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