6 perf-record - Run a command and record its profile into perf.data
11 'perf record' [-e <EVENT> | --event=EVENT] [-l] [-a] <command>
12 'perf record' [-e <EVENT> | --event=EVENT] [-l] [-a] -- <command> [<options>]
16 This command runs a command and gathers a performance counter profile
17 from it, into perf.data - without displaying anything.
19 This file can then be inspected later on, using 'perf report'.
25 Any command you can specify in a shell.
29 Select the PMU event. Selection can be:
31 - a symbolic event name (use 'perf list' to list all events)
33 - a raw PMU event (eventsel+umask) in the form of rNNN where NNN is a
34 hexadecimal event descriptor.
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
48 System-wide collection from all CPUs.
55 Record events on existing process ID (comma separated list).
59 Record events on existing thread ID (comma separated list).
63 Record events in threads owned by uid. Name or number.
67 Collect data with this RT SCHED_FIFO priority.
71 Collect data without buffering.
75 Event period to sample.
83 Child tasks do not inherit counters.
86 Profile at this frequency.
90 Number of mmap data pages. Must be a power of two.
93 Enables call-graph (stack chain/backtrace) recording.
96 Setup and enable call-graph (stack chain/backtrace) recording,
99 Allows specifying "fp" (frame pointer) or "dwarf"
100 (DWARF's CFI - Call Frame Information) as the method to collect
101 the information used to show the call graphs.
103 In some systems, where binaries are build with gcc
104 --fomit-frame-pointer, using the "fp" method will produce bogus
105 call graphs, using "dwarf", if available (perf tools linked to
106 the libunwind library) should be used instead.
110 Don't print any message, useful for scripting.
114 Be more verbose (show counter open errors, etc).
126 Sample timestamps. Use it with 'perf report -D' to see the timestamps,
135 Collect raw sample records from all opened counters (default for tracepoint counters).
139 Collect samples only on the list of CPUs provided. Multiple CPUs can be provided as a
140 comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2.
141 In per-thread mode with inheritance mode on (default), samples are captured only when
142 the thread executes on the designated CPUs. Default is to monitor all CPUs.
146 Do not update the builid cache. This saves some overhead in situations
147 where the information in the perf.data file (which includes buildids)
152 monitor only in the container (cgroup) called "name". This option is available only
153 in per-cpu mode. The cgroup filesystem must be mounted. All threads belonging to
154 container "name" are monitored when they run on the monitored CPUs. Multiple cgroups
155 can be provided. Each cgroup is applied to the corresponding event, i.e., first cgroup
156 to first event, second cgroup to second event and so on. It is possible to provide
157 an empty cgroup (monitor all the time) using, e.g., -G foo,,bar. Cgroups must have
158 corresponding events, i.e., they always refer to events defined earlier on the command
163 Enable taken branch stack sampling. Any type of taken branch may be sampled.
164 This is a shortcut for --branch-filter any. See --branch-filter for more infos.
168 Enable taken branch stack sampling. Each sample captures a series of consecutive
169 taken branches. The number of branches captured with each sample depends on the
170 underlying hardware, the type of branches of interest, and the executed code.
171 It is possible to select the types of branches captured by enabling filters. The
172 following filters are defined:
174 - any: any type of branches
175 - any_call: any function call or system call
176 - any_ret: any function return or system call return
177 - ind_call: any indirect branch
178 - u: only when the branch target is at the user level
179 - k: only when the branch target is in the kernel
180 - hv: only when the target is at the hypervisor level
183 The option requires at least one branch type among any, any_call, any_ret, ind_call.
184 The privilege levels may be omitted, in which case, the privilege levels of the associated
185 event are applied to the branch filter. Both kernel (k) and hypervisor (hv) privilege
186 levels are subject to permissions. When sampling on multiple events, branch stack sampling
187 is enabled for all the sampling events. The sampled branch type is the same for all events.
188 The various filters must be specified as a comma separated list: --branch-filter any_ret,u,k
189 Note that this feature may not be available on all processors.
193 Enable weightened sampling. An additional weight is recorded per sample and can be
194 displayed with the weight and local_weight sort keys. This currently works for TSX
195 abort events and some memory events in precise mode on modern Intel CPUs.
199 linkperf:perf-stat[1], linkperf:perf-list[1]