xref: /linux/tools/perf/Documentation/perf-record.txt (revision 2ba9268dd603d23e17643437b2246acb6844953b)
1perf-record(1)
2==============
3
4NAME
5----
6perf-record - Run a command and record its profile into perf.data
7
8SYNOPSIS
9--------
10[verse]
11'perf record' [-e <EVENT> | --event=EVENT] [-l] [-a] <command>
12'perf record' [-e <EVENT> | --event=EVENT] [-l] [-a] -- <command> [<options>]
13
14DESCRIPTION
15-----------
16This command runs a command and gathers a performance counter profile
17from it, into perf.data - without displaying anything.
18
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=::
29	Select the PMU event. Selection can be:
30
31        - 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 symbolically formed PMU event like 'pmu/param1=0x3,param2/' where
37	  'param1', 'param2', etc are defined as formats for the PMU in
38	  /sys/bus/event_sources/devices/<pmu>/format/*.
39
40	- a symbolically formed event like 'pmu/config=M,config1=N,config3=K/'
41
42          where M, N, K are numbers (in decimal, hex, octal format). Acceptable
43          values for each of 'config', 'config1' and 'config2' are defined by
44          corresponding entries in /sys/bus/event_sources/devices/<pmu>/format/*
45          param1 and param2 are defined as formats for the PMU in:
46          /sys/bus/event_sources/devices/<pmu>/format/*
47
48        - a hardware breakpoint event in the form of '\mem:addr[/len][:access]'
49          where addr is the address in memory you want to break in.
50          Access is the memory access type (read, write, execute) it can
51          be passed as follows: '\mem:addr[:[r][w][x]]'. len is the range,
52          number of bytes from specified addr, which the breakpoint will cover.
53          If you want to profile read-write accesses in 0x1000, just set
54          'mem:0x1000:rw'.
55          If you want to profile write accesses in [0x1000~1008), just set
56          'mem:0x1000/8:w'.
57
58--filter=<filter>::
59        Event filter.
60
61-a::
62--all-cpus::
63        System-wide collection from all CPUs.
64
65-l::
66        Scale counter values.
67
68-p::
69--pid=::
70	Record events on existing process ID (comma separated list).
71
72-t::
73--tid=::
74        Record events on existing thread ID (comma separated list).
75        This option also disables inheritance by default.  Enable it by adding
76        --inherit.
77
78-u::
79--uid=::
80        Record events in threads owned by uid. Name or number.
81
82-r::
83--realtime=::
84	Collect data with this RT SCHED_FIFO priority.
85
86--no-buffering::
87	Collect data without buffering.
88
89-c::
90--count=::
91	Event period to sample.
92
93-o::
94--output=::
95	Output file name.
96
97-i::
98--no-inherit::
99	Child tasks do not inherit counters.
100-F::
101--freq=::
102	Profile at this frequency.
103
104-m::
105--mmap-pages=::
106	Number of mmap data pages (must be a power of two) or size
107	specification with appended unit character - B/K/M/G. The
108	size is rounded up to have nearest pages power of two value.
109
110-g::
111	Enables call-graph (stack chain/backtrace) recording.
112
113--call-graph::
114	Setup and enable call-graph (stack chain/backtrace) recording,
115	implies -g.
116
117	Allows specifying "fp" (frame pointer) or "dwarf"
118	(DWARF's CFI - Call Frame Information) as the method to collect
119	the information used to show the call graphs.
120
121	In some systems, where binaries are build with gcc
122	--fomit-frame-pointer, using the "fp" method will produce bogus
123	call graphs, using "dwarf", if available (perf tools linked to
124	the libunwind library) should be used instead.
125
126-q::
127--quiet::
128	Don't print any message, useful for scripting.
129
130-v::
131--verbose::
132	Be more verbose (show counter open errors, etc).
133
134-s::
135--stat::
136	Per thread counts.
137
138-d::
139--data::
140	Sample addresses.
141
142-T::
143--timestamp::
144	Sample timestamps. Use it with 'perf report -D' to see the timestamps,
145	for instance.
