Ingo Molnar | e33e0a4 | 2009-04-20 15:58:01 +0200 | [diff] [blame] | 1 | perf-record(1) |
Ingo Molnar | c1c2365 | 2009-05-30 12:38:51 +0200 | [diff] [blame] | 2 | ============== |
Ingo Molnar | e33e0a4 | 2009-04-20 15:58:01 +0200 | [diff] [blame] | 3 | |
| 4 | NAME |
| 5 | ---- |
Ingo Molnar | 23ac9cb | 2009-05-27 09:33:18 +0200 | [diff] [blame] | 6 | perf-record - Run a command and record its profile into perf.data |
Ingo Molnar | e33e0a4 | 2009-04-20 15:58:01 +0200 | [diff] [blame] | 7 | |
| 8 | SYNOPSIS |
| 9 | -------- |
| 10 | [verse] |
| 11 | 'perf record' [-e <EVENT> | --event=EVENT] [-l] [-a] <command> |
Mike Galbraith | 9e096753 | 2009-05-28 16:25:34 +0200 | [diff] [blame] | 12 | 'perf record' [-e <EVENT> | --event=EVENT] [-l] [-a] -- <command> [<options>] |
Ingo Molnar | e33e0a4 | 2009-04-20 15:58:01 +0200 | [diff] [blame] | 13 | |
| 14 | DESCRIPTION |
| 15 | ----------- |
| 16 | This command runs a command and gathers a performance counter profile |
Ingo Molnar | 23ac9cb | 2009-05-27 09:33:18 +0200 | [diff] [blame] | 17 | from it, into perf.data - without displaying anything. |
Ingo Molnar | e33e0a4 | 2009-04-20 15:58:01 +0200 | [diff] [blame] | 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=:: |
Frederic Weisbecker | 1b290d6 | 2009-11-23 15:42:35 +0100 | [diff] [blame] | 29 | Select the PMU event. Selection can be: |
Ingo Molnar | e33e0a4 | 2009-04-20 15:58:01 +0200 | [diff] [blame] | 30 | |
Frederic Weisbecker | 1b290d6 | 2009-11-23 15:42:35 +0100 | [diff] [blame] | 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 | |
Cody P Schafer | f9ab9c1 | 2015-01-07 17:13:53 -0800 | [diff] [blame] | 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 | |
Kan Liang | 3d5d68a | 2015-07-08 04:44:54 -0400 | [diff] [blame] | 48 | There are also some params which are not defined in .../<pmu>/format/*. |
| 49 | These params can be used to set event defaults. |
| 50 | Here is a list of the params. |
| 51 | - 'period': Set event sampling period |
| 52 | |
| 53 | Note: If user explicitly sets options which conflict with the params, |
| 54 | the value set by the params will be overridden. |
| 55 | |
Jacob Shin | 3741eb9 | 2014-05-29 17:26:51 +0200 | [diff] [blame] | 56 | - a hardware breakpoint event in the form of '\mem:addr[/len][:access]' |
Frederic Weisbecker | 1b290d6 | 2009-11-23 15:42:35 +0100 | [diff] [blame] | 57 | where addr is the address in memory you want to break in. |
| 58 | Access is the memory access type (read, write, execute) it can |
Jacob Shin | 3741eb9 | 2014-05-29 17:26:51 +0200 | [diff] [blame] | 59 | be passed as follows: '\mem:addr[:[r][w][x]]'. len is the range, |
| 60 | number of bytes from specified addr, which the breakpoint will cover. |
Frederic Weisbecker | 1b290d6 | 2009-11-23 15:42:35 +0100 | [diff] [blame] | 61 | If you want to profile read-write accesses in 0x1000, just set |
| 62 | 'mem:0x1000:rw'. |
Jacob Shin | 3741eb9 | 2014-05-29 17:26:51 +0200 | [diff] [blame] | 63 | If you want to profile write accesses in [0x1000~1008), just set |
| 64 | 'mem:0x1000/8:w'. |
Shawn Bohrer | 08dbd7e | 2010-11-30 19:57:16 -0600 | [diff] [blame] | 65 | |
Namhyung Kim | 9a75606 | 2015-03-02 12:13:33 +0900 | [diff] [blame] | 66 | - a group of events surrounded by a pair of brace ("{event1,event2,...}"). |
