6 perf-stat - Run a command and gather performance counter statistics
11 'perf stat' [-e <EVENT> | --event=EVENT] [-a] <command>
12 'perf stat' [-e <EVENT> | --event=EVENT] [-a] -- <command> [<options>]
16 This command runs a command and gathers performance counter statistics
23 Any command you can specify in a shell.
28 Select the PMU event. Selection can be:
30 - a symbolic event name (use 'perf list' to list all events)
32 - a raw PMU event (eventsel+umask) in the form of rNNN where NNN is a
33 hexadecimal event descriptor.
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/*
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/*
47 child tasks do not inherit counters
50 stat events on existing process id (comma separated list)
54 stat events on existing thread id (comma separated list)
59 system-wide collection from all CPUs
63 scale/normalize counter values
67 print more detailed statistics, can be specified up to 3 times
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
75 repeat command and print average + stddev (max: 100). 0 means forever.
79 print large numbers with thousands' separators according to locale
83 Count only on the list of CPUs provided. Multiple CPUs can be provided as a
84 comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2.
85 In per-thread mode, this option is ignored. The -a option is still necessary
86 to activate system-wide monitoring. Default is to count on all CPUs.
90 Do not aggregate counts across all monitored CPUs in system-wide mode (-a).
91 This option is only valid in system-wide mode.
95 null run - don't start any counters
99 be more verbose (show counter open errors, etc)
102 --field-separator SEP::
103 print counts using a CSV-style output to make it easy to import directly into
104 spreadsheets. Columns are separated by the string specified in SEP.
108 monitor only in the container (cgroup) called "name". This option is available only
109 in per-cpu mode. The cgroup filesystem must be mounted. All threads belonging to
110 container "name" are monitored when they run on the monitored CPUs. Multiple cgroups
111 can be provided. Each cgroup is applied to the corresponding event, i.e., first cgroup
112 to first event, second cgroup to second event and so on. It is possible to provide
113 an empty cgroup (monitor all the time) using, e.g., -G foo,,bar. Cgroups must have
114 corresponding events, i.e., they always refer to events defined earlier on the command
119 Print the output into the designated file.
122 Append to the output file designated with the -o option. Ignored if -o is not specified.
126 Log output to fd, instead of stderr. Complementary to --output, and mutually exclusive
127 with 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
133 Pre and post measurement hooks, e.g.:
135 perf stat --repeat 10 --null --sync --pre 'make -s O=defconfig-build/clean' -- make -s -j64 O=defconfig-build/ bzImage
138 --interval-print msecs::
139 Print count deltas every N milliseconds (minimum: 10ms)
140 The 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'
144 Aggregate counts per processor socket for system-wide mode measurements. This
145 is a useful mode to detect imbalance between sockets. To enable this mode,
146 use --per-socket in addition to -a. (system-wide). The output includes the
147 socket number and the number of online processors on that socket. This is
148 useful to gauge the amount of aggregation.
151 Aggregate counts per physical processor for system-wide mode measurements. This
152 is a useful mode to detect imbalance between physical cores. To enable this mode,
153 use --per-core in addition to -a. (system-wide). The output includes the
154 core number and the number of online logical processors on that physical processor.
157 Aggregate counts per monitored threads, when monitoring threads (-t option)
158 or processes (-p option).
162 After starting the program, wait msecs before measuring. This is useful to
163 filter out the startup phase of the program, which is often very different.
168 Print statistics of transactional execution if supported.
173 $ perf stat -- make -j
175 Performance counter stats for 'make -j':
177 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
186 Wall-clock time elapsed: 719.554352 msecs
190 linkperf:perf-top[1], linkperf:perf-list[1]