1CYCLICTEST(8) System Manager's Manual CYCLICTEST(8)
2
6 cyclictest - High resolution test program
7
9 cyclictest [ -hfmnqrsvMS ] [-a proc ] [-A align ] [-b usec ] [-c clock
10 ] [-d dist ] [-h histogram ] [-i intv ] [--json filename ]
11 [-l loop ] [-o red ] [-p prio ] [-t num ] [-D time] [-w]
12 [-W] [-y policy ] [ -S | -U ]
13
16 These programs follow the usual GNU command line syntax, with long
17 options starting with two dashes ('--').
18 A summary of options is included below.
19 -a, --affinity[=PROC-SET]
21 Run threads on the set of processors given by PROC-SET. If
22 PROC-SET is not specified, all processors will be used. Threads
23 will be assigned to processors in the set in numeric order, in a
24 round-robin fashion.
25 The set of processors can be specified as A,B,C, or A-C, or A-
26 B,D-F, and so on*. The ! character can be used to negate a set.
27 For example, !B-D means to use all available CPUs except B
28 through D. The cpu numbers are the same as shown in the
29 processor field in /proc/cpuinfo. See numa(3) for more
30 information on specifying CPU sets. * Support for CPU sets
31 requires libnuma version >= 2. For libnuma v1, PROC-SET, if
32 specified, must be a single CPU number.
33 -A, --align=USEC
35 Align thread wakeups to a specific offset in microseconds
36 -b, --breaktrace=USEC
38 Send break trace command when latency > USEC
39 -c, --clock=CLOCK
41 select clock
42 0 = CLOCK_MONOTONIC (default)
43 1 = CLOCK_REALTIME
44 --default-system
46 Don't attempt to tune the system from cyclictest. Power
47 management is not suppressed. This might give poorer results,
48 but will allow you to discover if you need to tune the system.
49 -d, --distance=DIST
51 Distance of thread intervals in us, default = 500
52 -D, --duration=TIME
54 Specify a length for the test run.
55 Append 'm', 'h', or 'd' to specify minutes, hours or days.
56 -F, --fifo=<path>
58 Create a named pipe at path and write stats to it
59 -h, --histogram=US
61 Dump latency histogram to stdout after the run. US is the max
62 latency time to be be tracked in microseconds. This option runs
63 all threads at the same priority.
64 -H, --histofall=MAXLATENCYINUS
66 Same as -h except that an additional histogram column is
67 displayed at the right that contains summary data of all thread
68 histograms. If cyclictest runs a single thread only, the -H
69 option is equivalent to -h.
70 --histfile=<path>
72 Dump the latency histogram to <path> instead of stdout.
73 -i, --interval=INTV
75 Set the base interval of the thread(s) in microseconds (default
76 is 1000us). This sets the interval of the first thread. See also
77 -d.
78 --json=FILENAME
80 Write final results into FILENAME, JSON formatted.
81 --laptop
83 Save battery when running cyclictest. This will give you poorer
84 realtime results, but will not drain your battery so quickly.
85 --latency=PM_Q0S
87 power management latency target value. This value is written to
88 /dev/cpu_dma_latency and affects c-states. The default is 0
89 -l, --loops=LOOPS
91 Set the number of loops. The default is 0 (endless). This option
92 is useful for automated tests with a given number of test
93 cycles. Cyclictest is stopped once the number of timer intervals
94 has been reached.
95 --mainaffinity=CPUSET
97 Run the main thread on CPU #N. This only affects the main thread
98 and not the measurement threads
99 -m, --mlockall
101 Lock current and future memory allocations to prevent being
102 paged out
103 -M, --refresh_on_max
105 Delay updating the screen until a new max latency is hit.
106 (useful for running cyclictest on low-bandwidth connections)
107 -N, --nsecs
109 Show results in nanoseconds instead of microseconds, which is
110 the default unit.
111 -o, --oscope=RED
113 Oscilloscope mode, reduce verbose output by RED.
114 -p, --prio=PRIO
116 Set the priority of the first thread. The given priority is set
117 to the first test thread. Each further thread gets a lower
118 priority: Priority(Thread N) = max(Priority(Thread N-1) - 1, 0)
119 --policy=NAME
121 set the scheduler policy of the measurement threads where NAME
122 is one of: other, normal, batch, idle, fifo, rr
123 --priospread
125 spread priority levels starting at a specified value
126 -q, --quiet
128 Print a summary only on exit. Useful for automated tests, where
129 only the summary output needs to be captured.
130 -r, --relative
132 Use relative timers instead of absolute. The default behaviour
133 of the tests is to use absolute timers. This option is there for
134 completeness and should not be used for reproducible tests.
135 -R, --resolution
137 Check clock resolution, calling clock_gettime() many times. List
138 of lock_gettime() values will be reported with -X
139 --secaligned [USEC]
141 align thread wakeups to the next full second and apply the
142 optional offset.
143 -s, --system
145 Use sys_nanosleep and sys_setitimer instead of posix timers.
146 Note, that -s can only be used with one thread because itimers
147 are per process and not per thread. -s uses the nanosleep
148 syscall and is not restricted to one thread.
149 -S, --smp
151 Set options for standard testing on SMP systems. Equivalent to
152 using the options: "-t -a" as well keeping any specified
153 priority equal across all threads
154 --spike=<trigger>
156 record all spikes > trigger
157 --spike-nodes=[num of nodes]
159 These are the maximum number of spikes we can record.
160 The default is 1024 if not specified.
161 --smi Enable SMI count/detection on processors with SMI count support.
163 -t, --threads[=NUM]
165 Set the number of test threads (default is 1). Create NUM test
166 threads. If NUM is not specified, NUM is set to the number of
167 available CPUs. See -d, -i and -p for further information.
168 --tracemark
170 write a trace mark when -b latency is exceeded.
171 -u, --unbuffered
173 force unbuffered output for live processing
174 -v, --verbose
176 Output values on stdout for statistics. This option is used to
177 gather statistical information about the latency distribution.
178 The output is sent to stdout. The output format is:
179 n:c:v
181 where n=task number c=count v=latency value in us.
183 --dbg_cyclictest
185 Print info userful for debugging cyclictest
186 -x, --posix_timers
188 Use POSIX timers instead of clock_nanosleep.
189
192 numa(3), numactl(8),
193
195 cyclictest was written by Thomas Gleixner <tglx@linuxtronix.de>.
196 This manual page was written by Alessio Igor Bogani
198 <abogani@texware.it>, for the Debian project (but may be used by
199 others).
200 Updated by John Kacur <jkacur@redhat.com>
201 April 22, 2016 CYCLICTEST(8)