23 Any command you can specify in a shell.
24
Ingo Molnar20c84e92009-06-04 16:33:00 +020025
Ingo Molnar1d8c8b22009-04-20 15:52:29 +020026-e::
27--event=::
Cody P Schaferf9ab9c12015-01-07 17:13:53 -080028 Select the PMU event. Selection can be:
29
30 - a symbolic event name (use 'perf list' to list all events)
31
32 - a raw PMU event (eventsel+umask) in the form of rNNN where NNN is a
33 hexadecimal event descriptor.
34
35 - a symbolically formed event like 'pmu/param1=0x3,param2/' where
36 param1 and param2 are defined as formats for the PMU in
37 /sys/bus/event_sources/devices/<pmu>/format/*
38
39 - a symbolically formed event like 'pmu/config=M,config1=N,config2=K/'
40 where M, N, K are numbers (in decimal, hex, octal format).
41 Acceptable values for each of 'config', 'config1' and 'config2'
42 parameters are defined by corresponding entries in
43 /sys/bus/event_sources/devices/<pmu>/format/*
Ingo Molnar1d8c8b22009-04-20 15:52:29 +020044
Ingo Molnar20c84e92009-06-04 16:33:00 +020045-i::
Stephane Eranian2e6cdf92010-05-12 10:40:01 +020046--no-inherit::
47 child tasks do not inherit counters
Ingo Molnar20c84e92009-06-04 16:33:00 +020048-p::
49--pid=<pid>::
David Ahernb52956c2012-02-08 09:32:52 -070050 stat events on existing process id (comma separated list)
Shawn Bohrer8c207692010-11-30 19:57:19 -060051
52-t::
53--tid=<tid>::
David Ahernb52956c2012-02-08 09:32:52 -070054 stat events on existing thread id (comma separated list)
Shawn Bohrer8c207692010-11-30 19:57:19 -060055
Ingo Molnar20c84e92009-06-04 16:33:00 +020056
Ingo Molnar1d8c8b22009-04-20 15:52:29 +020057-a::
Shawn Bohrer8c207692010-11-30 19:57:19 -060058--all-cpus::
59 system-wide collection from all CPUs
Ingo Molnar1d8c8b22009-04-20 15:52:29 +020060
Brice Goglinb26bc5a2009-08-07 10:18:39 +020061-c::
Shawn Bohrer8c207692010-11-30 19:57:19 -060062--scale::
63 scale/normalize counter values
64
Borislav Petkovd7b60ba2016-03-07 16:44:44 -030065-d::
66--detailed::
67 print more detailed statistics, can be specified up to 3 times
68
69 -d: detailed events, L1 and LLC data cache
70 -d -d: more detailed events, dTLB and iTLB events
71 -d -d -d: very detailed events, adding prefetch events
72
Shawn Bohrer8c207692010-11-30 19:57:19 -060073-r::
74--repeat=<n>::
Frederik Deweerdta7e191c2013-03-01 13:02:27 -050075 repeat command and print average + stddev (max: 100). 0 means forever.
Ingo Molnar1d8c8b22009-04-20 15:52:29 +020076
Stephane Eranian5af52b52010-05-18 15:00:01 +020077-B::
Shawn Bohrer8c207692010-11-30 19:57:19 -060078--big-num::
Stephane Eranian5af52b52010-05-18 15:00:01 +020079 print large numbers with thousands' separators according to locale
80
Stephane Eranianc45c6ea2010-05-28 12:00:01 +020081-C::
82--cpu=::
Shawn Bohrer8c207692010-11-30 19:57:19 -060083Count only on the list of CPUs provided. Multiple CPUs can be provided as a
84comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2.
Stephane Eranianc45c6ea2010-05-28 12:00:01 +020085In per-thread mode, this option is ignored. The -a option is still necessary
86to activate system-wide monitoring. Default is to count on all CPUs.
87
Stephane Eranianf5b4a9c32010-11-16 11:05:01 +020088-A::
89--no-aggr::
90Do not aggregate counts across all monitored CPUs in system-wide mode (-a).
91This option is only valid in system-wide mode.
92
Shawn Bohrer8c207692010-11-30 19:57:19 -060093-n::
94--null::
95 null run - don't start any counters
96
97-v::
98--verbose::
99 be more verbose (show counter open errors, etc)
100
Stephane Eraniand7470b62010-12-01 18:49:05 +0200101-x SEP::
102--field-separator SEP::
103print counts using a CSV-style output to make it easy to import directly into
104spreadsheets. Columns are separated by the string specified in SEP.
105
Stephane Eranian023695d2011-02-14 11:20:01 +0200106-G name::
107--cgroup name::
108monitor only in the container (cgroup) called "name". This option is available only
109in per-cpu mode. The cgroup filesystem must be mounted. All threads belonging to
110container "name" are monitored when they run on the monitored CPUs. Multiple cgroups
111can be provided. Each cgroup is applied to the corresponding event, i.e., first cgroup
112to first event, second cgroup to second event and so on. It is possible to provide
113an empty cgroup (monitor all the time) using, e.g., -G foo,,bar. Cgroups must have
114corresponding events, i.e., they always refer to events defined earlier on the command
115line.
