tools/perf/Documentation/perf-record.txt

   1 perf-record(1)
   2 ==============
   3
   4 NAME
   5 ----
   6 perf-record - Run a command and record its profile into perf.data
   7
   8 SYNOPSIS
   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
  14 DESCRIPTION
  15 -----------
  16 This command runs a command and gathers a performance counter profile
  17 from it, into perf.data - without displaying anything.
  18
  19 This file can then be inspected later on, using 'perf report'.
  20
  21
  22 OPTIONS
  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           There are also some params which are not defined in .../<pmu>/format/*.
  49           These params can be used to overload default config values per event.
  50           Here is a list of the params.
  51           - 'period': Set event sampling period
  52           - 'time': Disable/enable time stamping. Acceptable values are 1 for
  53                     enabling time stamping. 0 for disabling time stamping.
  54                     The default is 1.
  55           Note: If user explicitly sets options which conflict with the params,
  56           the value set by the params will be overridden.
  57
  58         - a hardware breakpoint event in the form of '\mem:addr[/len][:access]'
  59           where addr is the address in memory you want to break in.
  60           Access is the memory access type (read, write, execute) it can
  61           be passed as follows: '\mem:addr[:[r][w][x]]'. len is the range,
  62           number of bytes from specified addr, which the breakpoint will cover.
  63           If you want to profile read-write accesses in 0x1000, just set
  64           'mem:0x1000:rw'.
  65           If you want to profile write accesses in [0x1000~1008), just set
  66           'mem:0x1000/8:w'.
  67
  68         - a group of events surrounded by a pair of brace ("{event1,event2,...}").
  69           Each event is separated by commas and the group should be quoted to
  70           prevent the shell interpretation.  You also need to use --group on
  71           "perf report" to view group events together.
  72
  73 --filter=<filter>::
  74         Event filter. This option should follow a event selector (-e) which
  75         selects tracepoint event(s). Multiple '--filter' options are combined
  76         using '&&'.
  77
  78 --exclude-perf::
  79         Don't record events issued by perf itself. This option should follow
  80         a event selector (-e) which selects tracepoint event(s). It adds a
  81         filter expression 'common_pid != $PERFPID' to filters. If other
  82         '--filter' exists, the new filter expression will be combined with
  83         them by '&&'.
  84
  85 -a::
  86 --all-cpus::
  87         System-wide collection from all CPUs.
  88
  89 -p::
  90 --pid=::
  91         Record events on existing process ID (comma separated list).
  92
  93 -t::
  94 --tid=::
  95         Record events on existing thread ID (comma separated list).
  96         This option also disables inheritance by default.  Enable it by adding
  97         --inherit.
  98
  99 -u::
 100 --uid=::
 101         Record events in threads owned by uid. Name or number.
 102
 103 -r::
 104 --realtime=::
 105         Collect data with this RT SCHED_FIFO priority.
 106
 107 --no-buffering::
 108         Collect data without buffering.
 109
 110 -c::
 111 --count=::
 112         Event period to sample.
 113
 114 -o::
 115 --output=::
 116         Output file name.
 117
 118 -i::
 119 --no-inherit::
 120         Child tasks do not inherit counters.
 121 -F::
 122 --freq=::
 123         Profile at this frequency.
 124
 125 -m::
 126 --mmap-pages=::
 127         Number of mmap data pages (must be a power of two) or size
 128         specification with appended unit character - B/K/M/G. The
 129         size is rounded up to have nearest pages power of two value.
 130         Also, by adding a comma, the number of mmap pages for AUX
 131         area tracing can be specified.
 132
 133 --group::
 134         Put all events in a single event group.  This precedes the --event
 135         option and remains only for backward compatibility.  See --event.
 136
 137 -g::
 138         Enables call-graph (stack chain/backtrace) recording.
 139
 140 --call-graph::
 141         Setup and enable call-graph (stack chain/backtrace) recording,
 142         implies -g.
 143
 144         Allows specifying "fp" (frame pointer) or "dwarf"
 145         (DWARF's CFI - Call Frame Information) or "lbr"
 146         (Hardware Last Branch Record facility) as the method to collect
 147         the information used to show the call graphs.
 148
 149         In some systems, where binaries are build with gcc
 150         --fomit-frame-pointer, using the "fp" method will produce bogus
 151         call graphs, using "dwarf", if available (perf tools linked to
 152         the libunwind library) should be used instead.
 153         Using the "lbr" method doesn't require any compiler options. It
 154         will produce call graphs from the hardware LBR registers. The
 155         main limition is that it is only available on new Intel
 156         platforms, such as Haswell. It can only get user call chain. It
 157         doesn't work with branch stack sampling at the same time.
 158
 159 -q::
 160 --quiet::
 161         Don't print any message, useful for scripting.
 162
 163 -v::
 164 --verbose::
 165         Be more verbose (show counter open errors, etc).
 166
 167 -s::
 168 --stat::
 169         Record per-thread event counts.  Use it with 'perf report -T' to see
 170         the values.
 171
 172 -d::
 173 --data::
 174         Record the sample addresses.
 175
 176 -T::
 177 --timestamp::
 178         Record the sample timestamps. Use it with 'perf report -D' to see the
 179         timestamps, for instance.
