6 perf-sched - Tool to trace/measure scheduler properties (latencies)
11 'perf sched' {record|latency|map|replay|script|timehist}
15 There are several variants of 'perf sched':
17 'perf sched record <command>' to record the scheduling events
18 of an arbitrary workload.
20 'perf sched latency' to report the per task scheduling latencies
21 and other scheduling properties of the workload.
23 'perf sched script' to see a detailed trace of the workload that
24 was recorded (aliased to 'perf script' for now).
26 'perf sched replay' to simulate the workload that was recorded
27 via perf sched record. (this is done by starting up mockup threads
28 that mimic the workload based on the events in the trace. These
29 threads can then replay the timings (CPU runtime and sleep patterns)
30 of the workload as it occurred when it was recorded - and can repeat
31 it a number of times, measuring its performance.)
33 'perf sched map' to print a textual context-switching outline of
34 workload captured via perf sched record. Columns stand for
35 individual CPUs, and the two-letter shortcuts stand for tasks that
36 are running on a CPU. A '*' denotes the CPU that had the event, and
37 a dot signals an idle CPU.
39 'perf sched timehist' provides an analysis of scheduling events.
42 perf sched record -- sleep 1
45 By default it shows the individual schedule events, including the wait
46 time (time between sched-out and next sched-in events for the task), the
47 task scheduling delay (time between wakeup and actually running) and run
50 time cpu task name wait time sch delay run time
51 [tid/pid] (msec) (msec) (msec)
52 -------------- ------ -------------------- --------- --------- ---------
53 79371.874569 [0011] gcc[31949] 0.014 0.000 1.148
54 79371.874591 [0010] gcc[31951] 0.000 0.000 0.024
55 79371.874603 [0010] migration/10[59] 3.350 0.004 0.011
56 79371.874604 [0011] <idle> 1.148 0.000 0.035
57 79371.874723 [0005] <idle> 0.016 0.000 1.383
58 79371.874746 [0005] gcc[31949] 0.153 0.078 0.022
61 Times are in msec.usec.
67 Input file name. (default: perf.data unless stdin is a fifo)
71 Be more verbose. (show symbol address, etc)
75 Display verbose dump of the sched data.
79 Don't complain, do it.
81 OPTIONS for 'perf sched map'
82 ----------------------------
85 Show only CPUs with activity. Helps visualizing on high core
89 Show just entries with activities for the given CPUs.
92 Highlight the given cpus.
95 Highlight the given pids.
97 OPTIONS for 'perf sched timehist'
98 ---------------------------------
108 Display call chains if present (default on).
111 Maximum number of functions to display in backtrace, default 5.
115 Only show events for the given CPU(s) (comma separated list).
119 Only show events for given process ID (comma separated list).
123 Only show events for given thread ID (comma separated list).
127 Show only a summary of scheduling by thread with min, max, and average
128 run times (in sec) and relative stddev.
132 Show all scheduling events followed by a summary by thread with min,
133 max, and average run times (in sec) and relative stddev.
135 --symfs=<directory>::
136 Look for files with symbols relative to this directory.
140 Show visual aid for sched switches by CPU: 'i' marks idle time,
141 's' are scheduler events.
149 Show migration events.
157 Show idle-related events only.
160 Only analyze samples within given time window: <start>,<stop>. Times
161 have the format seconds.microseconds. If start is not given (i.e., time
162 string is ',x.y') then analysis starts at the beginning of the file. If
163 stop time is not given (i.e, time string is 'x.y,') then analysis goes
167 Show task state when it switched out.
171 linkperf:perf-record[1]