| 67 | Each event is separated by commas and the group should be quoted to |
| 68 | prevent the shell interpretation. You also need to use --group on |
| 69 | "perf report" to view group events together. |
| 70 | |
Shawn Bohrer | 08dbd7e | 2010-11-30 19:57:16 -0600 | [diff] [blame] | 71 | --filter=<filter>:: |
Wang Nan | 4ba1faa | 2015-07-10 07:36:10 +0000 | [diff] [blame] | 72 | Event filter. This option should follow a event selector (-e) which |
| 73 | selects tracepoint event(s). Multiple '--filter' options are combined |
| 74 | using '&&'. |
| 75 | |
| 76 | --exclude-perf:: |
| 77 | Don't record events issued by perf itself. This option should follow |
| 78 | a event selector (-e) which selects tracepoint event(s). It adds a |
| 79 | filter expression 'common_pid != $PERFPID' to filters. If other |
| 80 | '--filter' exists, the new filter expression will be combined with |
| 81 | them by '&&'. |
Shawn Bohrer | 08dbd7e | 2010-11-30 19:57:16 -0600 | [diff] [blame] | 82 | |
Ingo Molnar | e33e0a4 | 2009-04-20 15:58:01 +0200 | [diff] [blame] | 83 | -a:: |
Shawn Bohrer | 08dbd7e | 2010-11-30 19:57:16 -0600 | [diff] [blame] | 84 | --all-cpus:: |
| 85 | System-wide collection from all CPUs. |
Ingo Molnar | e33e0a4 | 2009-04-20 15:58:01 +0200 | [diff] [blame] | 86 | |
Arnaldo Carvalho de Melo | 386c0b7 | 2009-08-05 10:04:53 -0300 | [diff] [blame] | 87 | -p:: |
| 88 | --pid=:: |
David Ahern | b52956c | 2012-02-08 09:32:52 -0700 | [diff] [blame] | 89 | Record events on existing process ID (comma separated list). |
Shawn Bohrer | 08dbd7e | 2010-11-30 19:57:16 -0600 | [diff] [blame] | 90 | |
| 91 | -t:: |
| 92 | --tid=:: |
David Ahern | b52956c | 2012-02-08 09:32:52 -0700 | [diff] [blame] | 93 | Record events on existing thread ID (comma separated list). |
Adrian Hunter | 69e7e5b | 2013-11-18 11:55:57 +0200 | [diff] [blame] | 94 | This option also disables inheritance by default. Enable it by adding |
| 95 | --inherit. |
Arnaldo Carvalho de Melo | 386c0b7 | 2009-08-05 10:04:53 -0300 | [diff] [blame] | 96 | |
Arnaldo Carvalho de Melo | 0d37aa3 | 2012-01-19 14:08:15 -0200 | [diff] [blame] | 97 | -u:: |
| 98 | --uid=:: |
| 99 | Record events in threads owned by uid. Name or number. |
| 100 | |
Arnaldo Carvalho de Melo | 386c0b7 | 2009-08-05 10:04:53 -0300 | [diff] [blame] | 101 | -r:: |
| 102 | --realtime=:: |
| 103 | Collect data with this RT SCHED_FIFO priority. |
Jiri Olsa | 563aecb | 2013-06-05 13:35:06 +0200 | [diff] [blame] | 104 | |
Arnaldo Carvalho de Melo | 509051e | 2014-01-14 17:52:14 -0300 | [diff] [blame] | 105 | --no-buffering:: |
Kirill Smelkov | acac03f | 2011-01-12 17:59:36 +0300 | [diff] [blame] | 106 | Collect data without buffering. |
Arnaldo Carvalho de Melo | 386c0b7 | 2009-08-05 10:04:53 -0300 | [diff] [blame] | 107 | |
Arnaldo Carvalho de Melo | 386c0b7 | 2009-08-05 10:04:53 -0300 | [diff] [blame] | 108 | -c:: |
| 109 | --count=:: |
| 110 | Event period to sample. |
| 111 | |
| 112 | -o:: |
| 113 | --output=:: |
| 114 | Output file name. |
| 115 | |
| 116 | -i:: |
Stephane Eranian | 2e6cdf9 | 2010-05-12 10:40:01 +0200 | [diff] [blame] | 117 | --no-inherit:: |
