Difference between revisions of "Real Time in Embedded Linux System"
From IGEP - ISEE Wiki
(→Using high-resolution timers) |
Davidof1990 (talk | contribs) |
||
| (4 intermediate revisions by 2 users not shown) | |||
| Line 7: | Line 7: | ||
== Using high-resolution timers == | == Using high-resolution timers == | ||
| − | Have a look at the [http://downloads. | + | Have a look at the [http://downloads.isee.biz/pub/files/examples/rttest.c rttest.c] source file |
Now compile this program: | Now compile this program: | ||
| Line 93: | Line 93: | ||
=== Testing the Xenomai scheduling latency === | === Testing the Xenomai scheduling latency === | ||
| − | Test done with linux-2.6.35.y-rt branch from | + | Test done with linux-2.6.35.y-rt branch from http://git.isee.biz/?p=pub/scm/linux-omap-2.6.git;a=shortlog;h=refs/heads/linux-2.6.35.y-rt |
Recompile your kernel with CONFIG_XENOMAI enabled and test the program with nothing special | Recompile your kernel with CONFIG_XENOMAI enabled and test the program with nothing special | ||
| Line 139: | Line 139: | ||
* Other resources | * Other resources | ||
** http://www.osadl.org/Realtime-Linux.projects-realtime-linux.0.html | ** http://www.osadl.org/Realtime-Linux.projects-realtime-linux.0.html | ||
| − | ** https://rt.wiki.kernel.org/index. | + | ** https://rt.wiki.kernel.org/index.php/Main_Page |
** https://www.osadl.org/fileadmin/dam/rtlws/12/Brown.pdf | ** https://www.osadl.org/fileadmin/dam/rtlws/12/Brown.pdf | ||
** http://embedded.communities.intel.com/servlet/JiveServlet/download/4435-1248/xeno_intel.pdf | ** http://embedded.communities.intel.com/servlet/JiveServlet/download/4435-1248/xeno_intel.pdf | ||
| − | [[Category:Software]][[Category:How to forge]] | + | [[Category:Software distributions]][[Category:How to forge]][[Category:Software applications]] |
Latest revision as of 10:17, 1 July 2015
Contents
Real-time - Timers and scheduling latency
Based on free training materials from Free Electrons covering real-time (http://free-electrons.com/docs/realtime/)
The purpose of this article is measure scheduling latency in different conditions.
Using high-resolution timers
Have a look at the rttest.c source file
Now compile this program:
# gcc -o rttest rttest.c -lrt
Execute the program on the board
Testing the voluntary preemption kernel
Test done with linux-2.6.35.y branch from git://git.igep.es/pub/scm/linux-omap-2.6.git
Recompile your kernel with CONFIG_PREEMPT_VOLUNTARY enabled and test the program with nothing special
Test the program with nothing special
# ./rttest Clock resolution (ns): 1 Measurement, please wait 1 minute... Samples: 153046 Min latency: 113 us Max latency: 7712 us Average latency: 292 us
Test your program and at the same time, add some workload to the board, by running
# stress --cpu 1 --io 2 --vm 2 --vm-bytes 128M --timeout 120s & # ./rttest Clock resolution (ns): 1 Measurement, please wait 1 minute... Samples: 90233 Min latency: 83 us Max latency: 28586 us Average latency: 558 us
Test your program again with the workload, but by running the program in the SCHED_FIFO scheduling class at priority 99
# stress --cpu 1 --io 2 --vm 2 --vm-bytes 128M --timeout 120s & # chrt -f -r 99 ./rttest Clock resolution (ns): 1 Measurement, please wait 1 minute... Samples: 210637 Min latency: 52 us Max latency: 2219 us Average latency: 181 us
Testing the preemptible kernel
Test done with linux-2.6.35.y branch from git://git.igep.es/pub/scm/linux-omap-2.6.git
Recompile your kernel with CONFIG_PREEMPT enabled and test the program with nothing special
# ./rttest Clock resolution (ns): 1 Measurement, please wait 1 minute... Samples: 154210 Min latency: 113 us Max latency: 9085 us Average latency: 288 us
Test your program and at the same time, add some workload to the board, by running
# stress --cpu 1 --io 2 --vm 2 --vm-bytes 128M --timeout 120s & # ./rttest Clock resolution (ns): 1 Measurement, please wait 1 minute... Samples: 113158 Min latency: 83 us Max latency: 15921 us Average latency: 421 us
Test your program again with the workload, but by running the program in the SCHED_FIFO scheduling class at priority 99
# stress --cpu 1 --io 2 --vm 2 --vm-bytes 128M --timeout 120s & # chrt -f -r 99 ./rttest Clock resolution (ns): 1 Measurement, please wait 1 minute... Samples: 235302 Min latency: 83 us Max latency: 296 us Average latency: 150 us
Testing the Xenomai scheduling latency
Test done with linux-2.6.35.y-rt branch from http://git.isee.biz/?p=pub/scm/linux-omap-2.6.git;a=shortlog;h=refs/heads/linux-2.6.35.y-rt
Recompile your kernel with CONFIG_XENOMAI enabled and test the program with nothing special
# ./rttest Clock resolution (ns): 1 Measurement, please wait 1 minute... Samples: 327833 Min latency: 73 us Max latency: 9702 us Average latency: 80 us
Test your program and at the same time, add some workload to the board, by running
# stress --cpu 1 --io 2 --vm 2 --vm-bytes 128M --timeout 120s & # ./rttest Clock resolution (ns): 1 Measurement, please wait 1 minute... Samples: 320301 Min latency: 63 us Max latency: 10006 us Average latency: 85 us
Test your program again with the workload, but by running the program in the SCHED_FIFO scheduling class at priority 99
# stress --cpu 1 --io 2 --vm 2 --vm-bytes 128M --timeout 120s & # chrt -f -r 99 ./rttest Clock resolution (ns): 1 Measurement, please wait 1 minute... Samples: 453705 Min latency: 18 us Max latency: 395 us Average latency: 30 us
Links
- Tools to ensure a system's realtime capabilities
- https://www.osadl.org/Single-View.111+M52212cb1379.0.html
- https://www.osadl.org/Realtime-Preempt-Kernel.kernel-rt.0.html
- https://www.osadl.org/Driver-to-Program-Signal-Tester.driver-to-program-signal-tester.0.html
- https://www.osadl.org/Latency-generator.latency-generator.0.html
- https://www.osadl.org/Visualize-the-temporal-relationship-of-L.taks-visualizer.0.html
- http://elinux.org/Realtime_Testing_Best_Practices
- Other resources
