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-p::
66--pid=::
67	Record events on existing process ID (comma separated list).
68
69-t::
70--tid=::
71        Record events on existing thread ID (comma separated list).
72        This option also disables inheritance by default.  Enable it by adding
73        --inherit.
74
75-u::
76--uid=::
77        Record events in threads owned by uid. Name or number.
78
79-r::
80--realtime=::
81	Collect data with this RT SCHED_FIFO priority.
82
83--no-buffering::
84	Collect data without buffering.
85
86-c::
87--count=::
88	Event period to sample.
89
90-o::
91--output=::
92	Output file name.
93
94-i::
95--no-inherit::
96	Child tasks do not inherit counters.
97-F::
98--freq=::
99	Profile at this frequency.
100
101-m::
102--mmap-pages=::
103	Number of mmap data pages (must be a power of two) or size
104	specification with appended unit character - B/K/M/G. The
105	size is rounded up to have nearest pages power of two value.
106
107-g::
108	Enables call-graph (stack chain/backtrace) recording.
109
110--call-graph::
111	Setup and enable call-graph (stack chain/backtrace) recording,
112	implies -g.
113
114	Allows specifying "fp" (frame pointer) or "dwarf"
115	(DWARF's CFI - Call Frame Information) or "lbr"
116	(Hardware Last Branch Record facility) as the method to collect
117	the information used to show the call graphs.
118
119	In some systems, where binaries are build with gcc
120	--fomit-frame-pointer, using the "fp" method will produce bogus
121	call graphs, using "dwarf", if available (perf tools linked to
122	the libunwind library) should be used instead.
123	Using the "lbr" method doesn't require any compiler options. It
124	will produce call graphs from the hardware LBR registers. The
125	main limition is that it is only available on new Intel
126	platforms, such as Haswell. It can only get user call chain. It
127	doesn't work with branch stack sampling at the same time.
128
129-q::
130--quiet::
131	Don't print any message, useful for scripting.
132
133-v::
134--verbose::
135	Be more verbose (show counter open errors, etc).
136
137-s::
138--stat::
139	Per thread counts.
140
141-d::
142--data::
143	Sample addresses.
144
145-T::
146--timestamp::
147	Sample timestamps. Use it with 'perf report -D' to see the timestamps,
148	for instance.
149
150-n::
151--no-samples::
152	Don't sample.
153
154-R::
155--raw-samples::
156Collect raw sample records from all opened counters (default for tracepoint counters).
157
158-C::
159--cpu::
160Collect samples only on the list of CPUs provided. Multiple CPUs can be provided as a
161comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2.
162In per-thread mode with inheritance mode on (default), samples are captured only when
163the thread executes on the designated CPUs. Default is to monitor all CPUs.
164
165-N::
166--no-buildid-cache::
167Do not update the buildid cache. This saves some overhead in situations
168where the information in the perf.data file (which includes buildids)
169is sufficient.
170
171-G name,...::
172--cgroup name,...::
173monitor only in the container (cgroup) called "name". This option is available only
174in per-cpu mode. The cgroup filesystem must be mounted. All threads belonging to
175container "name" are monitored when they run on the monitored CPUs. Multiple cgroups
176can be provided. Each cgroup is applied to the corresponding event, i.e., first cgroup
177to first event, second cgroup to second event and so on. It is possible to provide
178an empty cgroup (monitor all the time) using, e.g., -G foo,,bar. Cgroups must have
179corresponding events, i.e., they always refer to events defined earlier on the command
180line.
181
182-b::
183--branch-any::
184Enable taken branch stack sampling. Any type of taken branch may be sampled.
185This is a shortcut for --branch-filter any. See --branch-filter for more infos.
186
187-j::
188--branch-filter::
189Enable taken branch stack sampling. Each sample captures a series of consecutive
190taken branches. The number of branches captured with each sample depends on the
191underlying hardware, the type of branches of interest, and the executed code.
192It is possible to select the types of branches captured by enabling filters. The
193following filters are defined:
194
195        - any:  any type of branches
196        - any_call: any function call or system call
197        - any_ret: any function return or system call return
198        - ind_call: any indirect branch
199        - u:  only when the branch target is at the user level
200        - k: only when the branch target is in the kernel
201        - hv: only when the target is at the hypervisor level
202	- in_tx: only when the target is in a hardware transaction
203	- no_tx: only when the target is not in a hardware transaction
204	- abort_tx: only when the target is a hardware transaction abort
205	- cond: conditional branches
206
207+
208The option requires at least one branch type among any, any_call, any_ret, ind_call, cond.
209The privilege levels may be omitted, in which case, the privilege levels of the associated
210event are applied to the branch filter. Both kernel (k) and hypervisor (hv) privilege
211levels are subject to permissions.  When sampling on multiple events, branch stack sampling
212is enabled for all the sampling events. The sampled branch type is the same for all events.
213The various filters must be specified as a comma separated list: --branch-filter any_ret,u,k
214Note that this feature may not be available on all processors.
215
216--weight::
217Enable weightened sampling. An additional weight is recorded per sample and can be
218displayed with the weight and local_weight sort keys.  This currently works for TSX
219abort events and some memory events in precise mode on modern Intel CPUs.
220
221--transaction::
222Record transaction flags for transaction related events.
223
224--per-thread::
225Use per-thread mmaps.  By default per-cpu mmaps are created.  This option
226overrides that and uses per-thread mmaps.  A side-effect of that is that
227inheritance is automatically disabled.  --per-thread is ignored with a warning
228if combined with -a or -C options.
229
230-D::
231--delay=::
232After starting the program, wait msecs before measuring. This is useful to
233filter out the startup phase of the program, which is often very different.
234
235-I::
236--intr-regs::
237Capture machine state (registers) at interrupt, i.e., on counter overflows for
238each sample. List of captured registers depends on the architecture. This option
239is off by default.
240
241--running-time::
242Record running and enabled time for read events (:S)
243
244SEE ALSO
245--------
246linkperf:perf-stat[1], linkperf:perf-list[1]
247