Panagiotis Issaris panagiotis.issaris at mech.kuleuven.ac.be
Sun Oct 17 17:20:27 CEST 2004


On Sun, Oct 17, 2004 at 12:58:04PM +0200 or thereabouts, Philippe Gerum wrote:
> Hi,
> Running LiveCD 0.5, I noticed the following:
> o preempt and switch test entries are swapped for both UP and LXRT modes
> (e.g. #2 should be #3, #5 should be #6 and conversely).
> o Running the switches test does not bring much additional information,
> since preempt uses semaphores too and obviously does task switching
> heavily, so the figures of the former are part of the latter. If the
> LiveCD is about testing the sanity of one's hw configuration and
> detecting local RTAI software mis-configurations by comparing with known
> sane settings, having details such as "you've taken 987 ns to switch
> stacks" is basically useless IMHO. 

In the upcoming 0.6, I've removed both switches tests.  

Besides the jitter of the fast and slow tasks, are the other shown values 
(being min, avg, max latency) useful or are they somewhat redundant
because of the latency test?   

> o I need to hit the ENTER key to get the output of the LXRT latency test
> (which meanwhile seems to be running in the background), while it's not
> necessary for the UP test. It looks like the output is being piped and
> buffered, but I'm not sure of this. Is it?
I'm piping the output of "rtai-load" to "tee" to be able so see the 
output while it's running while capturing it to a file as well. 
> o I've got XXL latencies on my test box, which otherwise works fine
> using my development setup. Maybe it's a config problem. I'll check
> this.
Thanks. I'm curious about what might have caused it.

> o It would be nice to have some clock somewhere displaying the current
> test duration.
I think that will be difficult to achieve without modifying the sourcecode
of the tests.  Of course I can change the sourcecode, but for the moment
I wanted to be able to take a unpatched RTAI release or CVS sourcecode
and use it for the LiveCD.

That's also the reason why I used "tee" to capture the output of the 
LXRT latency task.  

What's your opinion on this? Better to change the test source code
for the binaries on the LiveCD?

With friendly regards,

Panagiotis Issaris
Katholieke Universiteit Leuven
Division Production Engineering,
Machine Design and Automation
Celestijnenlaan 300B              panagiotis.issaris at mech.kuleuven.ac.be
B-3001 Leuven Belgium                 http://www.mech.kuleuven.ac.be/pma

More information about the Rtai mailing list