Posix patchs to support affinity attribute in newlib

Jennifer Averett Jennifer.Averett at OARcorp.com
Mon Mar 3 18:01:02 UTC 2014


> -----Original Message-----
> From: gedare at gwmail.gwu.edu [mailto:gedare at gwmail.gwu.edu] On
> Behalf Of Gedare Bloom
> Sent: Monday, March 03, 2014 10:46 AM
> To: Sebastian Huber
> Cc: Jennifer Averett; rtems-devel at rtems.org
> Subject: Re: Posix patchs to support affinity attribute in newlib
> 
> On Mon, Mar 3, 2014 at 10:56 AM, Sebastian Huber
> <sebastian.huber at embedded-brains.de> wrote:
> > On 2014-03-03 16:36, Jennifer Averett wrote:
> >>>
> >>> >Is the implementation correct for CPU sets which are not pre-
> allocated?
> >>> >
> >>
> >> Until we have a system that supports over 32 CPU's the implementation
> >> can not be tested for this.  I minimized the locations that will have
> >> to be touched when we support this but I didn't think it was correct
> >> to add code that can not be tested.
> >>
> >
> > You can test the inline functions separately.  You can add an
> > _Assert() if someone hits a not implemented path.  If you copy an
> > attribute with somehow allocated CPU sets, then you need to allocate
> somehow memory for the copy.
> > This allocation may fail, so you cannot use a void function?
> >
> This is a good point. Most copy functions return the destination pointer.
> 

I'll change this inline to return an int and add Assert here and in the creation
of a thread.  It already Asserts in initialization if more than 32 processors are
present in the system.

> >
> > --
> > Sebastian Huber, embedded brains GmbH
> >
> > Address : Dornierstr. 4, D-82178 Puchheim, Germany
> > Phone   : +49 89 189 47 41-16
> > Fax     : +49 89 189 47 41-09
> > E-Mail  : sebastian.huber at embedded-brains.de
> > PGP     : Public key available on request.
> >
> > Diese Nachricht ist keine geschäftliche Mitteilung im Sinne des EHUG.
> > _______________________________________________
> > rtems-devel mailing list
> > rtems-devel at rtems.org
> > http://www.rtems.org/mailman/listinfo/rtems-devel




More information about the devel mailing list