| 118 | Child tasks do not inherit counters. |
Arnaldo Carvalho de Melo | 386c0b7 | 2009-08-05 10:04:53 -0300 | [diff] [blame] | 119 | -F:: |
| 120 | --freq=:: |
| 121 | Profile at this frequency. |
| 122 | |
| 123 | -m:: |
| 124 | --mmap-pages=:: |
Jiri Olsa | 27050f5 | 2013-09-01 12:36:13 +0200 | [diff] [blame] | 125 | Number of mmap data pages (must be a power of two) or size |
| 126 | specification with appended unit character - B/K/M/G. The |
| 127 | size is rounded up to have nearest pages power of two value. |
Adrian Hunter | e9db131 | 2015-04-09 18:53:46 +0300 | [diff] [blame] | 128 | Also, by adding a comma, the number of mmap pages for AUX |
| 129 | area tracing can be specified. |
Arnaldo Carvalho de Melo | 386c0b7 | 2009-08-05 10:04:53 -0300 | [diff] [blame] | 130 | |
Namhyung Kim | 9a75606 | 2015-03-02 12:13:33 +0900 | [diff] [blame] | 131 | --group:: |
| 132 | Put all events in a single event group. This precedes the --event |
| 133 | option and remains only for backward compatibility. See --event. |
| 134 | |
Arnaldo Carvalho de Melo | 386c0b7 | 2009-08-05 10:04:53 -0300 | [diff] [blame] | 135 | -g:: |
Jiri Olsa | 09b0fd4 | 2013-10-26 16:25:33 +0200 | [diff] [blame] | 136 | Enables call-graph (stack chain/backtrace) recording. |
| 137 | |
Arnaldo Carvalho de Melo | 386c0b7 | 2009-08-05 10:04:53 -0300 | [diff] [blame] | 138 | --call-graph:: |
Jiri Olsa | 09b0fd4 | 2013-10-26 16:25:33 +0200 | [diff] [blame] | 139 | Setup and enable call-graph (stack chain/backtrace) recording, |
| 140 | implies -g. |
| 141 | |
| 142 | Allows specifying "fp" (frame pointer) or "dwarf" |
Kan Liang | aad2b21 | 2015-01-05 13:23:04 -0500 | [diff] [blame] | 143 | (DWARF's CFI - Call Frame Information) or "lbr" |
| 144 | (Hardware Last Branch Record facility) as the method to collect |
Jiri Olsa | 09b0fd4 | 2013-10-26 16:25:33 +0200 | [diff] [blame] | 145 | the information used to show the call graphs. |
| 146 | |
| 147 | In some systems, where binaries are build with gcc |
| 148 | --fomit-frame-pointer, using the "fp" method will produce bogus |
| 149 | call graphs, using "dwarf", if available (perf tools linked to |
| 150 | the libunwind library) should be used instead. |
Kan Liang | aad2b21 | 2015-01-05 13:23:04 -0500 | [diff] [blame] | 151 | Using the "lbr" method doesn't require any compiler options. It |
| 152 | will produce call graphs from the hardware LBR registers. The |
| 153 | main limition is that it is only available on new Intel |
| 154 | platforms, such as Haswell. It can only get user call chain. It |
| 155 | doesn't work with branch stack sampling at the same time. |
Arnaldo Carvalho de Melo | 386c0b7 | 2009-08-05 10:04:53 -0300 | [diff] [blame] | 156 | |
Arnaldo Carvalho de Melo | b44308f | 2010-10-26 15:20:09 -0200 | [diff] [blame] | 157 | -q:: |
| 158 | --quiet:: |
| 159 | Don't print any message, useful for scripting. |
| 160 | |
Arnaldo Carvalho de Melo | 386c0b7 | 2009-08-05 10:04:53 -0300 | [diff] [blame] | 161 | -v:: |
| 162 | --verbose:: |
| 163 | Be more verbose (show counter open errors, etc). |
| 164 | |
| 165 | -s:: |
| 166 | --stat:: |
Namhyung Kim | 1f91d5f | 2015-05-10 00:19:42 +0900 | [diff] [blame] | 167 | Record per-thread event counts. Use it with 'perf report -T' to see |
