[Xenomai-help] RTDM timing problems

[Xenomai-help] RTDM timing problems

[Rodrigo Rosenfeld Rosas]

Hi,

I just can't understand what is happening.

The above code on my driver initialization gives the following results.

        start_time = rtdm_clock_read();
	rtdm_task_busy_sleep(43000);
	temp_time = rtdm_clock_read();
	   rtdm_printk(KERN_INFO "%u - %u = %u\n",(unsigned int) temp_time, (unsigned 
int) start_time, (unsigned int) (temp_time-start_time));

	   rtdm_printk(KERN_INFO "%u - %u = %u\n",(unsigned int) 
(temp_time&0xffffffff), (unsigned int) start_time&0xffffffff, (unsigned int) 
((temp_time-start_time)&0xffffffff));


As I expected, both printk gave the same (wrong?) result:
3175950208 - 2800623168 = 375327040

I was waiting something near to 43000, shouldn't I?

Please, can someone explain me what is going on...

Thanks,

Rodrigo.

	

	
		
_______________________________________________________ 
Yahoo! doce lar. Faça do Yahoo! sua homepage. 
http://br.yahoo.com/homepageset.html 

Re: [Xenomai-help] RTDM timing problems

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 1580 bytes --]

Rodrigo Rosenfeld Rosas wrote:
> Hi,
> 
> I just can't understand what is happening.
> 
> The above code on my driver initialization gives the following results.
> 
>         start_time = rtdm_clock_read();
> 	rtdm_task_busy_sleep(43000);
> 	temp_time = rtdm_clock_read();
> 	   rtdm_printk(KERN_INFO "%u - %u = %u\n",(unsigned int) temp_time, (unsigned 
> int) start_time, (unsigned int) (temp_time-start_time));
> 
> 	   rtdm_printk(KERN_INFO "%u - %u = %u\n",(unsigned int) 
> (temp_time&0xffffffff), (unsigned int) start_time&0xffffffff, (unsigned int) 
> ((temp_time-start_time)&0xffffffff));
> 
> 
> As I expected, both printk gave the same (wrong?) result:
> 3175950208 - 2800623168 = 375327040
> 
> I was waiting something near to 43000, shouldn't I?
> 
> Please, can someone explain me what is going on...
> 

Basically confirmed:

I-pipe: Domain Xenomai registered.
Xenomai: hal/x86 started.
Xenomai: real-time nucleus v2.1-rc2 (Champagne) loaded.
Xenomai: starting RTDM services.
4194262080 - 2680902336 = 1513359744


But when I additionally load xeno_posix to get the system timer started:

Xenomai: starting POSIX services.
3899754503 - 3899706087 = 48416


The point is that rtdm_clock_read() relies on the nucleus timer being
started. Otherwise, it just returns plain wrong numbers.

This makes me scream again for that damn timer being started
AUTOMATICALLY! I guess Philippe knows what I'm talking about... ;)
Meanwhile, we should state this in the docs. At least for 2.0.x, this
will remain true anyway.

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]

Re: [Xenomai-help] RTDM timing problems

[Jeroen Van den Keybus]

[-- Attachment #1: Type: text/plain, Size: 544 bytes --]

 Jan,

Can't you start it with a short Linux user-space app upon startup ? Doesn't
even need to go realtime.

I'm planning to try this as I want to start two unrelated Xenomai processes
which need the timer for timeouts (obviously in a perpetual effort to test
the Xenomai kernel ;-))

Jeroen.


This makes me scream again for that damn timer being started
> AUTOMATICALLY! I guess Philippe knows what I'm talking about... ;)
> Meanwhile, we should state this in the docs. At least for 2.0.x, this
> will remain true anyway.
>

[-- Attachment #2: Type: text/html, Size: 812 bytes --]

Re: [Xenomai-help] RTDM timing problems

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 640 bytes --]

Jeroen Van den Keybus wrote:
>  Jan,
> 
> Can't you start it with a short Linux user-space app upon startup ? Doesn't
> even need to go realtime.
> 
> I'm planning to try this as I want to start two unrelated Xenomai processes
> which need the timer for timeouts (obviously in a perpetual effort to test
> the Xenomai kernel ;-))
> 

