[MPlayer-dev-eng] [PATCH] very experimenal nvidia_vid

[D Richard Felker III]

On Fri, Sep 26, 2003 at 08:17:55PM +0400, Dmitry Baryshkov wrote:
> > > > >I see another problem: vidix API must be expanded to pass bpp to vidix
> > > > driver (for rivatv_overlay_colorkey). Also dunno, if BPP <15 or >24 are
> > > >supported by all cards. (IIRC RIVA128 didn't support even 16bpp).
> > > 
> > > It is not a problem as we can read those values directly from the card.
> > > I implemented it for nv03 but the registers for nv04 and later are
> > > different.
> > > Maybe the #if 0 block from rivafb helps you.
> > Will look at it. BTW: do you mean rivafb from rivatv, or from kernel?
> Ok. 0x1c9 always return 0x0. Others regs don't help.
> right BPIXEL seems to be 0x724, but... reading from it from X with
>  nvidia driver locks the whole box. Only SysRq helps :(nv driver works)
> Suppose I (or somebody other) will write implementation of bpp detection
> but it must be disable by default.
> Anyway, reading from PCRTC:0x28 helps, but it can't detect 15bpp mode.
> So, colorkey will be computed wrong, and overlay won't work.
> Maybe we should switch VIDIX from magneta colorkey to g-b only?

IMO just disable colorkey or set it to black. The problems with
picking the right rgb pixel value and the ugly magenta flashes users
observe briefly between movies just aren't worth it to make the window
overlap right. If you insist on using color key, at least use
r=0,g=0,b=1 so it's simple...

Rich