brintos

brintos / linux-shallow public Read only

0
0
Text · 7.2 KiB · 3db6495 Raw
234 lines · plain
1perf-sched(1)2=============3 4NAME5----6perf-sched - Tool to trace/measure scheduler properties (latencies)7 8SYNOPSIS9--------10[verse]11'perf sched' {record|latency|map|replay|script|timehist}12 13DESCRIPTION14-----------15There are several variants of 'perf sched':16 17  'perf sched record <command>' to record the scheduling events18  of an arbitrary workload.19 20  'perf sched latency' to report the per task scheduling latencies21  and other scheduling properties of the workload.22 23   Example usage:24       perf sched record -- sleep 125       perf sched latency26 27  -------------------------------------------------------------------------------------------------------------------------------------------28  Task                  |   Runtime ms  |  Count   | Avg delay ms    | Max delay ms    | Max delay start           | Max delay end          |29  -------------------------------------------------------------------------------------------------------------------------------------------30  perf:(2)              |      2.804 ms |       66 | avg:   0.524 ms | max:   1.069 ms | max start: 254752.314960 s | max end: 254752.316029 s31  NetworkManager:1343   |      0.372 ms |       13 | avg:   0.008 ms | max:   0.013 ms | max start: 254751.551153 s | max end: 254751.551166 s32  kworker/1:2-xfs:4649  |      0.012 ms |        1 | avg:   0.008 ms | max:   0.008 ms | max start: 254751.519807 s | max end: 254751.519815 s33  kworker/3:1-xfs:388   |      0.011 ms |        1 | avg:   0.006 ms | max:   0.006 ms | max start: 254751.519809 s | max end: 254751.519815 s34  sleep:147736          |      0.938 ms |        3 | avg:   0.006 ms | max:   0.007 ms | max start: 254751.313817 s | max end: 254751.313824 s35 36  It shows Runtime(time that a task spent actually running on the CPU),37  Count(number of times a delay was calculated) and delay(time that a38  task was ready to run but was kept waiting).39 40  Tasks with the same command name are merged and the merge count is41  given within (), However if -p option is used, pid is mentioned.42 43  'perf sched script' to see a detailed trace of the workload that44   was recorded (aliased to 'perf script' for now).45 46  'perf sched replay' to simulate the workload that was recorded47  via perf sched record. (this is done by starting up mockup threads48  that mimic the workload based on the events in the trace. These49  threads can then replay the timings (CPU runtime and sleep patterns)50  of the workload as it occurred when it was recorded - and can repeat51  it a number of times, measuring its performance.)52 53  'perf sched map' to print a textual context-switching outline of54  workload captured via perf sched record.  Columns stand for55  individual CPUs, and the two-letter shortcuts stand for tasks that56  are running on a CPU. A '*' denotes the CPU that had the event, and57  a dot signals an idle CPU.58 59  'perf sched timehist' provides an analysis of scheduling events.60    61    Example usage:62        perf sched record -- sleep 163        perf sched timehist64    65   By default it shows the individual schedule events, including the wait66   time (time between sched-out and next sched-in events for the task), the67   task scheduling delay (time between runnable and actually running) and68   run time for the task:69    70                time    cpu  task name             wait time  sch delay   run time71                             [tid/pid]                (msec)     (msec)     (msec)72      -------------- ------  --------------------  ---------  ---------  ---------73        79371.874569 [0011]  gcc[31949]                0.014      0.000      1.14874        79371.874591 [0010]  gcc[31951]                0.000      0.000      0.02475        79371.874603 [0010]  migration/10[59]          3.350      0.004      0.01176        79371.874604 [0011]  <idle>                    1.148      0.000      0.03577        79371.874723 [0005]  <idle>                    0.016      0.000      1.38378        79371.874746 [0005]  gcc[31949]                0.153      0.078      0.02279    ...80    81   Times are in msec.usec.82 83OPTIONS84-------85-i::86--input=<file>::87        Input file name. (default: perf.data unless stdin is a fifo)88 89-v::90--verbose::91        Be more verbose. (show symbol address, etc)92 93-D::94--dump-raw-trace=::95        Display verbose dump of the sched data.96 97-f::98--force::99	Don't complain, do it.100 101OPTIONS for 'perf sched latency'102-------------------------------103 104-C::105--CPU <n>::106        CPU to profile on.107 108-p::109--pids::110        latency stats per pid instead of per command name.111 112-s::113--sort <key[,key2...]>::114        sort by key(s): runtime, switch, avg, max115        by default it's sorted by "avg ,max ,switch ,runtime".116 117OPTIONS for 'perf sched map'118----------------------------119 120--compact::121	Show only CPUs with activity. Helps visualizing on high core122	count systems.123 124--cpus::125	Show just entries with activities for the given CPUs.126 127--color-cpus::128	Highlight the given cpus.129 130--color-pids::131	Highlight the given pids.132 133--task-name <task>::134	Map output only for the given task name(s). Separate the135	task names with a comma (without whitespace). The sched-out136	time is printed and is represented by '*-' for the given137	task name(s).138	('-' indicates other tasks while '.' is idle).139 140--fuzzy-name::141	Given task name(s) can be partially matched (fuzzy matching).142 143OPTIONS for 'perf sched timehist'144---------------------------------145-k::146--vmlinux=<file>::147    vmlinux pathname148 149--kallsyms=<file>::150    kallsyms pathname151 152-g::153--call-graph::154	Display call chains if present (default on).155 156--max-stack::157	Maximum number of functions to display in backtrace, default 5.158 159-C=::160--cpu=::161	Only show events for the given CPU(s) (comma separated list).162 163-p=::164--pid=::165	Only show events for given process ID (comma separated list).166 167-t=::168--tid=::169	Only show events for given thread ID (comma separated list).170 171-s::172--summary::173    Show only a summary of scheduling by thread with min, max, and average174    run times (in sec) and relative stddev.175 176-S::177--with-summary::178    Show all scheduling events followed by a summary by thread with min,179    max, and average run times (in sec) and relative stddev.180 181--symfs=<directory>::182    Look for files with symbols relative to this directory.183 184-V::185--cpu-visual::186	Show visual aid for sched switches by CPU: 'i' marks idle time,187	's' are scheduler events.188 189-w::190--wakeups::191	Show wakeup events.192 193-M::194--migrations::195	Show migration events.196 197-n::198--next::199	Show next task.200 201-I::202--idle-hist::203	Show idle-related events only.204 205--time::206	Only analyze samples within given time window: <start>,<stop>. Times207	have the format seconds.microseconds. If start is not given (i.e., time208	string is ',x.y') then analysis starts at the beginning of the file. If209	stop time is not given (i.e, time string is 'x.y,') then analysis goes210	to end of file.211 212--state::213	Show task state when it switched out.214 215--show-prio::216	Show task priority.217 218--prio::219	Only show events for given task priority(ies). Multiple priorities can be220	provided as a comma-separated list with no spaces: 0,120. Ranges of221	priorities are specified with -: 120-129. A combination of both can also be222	provided: 0,120-129.223 224OPTIONS for 'perf sched replay'225------------------------------226 227-r::228--repeat <n>::229	repeat the workload n times (0: infinite). Default is 10.230 231SEE ALSO232--------233linkperf:perf-record[1]234