[MPlayer-dev-eng] [PATCH] partial bigendian support for nvidia_vid

[Aurelien Jacobs]

On Mon, 24 Oct 2005 17:26:37 +0200
Sascha Sommer <saschasommer at freenet.de> wrote:

> Hi,
> 
> On Monday 24 October 2005 14:27, Diego Biurrun wrote:
> > Sascha,
> >
> > this removes to le2me_32 macros from nvidia_vid.c since apparently
> > the values are already in machine endian order.  It's not enough for
> > a complete bigendian port, but it is a necessary step and simplifies
> > the code somewhat.
> >
> > OK to commit?  You wrote this yourself, so I assume it is safe, but
> > I cannot test on x86, so I'd love to hear a quick go ahead for this.
> >
> 
> It should be safe. However I'm not sure if all cards that also work on
> PPC can  swap the bytes automatically and I also do not know if the
> swapping is  required. It might be possible that depending on the (X)
> driver the bytes are  swapped or not swapped. Maybe we can keep the
> current versions of the macros  under #if 0 for now.

I don't know if all nvidia cards are able to swap bytes, but anyway, to
do so, a driver as to enable this feature.
And IIRC, nvfb, rivafb and the nv X driver have a different behavior.
So one will enable automatic byte swap and the other won't. That's only
a vague remembering so I may be wrong, but you should at least test your
patch either with the framebuffer and with X.

Aurel