Why RTEMS timeslice is not as configured.

[Fernando Nicodemos]

Just one correction. CPU is TSC695FL, not TSC695F.

*Fernando G. Nicodemos*
*Diretor de Pesquisa e Desenvolvimento

***®*NCB - Sistemas **Embarcados Ltda.*
*23° 09’ 24.87”S**     45° 47’ 26.49”W*
São José dos Campos - SP - Brasil
t.:+55 (12) *3905.2243*                *
www.ncb.ind.br*

*AVISO:* Esta mensagem pode conter informações privilegiadas e seu conteúdo
é confidencial. É endereçada somente para destinatários conhecidos. Se você
não for um destinatário conhecido, não está autorizado a distribuir,
divulgar, imprimir, copiar e se utilizar desta mensagem e seu conteúdo.


2013/7/11 Fernando Nicodemos <fgnicodemos at ncb.ind.br>

> See answers bellow, in the text body.
>
> *Fernando G. Nicodemos*
> *Diretor de Pesquisa e Desenvolvimento
>
> ***®*NCB - Sistemas **Embarcados Ltda.*
> *23° 09’ 24.87”S**     45° 47’ 26.49”W*
> São José dos Campos - SP - Brasil
> t.:+55 (12) *3905.2243*                *
> www.ncb.ind.br*
>
>
> 2013/7/9 Joel Sherrill <joel.sherrill at oarcorp.com>
>
>>  On 7/8/2013 2:59 PM, Fernando Nicodemos wrote:
>>
>>  Dears,
>>
>>  I configured 4 equal Tasks in RTEMS and the table as follows:
>>
>> rtems_task_create(
>> Task_name[ 0 ],
>> 2,
>> RTEMS_MINIMUM_STACK_SIZE * 2,
>> RTEMS_PREEMPT | RTEMS_TIMESLICE | RTEMS_ASR | RTEMS_INTERRUPT_LEVEL(0),
>> RTEMS_GLOBAL,
>> &Task_id[ 0 ]
>> );
>>
>>  They will operate in round-robin.
>>
>>  #define CONFIGURE_MICROSECONDS_PER_TICK 10000
>>  #define CONFIGURE_TICKS_PER_TIME_SLICE 50
>>
>>  The result should show a timeslice of 500 milisseconds, but what i see
>> in osciloscope is more then 500ms, exactly 834ms. This code is running in a
>> ERC32 development kit from Atmel.
>>
>>   Where are the pins being written in the code to measure this?
>>
>
> FER-> ERC32 GPIO pins are written. Each GPIO pin 4-7 is alocated to one
> task. See task code bellow (this code is the same for the 4 tasks, except
> for the respective GPIO):
> /**********************
>  * RTEMS TASK 1
>  */
> rtems_task Task1(
>   rtems_task_argument unused
> )
> {
>   volatile uint32_t task1_wait_count = 2;
>   /**********************
>    * Generate a square output on GPIO4
>    * Manual loop-striping was applied for optimization
>    */
>   for(;;)
>   {
>     SET(GPIO4);
>     WAIT(task1_wait_count);
>     CLR(GPIO4_7);
>     WAIT(task1_wait_count);
>   }
> }
> Macros are:
> #define SET(gpio) ( *((volatile uint32_t *)ERC32_GPIDATR_ADDR) |= gpio )
> #define CLR(gpio) ( *((volatile uint32_t *)ERC32_GPIDATR_ADDR) &= ~gpio )
> #define WAIT(count) \
>     { \
>       volatile uint32_t i; \
>       for (i = 1; i == count; i++) { }; \
>     }
>
> #else
>
>
>>   Any ideas what is missing for this extra time?
>>
>>   What speed is the erc32?
>>
>> I am curious if the assumed CPU speed is different from your actual
>> board.
>>
>
> FER-> I did not change the default board values. I did not make any
> register configuration on core or system clocks. CPU is TSC695F wich leads
> a maximum 25MHz frequency. The board has two oscilators. One is 3.6864MHz
> and the other is 24MHz. So the CPU is running at 24MHz and sysclk is 12MHz
> (i measured in the sysclk pin and it is really 12MHz - remember that in
> ERC32 sysclk is automatically divided by 2).
>
>
>
>>   Thanks,
>>
>>  *Fernando G. Nicodemos*
>>
>>
>>
>> --
>> Joel Sherrill, Ph.D.             Director of Research & Development joel.sherrill at OARcorp.com        On-Line Applications Research
>> Ask me about RTEMS: a free RTOS  Huntsville AL 35805
>> Support Available                (256) 722-9985
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rtems.org/pipermail/users/attachments/20130711/429f56ec/attachment.html>