 180
 181 -P::
 182 --period::
 183         Record the sample period.
 184
 185 -n::
 186 --no-samples::
 187         Don't sample.
 188
 189 -R::
 190 --raw-samples::
 191 Collect raw sample records from all opened counters (default for tracepoint counters).
 192
 193 -C::
 194 --cpu::
 195 Collect samples only on the list of CPUs provided. Multiple CPUs can be provided as a
 196 comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2.
 197 In per-thread mode with inheritance mode on (default), samples are captured only when
 198 the thread executes on the designated CPUs. Default is to monitor all CPUs.
 199
 200 -N::
 201 --no-buildid-cache::
 202 Do not update the buildid cache. This saves some overhead in situations
 203 where the information in the perf.data file (which includes buildids)
 204 is sufficient.
 205
 206 -G name,...::
 207 --cgroup name,...::
 208 monitor only in the container (cgroup) called "name". This option is available only
 209 in per-cpu mode. The cgroup filesystem must be mounted. All threads belonging to
 210 container "name" are monitored when they run on the monitored CPUs. Multiple cgroups
 211 can be provided. Each cgroup is applied to the corresponding event, i.e., first cgroup
 212 to first event, second cgroup to second event and so on. It is possible to provide
 213 an empty cgroup (monitor all the time) using, e.g., -G foo,,bar. Cgroups must have
 214 corresponding events, i.e., they always refer to events defined earlier on the command
 215 line.
 216
 217 -b::
 218 --branch-any::
 219 Enable taken branch stack sampling. Any type of taken branch may be sampled.
 220 This is a shortcut for --branch-filter any. See --branch-filter for more infos.
 221
 222 -j::
 223 --branch-filter::
 224 Enable taken branch stack sampling. Each sample captures a series of consecutive
 225 taken branches. The number of branches captured with each sample depends on the
 226 underlying hardware, the type of branches of interest, and the executed code.
 227 It is possible to select the types of branches captured by enabling filters. The
 228 following filters are defined:
 229
 230         - any:  any type of branches
 231         - any_call: any function call or system call
 232         - any_ret: any function return or system call return
 233         - ind_call: any indirect branch
 234         - u:  only when the branch target is at the user level
 235         - k: only when the branch target is in the kernel
 236         - hv: only when the target is at the hypervisor level
 237         - in_tx: only when the target is in a hardware transaction
 238         - no_tx: only when the target is not in a hardware transaction
 239         - abort_tx: only when the target is a hardware transaction abort
 240         - cond: conditional branches
 241
 242 +
 243 The option requires at least one branch type among any, any_call, any_ret, ind_call, cond.
 244 The privilege levels may be omitted, in which case, the privilege levels of the associated
 245 event are applied to the branch filter. Both kernel (k) and hypervisor (hv) privilege
 246 levels are subject to permissions.  When sampling on multiple events, branch stack sampling
 247 is enabled for all the sampling events. The sampled branch type is the same for all events.
 248 The various filters must be specified as a comma separated list: --branch-filter any_ret,u,k
 249 Note that this feature may not be available on all processors.
 250
 251 --weight::
 252 Enable weightened sampling. An additional weight is recorded per sample and can be
 253 displayed with the weight and local_weight sort keys.  This currently works for TSX
 254 abort events and some memory events in precise mode on modern Intel CPUs.
 255
 256 --transaction::
 257 Record transaction flags for transaction related events.
 258
 259 --per-thread::
 260 Use per-thread mmaps.  By default per-cpu mmaps are created.  This option
 261 overrides that and uses per-thread mmaps.  A side-effect of that is that
 262 inheritance is automatically disabled.  --per-thread is ignored with a warning
 263 if combined with -a or -C options.
 264
 265 -D::
 266 --delay=::
 267 After starting the program, wait msecs before measuring. This is useful to
 268 filter out the startup phase of the program, which is often very different.
 269
 270 -I::
 271 --intr-regs::
 272 Capture machine state (registers) at interrupt, i.e., on counter overflows for
 273 each sample. List of captured registers depends on the architecture. This option
 274 is off by default.
 275
 276 --running-time::
 277 Record running and enabled time for read events (:S)
 278
 279 -k::
 280 --clockid::
 281 Sets the clock id to use for the various time fields in the perf_event_type
 282 records. See clock_gettime(). In particular CLOCK_MONOTONIC and
 283 CLOCK_MONOTONIC_RAW are supported, some events might also allow
 284 CLOCK_BOOTTIME, CLOCK_REALTIME and CLOCK_TAI.
 285
 286 -S::
 287 --snapshot::
 288 Select AUX area tracing Snapshot Mode. This option is valid only with an
 289 AUX area tracing event. Optionally the number of bytes to capture per
 290 snapshot can be specified. In Snapshot Mode, trace data is captured only when
 291 signal SIGUSR2 is received.
 292
 293 --proc-map-timeout::
 294 When processing pre-existing threads /proc/XXX/mmap, it may take a long time,
 295 because the file may be huge. A time out is needed in such cases.
 296 This option sets the time out limit. The default value is 500 ms.
 297
 298 --switch-events::
 299 Record context switch events i.e. events of type PERF_RECORD_SWITCH or
 300 PERF_RECORD_SWITCH_CPU_WIDE.
 301
 302 SEE ALSO
 303 --------
 304 linkperf:perf-stat[1], linkperf:perf-list[1]