-
Notifications
You must be signed in to change notification settings - Fork 0
/
README
58 lines (37 loc) · 1.79 KB
/
README
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
sysjitter v1.4
Copyright 2010-2017 David Riddoch <[email protected]>
sysjitter measures the extent to which the system impacts on user-level
code by causing jitter. It runs a thread on each processor core, and when
the thread is "knocked off" the core it measures how long for. At the end
of the run it outputs some summary statistics for each core, and
optionally the full raw data.
Prerequisites
-------------
- Supports x86, x86_64 and Power processors.
- Requires a version of gcc with support for built-in atomics.
Building
--------
make
Running
-------
It is important to run sysjitter on cores that are idle. If running on a
system with cpusets enabled, then run sysjitter as root.
sysjitter <threshold-nsec>
The argument <threshold-nsec> tells sysjitter to ignore any
"interruptions" shorter than the value given. Something upwards of a few
hundred nanoseconds is a reasonable choice here.
The default output gives summary statistics for each CPU core. Pipe the
output through "column -t" to make it a little easier on the eye.
You can get the raw data out with the --raw option. This includes the
timestamp and duration of each interruption on each CPU core. The data
is placed in a separate file for each CPU core.
By default it will run for 70 seconds. Override with the --runtime
option. e.g. For a 10 second run:
sysjitter --runtime 10 300
You can run on a specified subset of cores, given as a comma separated
list of cores and/or ranges. This example runs on cores 2, 4, 5 and 6
and reports interruptions of at least 20us:
sysjitter --cores 2,4-6 20000
Note that the cpu_mhz row is misnamed; this is actually the measured tick
rate of the CPU timestamp counter (in MHz), which may or may not be the
same as the clock frequency.