Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/tycho/clockperf
clocksource behavior microbenchmarks
https://github.com/tycho/clockperf
Last synced: 2 months ago
JSON representation
clocksource behavior microbenchmarks
- Host: GitHub
- URL: https://github.com/tycho/clockperf
- Owner: tycho
- License: isc
- Created: 2016-04-15T14:11:21.000Z (almost 9 years ago)
- Default Branch: master
- Last Pushed: 2023-06-01T20:13:01.000Z (over 1 year ago)
- Last Synced: 2024-08-03T01:25:49.792Z (6 months ago)
- Language: C
- Size: 172 KB
- Stars: 27
- Watchers: 5
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: COPYING
Awesome Lists containing this project
README
clockperf [](https://github.com/tycho/clockperf/actions) [](https://opensource.org/licenses/ISC)
=======
This is a small utility for measuring clocksource behavior on Linux, BSDs, OS
X, and Windows.
Building
--------
The build system is super straightforward:
```
$ make
```
Output Format
--------------
The **Cost(ns)** column is the time cost to read that clock in nanoseconds, as
measured by the TSC (except when measuring the cost of reading the TSC itself,
which is measured by looking at gettimeofday).
**Resol** is the observable tick rate of the clock. Note that if the clock
is so high resolution that we see a distinct value on every read, then the
observable resolution would merely be a function of how fast we could read it.
For this reason, a value of "----" indicates that the clock advances too
quickly for the resolution to be estimated with any precision.
**Mono** indicates whether the clock source is monotonic, with an additional
restriction. Not only must the clock only move forward, it must never return
the same value (i.e. high frequency).
**Fail** indicates the number of times the clock source failed to advance in >=
200 reads.
**Warp** indicates the number of times the clock jumped forward by an
unexpectedly large span (threshold for this is currently 1ms).
**Stal** indicates the number of times the clock source failed to advance in
more than 1, but less than 200 reads.
**Regr** indicates the number of times the clock moved backwards. This can
happen due to hypervisor behavior, NTP adjustments, and other clock changes.
It's extremely bad behavior if you use the clock source for timespan
measurement (e.g. profiling, benchmarks, etc).
Example Runs
------------
Linux 4.5.0 on x86_64 with
`/sys/devices/system/clocksource/clocksource0/current_clocksource` set to
`tsc`:
```
$ ./clockperf
clockperf v1.0.0
Invariant TSC: Yes
tsc resolution = 2401MHz
realtime resolution = 1000MHz
realtime_crs resolution = 100Hz
monotonic resolution = 1000MHz
monotonic_crs resolution = 100Hz
monotonic_raw resolution = 1000MHz
boottime resolution = 1000MHz
process resolution = 1000MHz
thread resolution = 1000MHz
clock resolution = 1000KHz
time resolution = 1Hz
Name Cost(ns) +/- Resol Mono Fail Warp Stal Regr
tsc 15.93 0.55% ---- Yes 0 0 0 0
gettimeofday 22.87 0.07% 1000KHz No 0 0 999 0
realtime 23.69 0.16% ---- Yes 0 0 0 0
realtime_crs 9.68 0.35% 100Hz No 999 0 0 0
8.34 84.91%
monotonic 23.40 0.10% ---- Yes 0 0 0 0
monotonic_crs 9.03 0.63% 100Hz No 999 0 0 0
8.34 84.91%
monotonic_raw 73.34 0.08% ---- Yes 0 0 0 0
boottime 77.66 0.17% ---- Yes 0 0 0 0
process 132.47 0.08% 143MHz No 0 0 0 0
thread 125.87 0.04% 1000MHz No 0 0 0 0
clock 134.26 0.09% 1000KHz No 0 0 999 0
getrusage 210.87 0.06% 100Hz No 995 4 4 0
ftime 27.86 0.03% 1000Hz No 994 0 5 0
time 5.36 0.05% 1Hz No 1000 0 0 0
```
Linux 4.5.0 on x86_64 with
`/sys/devices/system/clocksource/clocksource0/current_clocksource` set to
`hpet`:
```
$ ./clockperf
clockperf v1.0.0
Invariant TSC: Yes
tsc resolution = 2402MHz
realtime resolution = 1000MHz
realtime_crs resolution = 100Hz
monotonic resolution = 1000MHz
monotonic_crs resolution = 100Hz
monotonic_raw resolution = 1000MHz
boottime resolution = 1000MHz
process resolution = 1000MHz
thread resolution = 1000MHz
clock resolution = 1000KHz
time resolution = 1Hz
Name Cost(ns) +/- Resol Mono Fail Warp Stal Regr
tsc 16.13 0.51% ---- Yes 0 0 0 0
gettimeofday 564.88 0.13% 1000KHz No 0 0 292 0
realtime 567.63 0.94% ---- Yes 0 0 0 0
realtime_crs 9.71 0.55% 100Hz No 999 0 0 0
8.34 84.91%
monotonic 561.19 0.06% ---- Yes 0 0 0 0
monotonic_crs 8.84 0.49% 100Hz No 999 0 0 0
8.34 84.91%
monotonic_raw 618.30 0.82% ---- Yes 0 0 0 0
boottime 619.19 0.12% ---- Yes 0 0 0 0
process 132.17 0.25% 56MHz No 0 0 0 0
thread 125.85 0.07% 40MHz No 0 0 0 0
clock 134.24 0.11% 1000KHz No 0 0 999 0
getrusage 210.51 0.11% 100Hz No 995 4 4 0
ftime 570.64 0.28% 1000Hz No 888 0 111 0
time 5.36 0.45% 1Hz No 1000 0 0 0
```
Linux 3.10.40 on ARMv7a Tegra TK1 Jetson board with
`/sys/devices/system/clocksource/clocksource0/available_clocksource` set to
`arch_sys_counter`:
```
$ ./clockperf
clockperf v1.0.0
realtime resolution = 1000MHz
realtime_crs resolution = 1000Hz
monotonic resolution = 1000MHz
monotonic_crs resolution = 1000Hz
monotonic_raw resolution = 1000MHz
boottime resolution = 1000MHz
process resolution = 1000MHz
thread resolution = 1000MHz
clock resolution = 1000KHz
time resolution = 1Hz
Name Cost(ns) +/- Resol Mono Fail Warp Stal Regr
gettimeofday 681.59 20.29% ---- No 0 0 848 0
681.59 20.29%
realtime 266.57 0.18% ---- Yes 0 0 0 0
realtime_crs 234.29 0.08% 1000Hz No 954 0 45 0
monotonic 269.15 0.41% ---- Yes 0 0 0 0
monotonic_crs 244.34 0.14% 1000Hz No 952 0 46 0
monotonic_raw 282.09 0.18% ---- Yes 0 0 0 0
boottime 271.64 0.07% ---- Yes 0 0 0 0
process 719.18 0.22% ---- Yes 0 0 0 0
thread 635.93 0.34% ---- Yes 0 0 0 0
clock 728.26 0.08% 1000KHz No 0 0 225 0
getrusage 867.63 0.23% 1000Hz No 833 0 166 0
ftime 287.70 0.04% 1000Hz No 944 0 55 0
time 265.54 0.22% 1Hz No 999 0 0 0
166.77 204.50%
```