Re: QUESTIONS ABOUT PREEMPT_RT on S3C24X0

gisspirit wrote:
>> gisspirit wrote:
>>> Hi,all
>>>   This is the first time that I send a email to the linux arm mailing list.
>>> Recently I began to study CONFIG_PREEMPT_RT  in  the Real-Time Linux  WIFI,unfortunately I find that  CONFIG_PREEMPT_RT does not perform well on 
>>> all the arms(refer to 
>>> "1.6Which architectures does the CONFIG_PREEMPT_RT patch support? " 
>>> on the web sitehttp://rt.wiki.kernel.org/index.php/Frequently_Asked_Questions).
>>> But I find the some platforms tested and in use with CONFIG_PREEMPT_RT ,such as
>>> Atmel AT91RM9200-EK ,(refer to http://rt.wiki.kernel.org/index.php/CONFIG_PREEMPT_RT_Patch#Platforms_Tested_and_in_Use_with_CONFIG_PREEMPT_RT)
>>>     
>>>     What I want to konw is that whether someone has tested S3C24X0 with CONFIG_PREEMPT_RT.
>>>     Best wishes to all!
>> I do not know about S3C24X0, but have tested the PREEMPT_RT patch on
>> Intel IXP465, and it seemed to work well enough. However, this is not
>> the reason for my post. What I wanted to tell you is that if you are
>> concerned about performance, there is something which you should really
>> test, it is the FCSE patch, last posted on this list here:
>> http://lists.arm.linux.org.uk/lurker/message/20090505.073913.9e5b2512.en.html
>>
>> You can combine it with the PREEMPT_RT patch, and on the platforms I
>> tested it, when used in "guaranteed" mode, it improves scheduling
>> latency somewhere between 50us and 100us.
>>
>> -- 
>>                                                 Gilles.
>   I appreciate your help very much! 
>   S3C24X0 is short for S3C2410 or S3C2440. It is  ARM920T architecture produced by SUMSANG ELECTRONIC.
>    There is another question which I want to ask you is that how I can test the various latency in the linux kernel?

Please do not drop the mailing list from the CC. Your questions may
interest other people.

If you want to measure user-space scheduling latency, you may use the
cyclictest tool:
http://rt.wiki.kernel.org/index.php/Cyclictest
Note however that to get meaningful results, you have to run Linux timer
in one-shot mode (that is enable high-resolution timers, supposing that
it is supported by your platform).

For in-kernel measurement, the built-in latency tracing tool have
unfortunately too much overhead on the ARM platform where I could test
them to give meaningful results, but setting up a small driver to do
these measurements should not be too hard.

-- 
                                                 Gilles.

-------------------------------------------------------------------
List admin: http://lists.arm.linux.org.uk/mailman/listinfo/linux-arm
FAQ:        http://www.arm.linux.org.uk/mailinglists/faq.php
Etiquette:  http://www.arm.linux.org.uk/mailinglists/etiquette.php