How many Timers Can I Create???

[Alex]

Ok, but I am still looking to this behaiviour as strange, because the pc I am using has 256 MB of ram and if the rtems see memory as a flat fashion, It looks like that rtems is waisting memory in timers... Sorry about my statement ... I am beginner in rtems, but now I am interesting in find out the reason... I think we all!

So if with 256 mb rtems only allows to create 10 timers, well If we need 20 timers We have to put ~512 Mb on the PC. It looks very strange...

Alex



----- Mensagem Original -----
De: Joel Sherrill <joel.sherrill at OARcorp.com>
Data: Segunda-Feira, 12 de Abril de 2004, 17:18
Assunto: Re: How many Timers Can I Create???

> Kamen Penev wrote:
> 
> > Joel Sherrill wrote:
> > 
> >> Alex wrote:
> >>
> >>> Hi,
> >>>
> >>> Does RTEMS has a confifured maximum number o timers it 
> supports, or 
> >>> it naturally crashs when too much timers are created?
> >>>
> >>> In an test I am doing I have the following:
> >>>
> >>> #define CONFIGURE_MAXIMUM_POSIX_TIMERS 150
> >>> #define CONFIGURE_MAXIMUM_ 150
> >>>
> >>> Then I have source code to create 100 timers, but when the 
> program 
> >>> starts it hangs, even before starting sreating the first 
> timer. The 
> >>> timers are created with the timer_create POSIX function.
> >>
> >>
> >>
> >> Sounds like the BSP was unable to assign the memory for that much
> >> workspace.
> > 
> > 
> > 
> > I see this in confdefs.h (ss-20030417):
> > 
> > #define CONFIGURE_MEMORY_FOR_POSIX_TIMERS(_timers) \
> >  ((_timers) * \
> >   ( 0 ) )
> > 
> > Maybe this is why?
> 
> That would explain it also. :)
> 
> I would guess that the proper definitions in
> cpukit/sapi/include/confdefs.h should be:
> 
> #define CONFIGURE_MEMORY_FOR_POSIX_TIMERS(_timers) \
>   ((_timers) * sizeof(POSIX_Timer_Control))
> 
> And since POSIX timers implicitly use a Classic API timer:
> 
> #define CONFIGURE_MEMORY_FOR_TIMERS(_timers) \
>   ((_timers + CONFIGURE_MAXIMUM_POSIX_TIMERS) * \
>      ( sizeof(Timer_Control) + CONFIGURE_OBJECT_TABLE_STUFF ) )
> 
> My earlier statement about the bspstart.c failing still
> apply.  But this time you probably got slightly further
> in initialization before things failed.  RTEMS initialization
> probably ran out of workspace duging the preallocation of
> object control blocks.
> 
> If this works, let us know so a PR can be filed.
> 
> > Kamen
> > 
> 
> 
> 
___________________________________________________________________________________________
IOL Flash. A net normal em versao acelerada.
Promocao: Ligue-se a Internet pelo IOL Flash e ganhe um Vale de €10 da Galp!
Saiba como: http://www.iol.pt/central_utilizador/iol_flash/promocao.php