129 lines · plain
1perf-timechart(1)2=================3 4NAME5----6perf-timechart - Tool to visualize total system behavior during a workload7 8SYNOPSIS9--------10[verse]11'perf timechart' [<timechart options>] {record} [<record options>]12 13DESCRIPTION14-----------15There are two variants of perf timechart:16 17 'perf timechart record <command>' to record the system level events18 of an arbitrary workload. By default timechart records only scheduler19 and CPU events (task switches, running times, CPU power states, etc),20 but it's possible to record IO (disk, network) activity using -I argument.21 22 'perf timechart' to turn a trace into a Scalable Vector Graphics file,23 that can be viewed with popular SVG viewers such as 'Inkscape'. Depending24 on the events in the perf.data file, timechart will contain scheduler/cpu25 events or IO events.26 27 In IO mode, every bar has two charts: upper and lower.28 Upper bar shows incoming events (disk reads, ingress network packets).29 Lower bar shows outgoing events (disk writes, egress network packets).30 There are also poll bars which show how much time application spent31 in poll/epoll/select syscalls.32 33TIMECHART OPTIONS34-----------------35-o::36--output=::37 Select the output file (default: output.svg)38-i::39--input=::40 Select the input file (default: perf.data unless stdin is a fifo)41-w::42--width=::43 Select the width of the SVG file (default: 1000)44-P::45--power-only::46 Only output the CPU power section of the diagram47-T::48--tasks-only::49 Don't output processor state transitions50-p::51--process::52 Select the processes to display, by name or PID53-f::54--force::55 Don't complain, do it.56--symfs=<directory>::57 Look for files with symbols relative to this directory.58-n::59--proc-num::60 Print task info for at least given number of tasks.61-t::62--topology::63 Sort CPUs according to topology.64--highlight=<duration_nsecs|task_name>::65 Highlight tasks (using different color) that run more than given66 duration or tasks with given name. If number is given it's interpreted67 as number of nanoseconds. If non-numeric string is given it's68 interpreted as task name.69--io-skip-eagain::70 Don't draw EAGAIN IO events.71--io-min-time=<nsecs>::72 Draw small events as if they lasted min-time. Useful when you need73 to see very small and fast IO. It's possible to specify ms or us74 suffix to specify time in milliseconds or microseconds.75 Default value is 1ms.76--io-merge-dist=<nsecs>::77 Merge events that are merge-dist nanoseconds apart.78 Reduces number of figures on the SVG and makes it more render-friendly.79 It's possible to specify ms or us suffix to specify time in80 milliseconds or microseconds.81 Default value is 1us.82 83RECORD OPTIONS84--------------85-P::86--power-only::87 Record only power-related events88-T::89--tasks-only::90 Record only tasks-related events91-I::92--io-only::93 Record only io-related events94-g::95--callchain::96 Do call-graph (stack chain/backtrace) recording97 98EXAMPLES99--------100 101$ perf timechart record git pull102 103 [ perf record: Woken up 13 times to write data ]104 [ perf record: Captured and wrote 4.253 MB perf.data (~185801 samples) ]105 106$ perf timechart107 108 Written 10.2 seconds of trace to output.svg.109 110Record system-wide timechart:111 112 $ perf timechart record113 114 then generate timechart and highlight 'gcc' tasks:115 116 $ perf timechart --highlight gcc117 118Record system-wide IO events:119 120 $ perf timechart record -I121 122 then generate timechart:123 124 $ perf timechart125 126SEE ALSO127--------128linkperf:perf-record[1]129