Thread (14 messages) 14 messages, 5 authors, 2007-10-08

Re: Cyclictest expected results

From: Robert Schwebel <hidden>
Date: 2007-09-30 18:32:59

On Sun, Sep 30, 2007 at 09:07:19PM +0530, Jaswinder Singh wrote:
Here are my results on Intel Pentium 4 :

Command line :  ./cyclictest -t 1 -p 80 -i 10000 -l 10000 -q
Can you post the output of /proc/timer_list?
2.6.23-rc8-rt1 :
WARNING: High resolution timers not available
T: 0 ( 2003) P:80 I:10000 C:  10000 Min:   1500 Act: 6129 Avg: 4502 Max:    7832
That one should give you sane results.
From where I can get Linux-2.6.16-hrt5 and Linux-2.6.16-rt12 patches.
I don't think it makes much sense to play with such an old kernel any more.
CONFIG_HIGH_RES_TIMERS=y are defined in .config then why I am getting
"WARNING: High resolution timers not available".

Why my results are not so impressive in realtime even though I defined:
CONFIG_PREEMPT_RT=y
CONFIG_PREEMPT=y
CONFIG_PREEMPT_SOFTIRQS=y
CONFIG_PREEMPT_HARDIRQS=y
CONFIG_PREEMPT_BKL=y
Doesn't look too bad. Are you sure you actually _bootet_ that kernel? :)
Min, Avg and Max are showing cyclic latencies. Ideally It should be
ZERO. Am I right ?
You won't get zero. On x86 these latencies should be well below 100 us
maximum time, with an average in the lower 10 us range.

Robert
-- 
Pengutronix - Linux Solutions for Science and Industry
Entwicklungszentrum Nord     http://www.pengutronix.de
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help