* [Xenomai-core] native syscalls
@ 2009-04-09 2:53 Steven Seeger
2009-04-09 9:25 ` Gilles Chanteperdrix
0 siblings, 1 reply; 3+ messages in thread
From: Steven Seeger @ 2009-04-09 2:53 UTC (permalink / raw)
To: xenomai
[-- Attachment #1: Type: text/plain, Size: 3622 bytes --]
Can xenomai preempt its own system calls? We ran latency with our
app. We had a thread waiting on a cond, and when we signaled that
cond, the latency jumped from ~70us to almost 400us even though
latency runs a higher priority thread than the thread that we wake up.
Is there interrupt disabling that accounts for this?
Here are the results as taken by Mark S.
== All results in microseconds
warming up...
RTT| 00:00:01 (periodic user-mode task, 100 us period, priority 99)
RTH|-----lat min|-----lat avg|-----lat max|-overrun|----lat best|---
lat worst
RTD| 14.934| 27.179| 65.767| 0| 14.934|
65.767
RTD| 15.151| 26.798| 70.954| 0| 14.934|
70.954
RTD| 15.409| 27.152| 66.890| 0| 14.934|
70.954
RTD| 15.947| 27.316| 71.168| 0| 14.934|
71.168
RTD| 15.746| 26.878| 71.188| 0| 14.934|
71.188
RTD| 15.489| 26.631| 74.476| 0| 14.934|
74.476
RTD| 16.659| 26.527| 67.425| 0| 14.934|
74.476
RTD| 14.567| 26.972| 59.448| 0| 14.567|
74.476
RTD| 15.987| 27.837| 74.092| 0| 14.567|
74.476
RTD| 16.241| 28.352| 66.740| 0| 14.567|
74.476
RTD| 13.982| 27.139| 65.630| 0| 13.982|
74.476
RTD| 13.982| 33.355| 65.286| 3| 13.892|
388.473
RTD| 13.982| 33.355| 65.286| 3| 13.892|
388.473
RTD| 13.982| 33.355| 65.286| 3| 13.892|
388.473
RTD| 13.982| 33.355| 65.286| 3| 13.892|
388.473
RTD| 13.982| 33.355| 65.286| 3| 13.892|
388.473
RTD| 13.791| 33.465| 67.351| 3| 12.702|
388.473
RTD| 13.791| 33.465| 67.351| 3| 12.702|
388.473
RTD| 13.791| 33.465| 67.351| 3| 12.702|
388.473
RTD| 13.791| 33.465| 67.351| 3| 12.702|
388.473
RTD| 13.761| 33.722| 90.066| 3| 12.702|
388.473
RTT| 00:00:22 (periodic user-mode task, 100 us period, priority 99)
RTH|-----lat min|-----lat avg|-----lat max|-overrun|----lat best|---
lat worst
RTD| 13.283| 29.361| 59.140| 3| 12.702|
388.473
RTD| 11.502| 29.040| 58.235| 3| 11.502|
388.473
RTD| 10.530| 28.947| 67.027| 3| 10.530|
388.473
RTD| 14.410| 31.760| 65.229| 3| 10.530|
388.473
RTD| 13.207| 31.286| 73.607| 3| 10.530|
388.473
RTD| 13.454| 30.678| 71.275| 3| 10.530|
388.473
RTD| 14.730| 31.122| 64.090| 3| 10.530|
388.473
RTD| 14.881| 30.397| 80.705| 3| 10.530|
388.473
RTD| 9.848| 30.534| 66.659| 3| 9.848|
388.473
RTD| 14.039| 30.678| 63.525| 3| 9.848|
388.473
RTD| 13.290| 28.973| 71.589| 3| 9.848|
388.473
RTD| 16.228| 23.957| 80.598| 3| 9.848|
388.473
RTD| 17.304| 26.467| 74.018| 3| 9.848|
388.473
RTD| 16.137| 26.494| 73.226| 3| 9.848|
388.473
^
C
---|
------------|
------------|------------|--------|-------------------------
Thanks,
Steven
[-- Attachment #2: Type: text/html, Size: 9306 bytes --]
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [Xenomai-core] native syscalls
2009-04-09 2:53 [Xenomai-core] native syscalls Steven Seeger
@ 2009-04-09 9:25 ` Gilles Chanteperdrix
2009-04-09 11:02 ` Steven Seeger
0 siblings, 1 reply; 3+ messages in thread
From: Gilles Chanteperdrix @ 2009-04-09 9:25 UTC (permalink / raw)
To: Steven Seeger; +Cc: xenomai
Steven Seeger wrote:
> Can xenomai preempt its own system calls? We ran latency with our
> app. We had a thread waiting on a cond, and when we signaled that
> cond, the latency jumped from ~70us to almost 400us even though
> latency runs a higher priority thread than the thread that we wake up.
> Is there interrupt disabling that accounts for this?
>
> Here are the results as taken by Mark S.
As I told you in a voluntarily detailed private mail, only the traces
made by the I-pipe tracer are of any interest for us.
--
Gilles.
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [Xenomai-core] native syscalls
2009-04-09 9:25 ` Gilles Chanteperdrix
@ 2009-04-09 11:02 ` Steven Seeger
0 siblings, 0 replies; 3+ messages in thread
From: Steven Seeger @ 2009-04-09 11:02 UTC (permalink / raw)
To: Gilles Chanteperdrix; +Cc: xenomai
On Apr 9, 2009, at 5:25 AM, Gilles Chanteperdrix wrote:
> As I told you in a voluntarily detailed private mail, only the traces
> made by the I-pipe tracer are of any interest for us.
Gilles, I will send that email to Mark who has hardware to run on. I
apologize for not doing this first but the email you speak of was lost
among the 20 or so we had yesterday. :)
Steven
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2009-04-09 11:02 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2009-04-09 2:53 [Xenomai-core] native syscalls Steven Seeger
2009-04-09 9:25 ` Gilles Chanteperdrix
2009-04-09 11:02 ` Steven Seeger
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.