[MPlayer-dev-eng] VIDIX on PPC (was: Re: [MPlayer-cvslog] CVS: main/vidix/drivers nvidia_vid.c, 1.24, 1.25)

[Reimar Döffinger]

Hi,
On Thu, Sep 08, 2005 at 03:12:24AM +0200, Diego Biurrun wrote:
> On Wed, Sep 07, 2005 at 08:32:12PM +0200, Sascha Sommer wrote:
> > You could also try to remove the le2me_32 from the VID_RD32 /VID_WR32 macros.
> > I just read that nvidia cards can byteswap the config registers.

Yes, but only >= NV10 it seems, this is from nv10GetConfig in the X
driver:
#if X_BYTE_ORDER == X_BIG_ENDIAN
  /* turn on big endian register access */
  if(!(pNv->PMC[0x0004/4] & 0x01000001)) {
    pNv->PMC[0x0004/4] = 0x01000001;
    mem_barrier();
  }
#endif
No such thing in the nv4GetConfig function...

> This appears to be a step in the right direction, although I get a
> _really_ funky result.  The video now alternates between the correct
> image and a hard to describe effect.  Across the whole width of the
> video window artifacts cover the whole height of the monitor.  The top
> third are big pink and green horizontal stripes with speckled blocks in
> them, the bottom two thirds are covered with vertical stripes of all
> colors with a heavy pink predominance.

Try -nodouble.
Also try with rebooting and not starting X, the registers should not be
set to bigendian then I think.
If you can get a svgalib version that works on PPC also try -vo
svga:vidix

Greetings,
Reimar Döffinger