| 168 | the values. |
Arnaldo Carvalho de Melo | 386c0b7 | 2009-08-05 10:04:53 -0300 | [diff] [blame] | 169 | |
| 170 | -d:: |
| 171 | --data:: |
Peter Zijlstra | 5610032 | 2015-06-10 16:48:50 +0200 | [diff] [blame] | 172 | Record the sample addresses. |
Arnaldo Carvalho de Melo | 386c0b7 | 2009-08-05 10:04:53 -0300 | [diff] [blame] | 173 | |
Arnaldo Carvalho de Melo | 9c90a61 | 2010-12-02 10:25:28 -0200 | [diff] [blame] | 174 | -T:: |
| 175 | --timestamp:: |
Peter Zijlstra | 5610032 | 2015-06-10 16:48:50 +0200 | [diff] [blame] | 176 | Record the sample timestamps. Use it with 'perf report -D' to see the |
| 177 | timestamps, for instance. |
| 178 | |
| 179 | -P:: |
| 180 | --period:: |
| 181 | Record the sample period. |
Arnaldo Carvalho de Melo | 9c90a61 | 2010-12-02 10:25:28 -0200 | [diff] [blame] | 182 | |
Arnaldo Carvalho de Melo | 386c0b7 | 2009-08-05 10:04:53 -0300 | [diff] [blame] | 183 | -n:: |
| 184 | --no-samples:: |
| 185 | Don't sample. |
Ingo Molnar | e33e0a4 | 2009-04-20 15:58:01 +0200 | [diff] [blame] | 186 | |
Frederic Weisbecker | ec7ba4e | 2009-08-31 03:32:03 +0200 | [diff] [blame] | 187 | -R:: |
| 188 | --raw-samples:: |
Frederic Weisbecker | bdef3b0 | 2010-04-14 20:05:17 +0200 | [diff] [blame] | 189 | Collect raw sample records from all opened counters (default for tracepoint counters). |
Frederic Weisbecker | ec7ba4e | 2009-08-31 03:32:03 +0200 | [diff] [blame] | 190 | |
Stephane Eranian | c45c6ea | 2010-05-28 12:00:01 +0200 | [diff] [blame] | 191 | -C:: |
| 192 | --cpu:: |
Shawn Bohrer | 08dbd7e | 2010-11-30 19:57:16 -0600 | [diff] [blame] | 193 | Collect samples only on the list of CPUs provided. Multiple CPUs can be provided as a |
| 194 | comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2. |
Stephane Eranian | c45c6ea | 2010-05-28 12:00:01 +0200 | [diff] [blame] | 195 | In per-thread mode with inheritance mode on (default), samples are captured only when |
| 196 | the thread executes on the designated CPUs. Default is to monitor all CPUs. |
| 197 | |
Stephane Eranian | a1ac1d3 | 2010-06-17 11:39:01 +0200 | [diff] [blame] | 198 | -N:: |
| 199 | --no-buildid-cache:: |
Masanari Iida | 96355f2 | 2014-09-10 00:18:50 +0900 | [diff] [blame] | 200 | Do not update the buildid cache. This saves some overhead in situations |
Stephane Eranian | a1ac1d3 | 2010-06-17 11:39:01 +0200 | [diff] [blame] | 201 | where the information in the perf.data file (which includes buildids) |
| 202 | is sufficient. |
| 203 | |
Stephane Eranian | 023695d | 2011-02-14 11:20:01 +0200 | [diff] [blame] | 204 | -G name,...:: |
| 205 | --cgroup name,...:: |
| 206 | monitor only in the container (cgroup) called "name". This option is available only |
| 207 | in per-cpu mode. The cgroup filesystem must be mounted. All threads belonging to |
| 208 | container "name" are monitored when they run on the monitored CPUs. Multiple cgroups |
| 209 | can be provided. Each cgroup is applied to the corresponding event, i.e., first cgroup |
| 210 | to first event, second cgroup to second event and so on. It is possible to provide |
| 211 | an empty cgroup (monitor all the time) using, e.g., -G foo,,bar. Cgroups must have |