Of course, you can. And this is also what we are doing in our software
so far. Anyway, the point is that this step is superfluous if only
aperiodic mode is available. Moreover, it's inconsistent compared to
other skin's interfaces. But there is hope we will overcome it one day. :)

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]

Re: [Xenomai-help] RTDM timing problems

[Rodrigo Rosenfeld Rosas]

Em Segunda 30 Janeiro 2006 16:42, Jan Kiszka escreveu:

>The point is that rtdm_clock_read() relies on the nucleus timer being
>started. Otherwise, it just returns plain wrong numbers.
>
>This makes me scream again for that damn timer being started
>AUTOMATICALLY! I guess Philippe knows what I'm talking about... ;)
>Meanwhile, we should state this in the docs. At least for 2.0.x, this
>will remain true anyway.

What can I do for now? What is the relation between the system time and the 
timer?

Thank you Jan,

Rodrigo.

	

	
		
_______________________________________________________ 
Yahoo! doce lar. Faça do Yahoo! sua homepage. 
http://br.yahoo.com/homepageset.html 

Re: [Xenomai-help] RTDM timing problems

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 669 bytes --]

Rodrigo Rosenfeld Rosas wrote:
> Em Segunda 30 Janeiro 2006 16:42, Jan Kiszka escreveu:
> 
>> The point is that rtdm_clock_read() relies on the nucleus timer being
>> started. Otherwise, it just returns plain wrong numbers.
>>
>> This makes me scream again for that damn timer being started
>> AUTOMATICALLY! I guess Philippe knows what I'm talking about... ;)
>> Meanwhile, we should state this in the docs. At least for 2.0.x, this
>> will remain true anyway.
> 
> What can I do for now? What is the relation between the system time and the 
> timer?

Just start the timer before using rtdm_clock_read(). This is now also
stated in the docs.

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]

Re: [Xenomai-help] RTDM timing problems

[Rodrigo Rosenfeld Rosas]

Em Quinta 02 Fevereiro 2006 10:48, voc=EA escreveu:

>Rodrigo Rosenfeld Rosas wrote:
>> Em Segunda 30 Janeiro 2006 16:42, Jan Kiszka escreveu:
>>> The point is that rtdm_clock_read() relies on the nucleus timer being
>>> started. Otherwise, it just returns plain wrong numbers.
>>>
>>> This makes me scream again for that damn timer being started
>>> AUTOMATICALLY! I guess Philippe knows what I'm talking about... ;)
>>> Meanwhile, we should state this in the docs. At least for 2.0.x, this
>>> will remain true anyway.
>>
>> What can I do for now? What is the relation between the system time and
>> the timer?
>
>Just start the timer before using rtdm_clock_read().

Sorry, It is not clear to me how can I do that.

Regards,

Rodrigo.

	

	
		
_______________________________________________________ 
Yahoo! doce lar. Faça do Yahoo! sua homepage. 
http://br.yahoo.com/homepageset.html 

Re: [Xenomai-help] RTDM timing problems

[Jan Kiszka]

[-- Attachment #1: Type: text/plain, Size: 1061 bytes --]

Rodrigo Rosenfeld Rosas wrote:
> Em Quinta 02 Fevereiro 2006 10:48, você escreveu:
> 
>> Rodrigo Rosenfeld Rosas wrote:
>>> Em Segunda 30 Janeiro 2006 16:42, Jan Kiszka escreveu:
>>>> The point is that rtdm_clock_read() relies on the nucleus timer being
>>>> started. Otherwise, it just returns plain wrong numbers.
>>>>
>>>> This makes me scream again for that damn timer being started
>>>> AUTOMATICALLY! I guess Philippe knows what I'm talking about... ;)
>>>> Meanwhile, we should state this in the docs. At least for 2.0.x, this
>>>> will remain true anyway.
>>> What can I do for now? What is the relation between the system time and
>>> the timer?
>> Just start the timer before using rtdm_clock_read().
> 
> Sorry, It is not clear to me how can I do that.

rt_timer_start() if you use the native API in your application. But I
would suggest to maintain a simple helper tool just for starting /
stopping the timer for native apps as long as this is explicitly needed.
As I indicated, plans exist to overcome this in 2.1.

Jan



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]