146
147-n::
148--no-samples::
149	Don't sample.
150
151-R::
152--raw-samples::
153Collect raw sample records from all opened counters (default for tracepoint counters).
154
155-C::
156--cpu::
157Collect samples only on the list of CPUs provided. Multiple CPUs can be provided as a
158comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2.
159In per-thread mode with inheritance mode on (default), samples are captured only when
160the thread executes on the designated CPUs. Default is to monitor all CPUs.
161
162-N::
163--no-buildid-cache::
164Do not update the buildid cache. This saves some overhead in situations
165where the information in the perf.data file (which includes buildids)
166is sufficient.
167
168-G name,...::
169--cgroup name,...::
170monitor only in the container (cgroup) called "name". This option is available only
171in per-cpu mode. The cgroup filesystem must be mounted. All threads belonging to
172container "name" are monitored when they run on the monitored CPUs. Multiple cgroups
173can be provided. Each cgroup is applied to the corresponding event, i.e., first cgroup
174to first event, second cgroup to second event and so on. It is possible to provide
175an empty cgroup (monitor all the time) using, e.g., -G foo,,bar. Cgroups must have
176corresponding events, i.e., they always refer to events defined earlier on the command
177line.
178
179-b::
180--branch-any::
181Enable taken branch stack sampling. Any type of taken branch may be sampled.
182This is a shortcut for --branch-filter any. See --branch-filter for more infos.
183
184-j::
185--branch-filter::
186Enable taken branch stack sampling. Each sample captures a series of consecutive
187taken branches. The number of branches captured with each sample depends on the
188underlying hardware, the type of branches of interest, and the executed code.
189It is possible to select the types of branches captured by enabling filters. The
190following filters are defined:
191
192        - any:  any type of branches
193        - any_call: any function call or system call
194        - any_ret: any function return or system call return
195        - ind_call: any indirect branch
196        - u:  only when the branch target is at the user level
197        - k: only when the branch target is in the kernel
198        - hv: only when the target is at the hypervisor level
199	- in_tx: only when the target is in a hardware transaction
200	- no_tx: only when the target is not in a hardware transaction
201	- abort_tx: only when the target is a hardware transaction abort
202	- cond: conditional branches
203
204+
205The option requires at least one branch type among any, any_call, any_ret, ind_call, cond.
206The privilege levels may be omitted, in which case, the privilege levels of the associated
207event are applied to the branch filter. Both kernel (k) and hypervisor (hv) privilege
208levels are subject to permissions.  When sampling on multiple events, branch stack sampling
209is enabled for all the sampling events. The sampled branch type is the same for all events.
210The various filters must be specified as a comma separated list: --branch-filter any_ret,u,k
211Note that this feature may not be available on all processors.
212
213--weight::
214Enable weightened sampling. An additional weight is recorded per sample and can be
215displayed with the weight and local_weight sort keys.  This currently works for TSX
216abort events and some memory events in precise mode on modern Intel CPUs.
217
218--transaction::
219Record transaction flags for transaction related events.
220
221--per-thread::
222Use per-thread mmaps.  By default per-cpu mmaps are created.  This option
223overrides that and uses per-thread mmaps.  A side-effect of that is that
224inheritance is automatically disabled.  --per-thread is ignored with a warning
225if combined with -a or -C options.
226
227-D::
228--delay=::
229After starting the program, wait msecs before measuring. This is useful to
230filter out the startup phase of the program, which is often very different.
231
232-I::
233--intr-regs::
234Capture machine state (registers) at interrupt, i.e., on counter overflows for
235each sample. List of captured registers depends on the architecture. This option
236is off by default.
237
238SEE ALSO
239--------
240linkperf:perf-stat[1], linkperf:perf-list[1]
241