| 212 | corresponding events, i.e., they always refer to events defined earlier on the command |
| 213 | line. |
| 214 | |
Roberto Agostino Vitillo | bdfebd8 | 2012-02-09 23:21:02 +0100 | [diff] [blame] | 215 | -b:: |
Stephane Eranian | a5aabda | 2012-03-08 23:47:45 +0100 | [diff] [blame] | 216 | --branch-any:: |
| 217 | Enable taken branch stack sampling. Any type of taken branch may be sampled. |
| 218 | This is a shortcut for --branch-filter any. See --branch-filter for more infos. |
| 219 | |
| 220 | -j:: |
| 221 | --branch-filter:: |
Roberto Agostino Vitillo | bdfebd8 | 2012-02-09 23:21:02 +0100 | [diff] [blame] | 222 | Enable taken branch stack sampling. Each sample captures a series of consecutive |
| 223 | taken branches. The number of branches captured with each sample depends on the |
| 224 | underlying hardware, the type of branches of interest, and the executed code. |
| 225 | It is possible to select the types of branches captured by enabling filters. The |
| 226 | following filters are defined: |
| 227 | |
Stephane Eranian | a5aabda | 2012-03-08 23:47:45 +0100 | [diff] [blame] | 228 | - any: any type of branches |
Roberto Agostino Vitillo | bdfebd8 | 2012-02-09 23:21:02 +0100 | [diff] [blame] | 229 | - any_call: any function call or system call |
| 230 | - any_ret: any function return or system call return |
Anshuman Khandual | 2e49a94 | 2012-05-18 14:16:50 +0530 | [diff] [blame] | 231 | - ind_call: any indirect branch |
Roberto Agostino Vitillo | bdfebd8 | 2012-02-09 23:21:02 +0100 | [diff] [blame] | 232 | - u: only when the branch target is at the user level |
| 233 | - k: only when the branch target is in the kernel |
| 234 | - hv: only when the target is at the hypervisor level |
Andi Kleen | 0126d49 | 2013-09-20 07:40:42 -0700 | [diff] [blame] | 235 | - in_tx: only when the target is in a hardware transaction |
| 236 | - no_tx: only when the target is not in a hardware transaction |
| 237 | - abort_tx: only when the target is a hardware transaction abort |
Anshuman Khandual | 3e39db4 | 2014-05-22 12:50:10 +0530 | [diff] [blame] | 238 | - cond: conditional branches |
Roberto Agostino Vitillo | bdfebd8 | 2012-02-09 23:21:02 +0100 | [diff] [blame] | 239 | |
| 240 | + |
Anshuman Khandual | 3e39db4 | 2014-05-22 12:50:10 +0530 | [diff] [blame] | 241 | The option requires at least one branch type among any, any_call, any_ret, ind_call, cond. |
Masanari Iida | 9c76820 | 2012-11-30 14:10:25 +0900 | [diff] [blame] | 242 | The privilege levels may be omitted, in which case, the privilege levels of the associated |
Stephane Eranian | a5aabda | 2012-03-08 23:47:45 +0100 | [diff] [blame] | 243 | event are applied to the branch filter. Both kernel (k) and hypervisor (hv) privilege |
| 244 | levels are subject to permissions. When sampling on multiple events, branch stack sampling |
| 245 | is enabled for all the sampling events. The sampled branch type is the same for all events. |
| 246 | The various filters must be specified as a comma separated list: --branch-filter any_ret,u,k |
| 247 | Note that this feature may not be available on all processors. |
Roberto Agostino Vitillo | bdfebd8 | 2012-02-09 23:21:02 +0100 | [diff] [blame] | 248 | |
