All of lore.kernel.org
 help / color / mirror / Atom feed
From: Gilles Chanteperdrix <gilles.chanteperdrix@xenomai.org>
To: BOUIN Alexandre <abouin@domain.hid>
Cc: adeos-main@gna.org
Subject: Re: [Adeos-main] New patch for AT91SAM9261 adeos support for 2.6.19 kernel
Date: Tue, 24 Apr 2007 15:26:50 +0200	[thread overview]
Message-ID: <462E059A.2010607@domain.hid> (raw)
In-Reply-To: <403B28AE6257B548B66BB88414AC78330174BBDF@STEGOSAURE.adetel.com>

BOUIN Alexandre wrote:
>>It is a high.
> 
> by default, nucleus latency is set to 9342. I can diminish this value
> to 323µs only. Is it possible I've done something wrong ? Is there
> another way to reduce theses latencies ?
> 

The latency value is a value in nanoseconds, so 9342 is 9us. It is a
value used by the nucleus to compensate for the systematic latency of
your system. Usually, I set it to approximately the minimum value
returned by latency (plus the current value of the latency) minus a
small saftety margin, so that latency minimum value is close to 0 and I
do not get negative values.

> 
>>You should get plenty of negative values.
> 
> I've sent you my histogram and I don't see any negative value.
> Could you precise what was your thought ?

Look at the code:
static inline void add_histogram(long *histogram, long addval)
{
        /* bucketsize steps */
        long inabs =
            rt_timer_tsc2ns(addval >= 0 ? addval : -addval) / bucketsize;
        histogram[inabs < histogram_size ? inabs : histogram_size - 1]++;
}

So, you will never see a negative value in the histogram.


> 
> 
>>What is your kernel configuration, do not you have some
> 
> statistics/debug options enabled ?
> It is right : I've let some kernel debug options while performing these tests.
> I change it right now. 

I was thinking about Xenomai debug/statistics option. Xenomai nucleus
statistics collection, for instance, is enabled by default and should be
disabled for best performance. Check also that you did not enable
Xenomai debug options, or the I-pipe tracer.

> 
> Test was launched without any kernel debug options, and we got a worst
> case of 197µs now.
> Here is a sum up of our results : 
> [Board] : 			 AT91SAM9261EK
> [Kernel version] : 	 2.6.19-at91-xenomai-adeos
> [Clocks] : 			 CPU 198 MHz, master 99 MHz, main 18.432 MHz
> [Nucleus latency] : 	 323µs
> [Root] : 			 nfs
> [Kernel debug options] : No
> [Load Softwares] :	 calibrator + ping host + bonnie++
> [Duration] : 		 > 15h
> [Best Latency] : 		 45µs
> [Average Latency] : 	 82.5µs
> [Worst Latency] : 	 197µs

The best way to give us all the useful informations about your platform
and configuration is to use the xeno-test script. Sorry I did not
mention this earlier. You will have to use the -W option to use your
custom load generation script.

-- 
                                                 Gilles Chanteperdrix


  reply	other threads:[~2007-04-24 13:26 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <403B28AE6257B548B66BB88414AC78330174BBDE@STEGOSAURE.adetel.com>
2007-04-24 13:14 ` [Adeos-main] New patch for AT91SAM9261 adeos support for 2.6.19 kernel BOUIN Alexandre
2007-04-24 13:26   ` Gilles Chanteperdrix [this message]
     [not found] <403B28AE6257B548B66BB88414AC78330174BBEA@STEGOSAURE.adetel.com>
2007-05-03 11:45 ` BOUIN Alexandre
     [not found] <403B28AE6257B548B66BB88414AC78330174BBD8@domain.hid>
2007-04-20 15:21 ` BOUIN Alexandre
2007-04-20 15:27   ` Gilles Chanteperdrix
2007-04-20 14:01 Gregory CLEMENT

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=462E059A.2010607@domain.hid \
    --to=gilles.chanteperdrix@xenomai.org \
    --cc=abouin@domain.hid \
    --cc=adeos-main@gna.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
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.