Le dimanche 14 février 2010 à 03:33 +0200, Ville Syrjälä a écrit :
> On Sun, Feb 14, 2010 at 12:54:09AM +0100, Lionel Landwerlin wrote:
> > Le samedi 13 février 2010 à 23:41 +0200, Ville Syrjälä a écrit :
> > > On Sat, Feb 13, 2010 at 07:49:37PM +0100, Lionel Landwerlin wrote:
> > > > The null system avoids dependency on a particular system (like fbdev, 
> > > > devmem, etc...)
> > > > when you're writing a gfxdriver for relaying on proprietary interfaces, 
> > > > which can't be
> > > > exposed in an opensource project.
> > > > The behavior of this system is to answer "no" to any memory allocation 
> > > > request, so the
> > > > requests are redirected to the gfxdriver which implements its own 
> > > > allocation system.
> > > > 
> > > > Signed-off-by: Lionel Landwerlin <llandwer...@gmail.com>
> > > > ---
> > > >  configure.in                     |   68 +++++++-----
> > > >  src/core/system.h                |    7 +-
> > > >  src/misc/conf.c                  |   12 +-
> > > >  systems/Makefile.am              |   10 ++-
> > > >  systems/null/Makefile.am         |   38 +++++++
> > > >  systems/null/null.c              |  187 
> > > > +++++++++++++++++++++++++++++++++
> > > >  systems/null/null.h              |   14 +++
> > > >  systems/null/null_surface_pool.c |  211 
> > > > ++++++++++++++++++++++++++++++++++++++
> > > 
> > > Why exactly do you need this dummy pool when it does nothing?
> > 
> > It's just a suggestion. Some people are working with proprietary source
> > and non standard driver interfaces. None of the existing systems matches
> > such interfaces. Instead of integrating a specific system in the
> > DirectFB's sources (which can't be released anyway), we propose the
> > following solution.
> 
> I can understand the reason for the null system module but not the null
> surface pool.
> 
> > In order to allocate video memory through non standard interfaces, we
> > could rely on such "null" system that implicitly redirects the memory
> > allocations requests to the next memory pool. Hopefully, the gfxdriver
> > might have registred its own buffer allocation pool.
> 
> But it would work exactly the same without the null surface pool, right?
> I don't think there's any rule that a system module must register a
> surface pool.
> 

Oh... that's possible. I though a pool registration was requiered for
every system.


_______________________________________________
directfb-dev mailing list
directfb-dev@directfb.org
http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-dev

Reply via email to