Andi Kleen | 0548429 | 2013-01-24 16:10:29 +0100 | [diff] [blame] | 249 | --weight:: |
| 250 | Enable weightened sampling. An additional weight is recorded per sample and can be |
| 251 | displayed with the weight and local_weight sort keys. This currently works for TSX |
| 252 | abort events and some memory events in precise mode on modern Intel CPUs. |
| 253 | |
Andi Kleen | 475eeab | 2013-09-20 07:40:43 -0700 | [diff] [blame] | 254 | --transaction:: |
| 255 | Record transaction flags for transaction related events. |
| 256 | |
Adrian Hunter | 3aa5939 | 2013-11-15 15:52:29 +0200 | [diff] [blame] | 257 | --per-thread:: |
| 258 | Use per-thread mmaps. By default per-cpu mmaps are created. This option |
| 259 | overrides that and uses per-thread mmaps. A side-effect of that is that |
| 260 | inheritance is automatically disabled. --per-thread is ignored with a warning |
| 261 | if combined with -a or -C options. |
Adrian Hunter | 539e6bb | 2013-11-01 15:51:34 +0200 | [diff] [blame] | 262 | |
Arnaldo Carvalho de Melo | a6205a3 | 2014-01-14 17:58:12 -0300 | [diff] [blame] | 263 | -D:: |
| 264 | --delay=:: |
Andi Kleen | 6619a53 | 2014-01-11 13:38:27 -0800 | [diff] [blame] | 265 | After starting the program, wait msecs before measuring. This is useful to |
| 266 | filter out the startup phase of the program, which is often very different. |
| 267 | |
Stephane Eranian | 4b6c517 | 2014-09-24 13:48:41 +0200 | [diff] [blame] | 268 | -I:: |
| 269 | --intr-regs:: |
| 270 | Capture machine state (registers) at interrupt, i.e., on counter overflows for |
| 271 | each sample. List of captured registers depends on the architecture. This option |
| 272 | is off by default. |
| 273 | |
Andi Kleen | 85c273d | 2015-02-24 15:13:40 -0800 | [diff] [blame] | 274 | --running-time:: |
| 275 | Record running and enabled time for read events (:S) |
| 276 | |
Peter Zijlstra | 814c8c3 | 2015-03-31 00:19:31 +0200 | [diff] [blame] | 277 | -k:: |
| 278 | --clockid:: |
| 279 | Sets the clock id to use for the various time fields in the perf_event_type |
| 280 | records. See clock_gettime(). In particular CLOCK_MONOTONIC and |
| 281 | CLOCK_MONOTONIC_RAW are supported, some events might also allow |
| 282 | CLOCK_BOOTTIME, CLOCK_REALTIME and CLOCK_TAI. |
| 283 | |
Adrian Hunter | 2dd6d8a | 2015-04-30 17:37:32 +0300 | [diff] [blame] | 284 | -S:: |
| 285 | --snapshot:: |
| 286 | Select AUX area tracing Snapshot Mode. This option is valid only with an |
| 287 | AUX area tracing event. Optionally the number of bytes to capture per |
| 288 | snapshot can be specified. In Snapshot Mode, trace data is captured only when |
| 289 | signal SIGUSR2 is received. |
| 290 | |
Kan Liang | 9d9cad7 | 2015-06-17 09:51:11 -0400 | [diff] [blame] | 291 | --proc-map-timeout:: |
| 292 | When processing pre-existing threads /proc/XXX/mmap, it may take a long time, |
| 293 | because the file may be huge. A time out is needed in such cases. |
| 294 | This option sets the time out limit. The default value is 500 ms. |
| 295 | |
Ingo Molnar | e33e0a4 | 2009-04-20 15:58:01 +0200 | [diff] [blame] | 296 | SEE ALSO |
| 297 | -------- |
Thomas Gleixner | 386b05e | 2009-06-06 14:56:33 +0200 | [diff] [blame] | 298 | linkperf:perf-stat[1], linkperf:perf-list[1] |