116
Stephane Eranian4aa90152011-08-15 22:22:33 +0200117-o file::
Jim Cromie56f3bae2011-09-07 17:14:00 -0600118--output file::
Stephane Eranian4aa90152011-08-15 22:22:33 +0200119Print the output into the designated file.
120
121--append::
122Append to the output file designated with the -o option. Ignored if -o is not specified.
123
Jim Cromie56f3bae2011-09-07 17:14:00 -0600124--log-fd::
125
126Log output to fd, instead of stderr. Complementary to --output, and mutually exclusive
127with it. --append may be used here. Examples:
128 3>results perf stat --log-fd 3 -- $cmd
129 3>>results perf stat --log-fd 3 --append -- $cmd
130
Peter Zijlstra1f16c572012-10-23 13:40:14 +0200131--pre::
132--post::
133 Pre and post measurement hooks, e.g.:
134
135perf stat --repeat 10 --null --sync --pre 'make -s O=defconfig-build/clean' -- make -s -j64 O=defconfig-build/ bzImage
Jim Cromie56f3bae2011-09-07 17:14:00 -0600136
Stephane Eranian13370a92013-01-29 12:47:44 +0100137-I msecs::
138--interval-print msecs::
Kan Liang19afd102015-10-02 05:04:34 -0400139Print count deltas every N milliseconds (minimum: 10ms)
140The overhead percentage could be high in some cases, for instance with small, sub 100ms intervals. Use with caution.
141 example: 'perf stat -I 1000 -e cycles -a sleep 5'
Jim Cromie56f3bae2011-09-07 17:14:00 -0600142
Stephane Eraniand4304952013-02-14 13:57:28 +0100143--per-socket::
Stephane Eraniand7e7a452013-02-06 15:46:02 +0100144Aggregate counts per processor socket for system-wide mode measurements. This
145is a useful mode to detect imbalance between sockets. To enable this mode,
Stephane Eraniand4304952013-02-14 13:57:28 +0100146use --per-socket in addition to -a. (system-wide). The output includes the
Stephane Eraniand7e7a452013-02-06 15:46:02 +0100147socket number and the number of online processors on that socket. This is
148useful to gauge the amount of aggregation.
149
Stephane Eranian12c08a92013-02-14 13:57:29 +0100150--per-core::
151Aggregate counts per physical processor for system-wide mode measurements. This
152is a useful mode to detect imbalance between physical cores. To enable this mode,
153use --per-core in addition to -a. (system-wide). The output includes the
154core number and the number of online logical processors on that physical processor.
155
Jiri Olsa32b8af82015-06-26 11:29:27 +0200156--per-thread::
157Aggregate counts per monitored threads, when monitoring threads (-t option)
158or processes (-p option).
159
Andi Kleen41191682013-08-02 17:41:11 -0700160-D msecs::
Andi Kleen8f3dd2b2014-01-07 14:14:06 -0800161--delay msecs::
Andi Kleen41191682013-08-02 17:41:11 -0700162After starting the program, wait msecs before measuring. This is useful to
163filter out the startup phase of the program, which is often very different.
164
Andi Kleen4cabc3d2013-08-21 16:47:26 -0700165-T::
166--transaction::
167
168Print statistics of transactional execution if supported.
169
Ingo Molnar1d8c8b22009-04-20 15:52:29 +0200170EXAMPLES
171--------
172
Ingo Molnar20c84e92009-06-04 16:33:00 +0200173$ perf stat -- make -j
Ingo Molnar1d8c8b22009-04-20 15:52:29 +0200174
Ingo Molnar20c84e92009-06-04 16:33:00 +0200175 Performance counter stats for 'make -j':
Ingo Molnar1d8c8b22009-04-20 15:52:29 +0200176
Ingo Molnar20c84e92009-06-04 16:33:00 +0200177 8117.370256 task clock ticks # 11.281 CPU utilization factor
178 678 context switches # 0.000 M/sec
179 133 CPU migrations # 0.000 M/sec
180 235724 pagefaults # 0.029 M/sec
181 24821162526 CPU cycles # 3057.784 M/sec
182 18687303457 instructions # 2302.138 M/sec
183 172158895 cache references # 21.209 M/sec
184 27075259 cache misses # 3.335 M/sec
Ingo Molnar1d8c8b22009-04-20 15:52:29 +0200185
Ingo Molnar20c84e92009-06-04 16:33:00 +0200186 Wall-clock time elapsed: 719.554352 msecs
Ingo Molnar1d8c8b22009-04-20 15:52:29 +0200187
188SEE ALSO
189--------
Thomas Gleixner386b05e2009-06-06 14:56:33 +0200190linkperf:perf-top[1], linkperf:perf-list[1]