[MPlayer-dev-eng] [PATCH] nvidia_vid update

[Dmitry Baryshkov]

On Tue, Jan 06, 2004 at 06:13:32PM +0100, Sascha Sommer wrote:
> >> > -    unsigned int next_frame;
> >> > +    unsigned int cur_frame;
> >>
> >> Looks like pure cosmetics. Is this really needed?
> >
> >It's cosmetics :) But IMHO a bit of cosmetics is better, than next_frame
> >variable pointing to current frame and not to the next one.
> 
> Dunno and don't care. Feel free to commit.

Thank you.

> 
> >According to nv4ref.h, only 2 last bits define bpp:
> >/* Controls what the format of the framebuffer is */
> >#define NV_PCRTC_PIXEL                       0x28
> >#define NV_PCRTC_PIXEL_MODE                  7:7
> >#define NV_PCRTC_PIXEL_MODE_TV               0x01
> >#define NV_PCRTC_PIXEL_MODE_VGA              0x00
> >#define NV_PCRTC_PIXEL_TV_MODE               6:6
> >#define NV_PCRTC_PIXEL_TV_MODE_NTSC          0x00
> >#define NV_PCRTC_PIXEL_TV_MODE_PAL           0x01
> >#define NV_PCRTC_PIXEL_TV_HORIZ_ADJUST       5:3
> >#define NV_PCRTC_PIXEL_FORMAT                1:0
> >#define NV_PCRTC_PIXEL_FORMAT_VGA            0x00
> >#define NV_PCRTC_PIXEL_FORMAT_8BPP           0x01
> >#define NV_PCRTC_PIXEL_FORMAT_16BPP          0x02
> >#define NV_PCRTC_PIXEL_FORMAT_32BPP          0x03
> 
> Seems to be the same for nv03. Can someone please test
> with Geforce.
> 
> >> >   /* get 9th bit in CRTC_OVERFLOW*/
> >> >   info->screen_y |= (VID_RD08(info->chip.PCIO,0x03D5) &0x02)<<7;
> >> >   /* and the 10th in CRTC_OVERFLOW*/
> >> >+  VID_WR08(info->chip.PCIO,0x03D4,0x07);
> >>
> >> the register shouldn't change. Does it?
> >
> >Just to be sure. At least it may change (multitasking :).
> 
> I'm against commiting this change. If it would change
> between the two reads it might aswell change between
> the write and the read.

OK.

> 
> >> >+ while (bps == 0)
> >> >+ {
> >> >+  if (info->chip.arch != NV_ARCH_03)
> >> >+   bps = info->chip.PGRAPH[0x00000670/4];
> >> >+  else
> >> >+   bps = info->chip.PGRAPH[0x00000650/4];
> >> >+//  if (bps == 0) fprintf(stderr, "bps returned 0: %d\n");
> >> >+ }
> >> >      x = info->wx - (pan % bps) * 8 / info->depth;
> >> >      y = info->wy - (pan / bps);
> >> >+    }
> >> >+
> >>
> >> I'm currently unable to test it, but afair this register is also 0 in
> >> textmode.
> >
> >Yes, but in text mode info->depth == 0, so the whole block is skipped.
> 
> Please leave this code as it is now unlike you can give me a reason why your
> changes are needed.
> 

It's to get panning work. Under X11 lower resolution (not by XRandR
extension, but with Ctrl-Alt-Plus), Move your desktop down. Now overlaying
point is wrong :(
Horizontal moving of desktop works, but not vertical, because bps isn't
correct in such modes.

This part of patch corrects this. Now, if you have doublebuffering
enabled, it automagically works.

-- 
With best wishes
Dmitry Baryshkov.