Re: Re: xcalibur codes

On Friday 24 June 2005 19:30, Kevin Shepherd wrote:
> Mathew Saxton wrote:
> > Just wondering if this went anywhere.. Good luck with it..
>
> Thank you.
> Progress so far:
> I have incorporated the new code tables into the video driver
> and compiled this into the 2.4.30 kernel.
> This seems to be setting the xcalibur tv encoder, but not producing
> a valid setup, because the screen becomes a garbled mess.
> I believe that the riva setup, through PCI write, needs the
> associated timing to correspond to these tv encoder settings.
> In order to find these out, I now plan to run a debug bios to
> trace the exact numbers sent to the nvidia chip ( through pci )
> and the xcalibur chip ( through SMBus ).  This way I should
> be able to calculate back to the pixel clock, horz and vert
> margins, and to check that the 0x90 range captures all encoder
> config.
>
> Oh for some xcalibur docs!
> I heard the rumour that it might be an SIS design, and found
> that the SiS301C video bridge looked most likely, but there
> are no docs for this that I could find - only the X driver.
>
> I expect that my reverse engineering approach will lead to
> success, docs would simply save time.
>
> I will post any further progress.

Run xbv under linux (make yourself a kernel that does not do any video init  - 
ie no xboxfb, and look at the NVHTOTAL and NVVTOTAL, CRTC_HTOTAL, and 
CRTC_VTOTAL values.

Try it for 480p first, and see how those values are hardwired into the 
existing xcalibur driver, and then you will be simply able to plug in the new 
ones.  That's how I got 480p working.  The simple reason I didn't get 720p 
and 1080i sorted is that I had no games that supported those modes that I 
could then exploit, in order to dump the chip settings.

David

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iD8DBQBCvGDFOIrGtzS07WgRApQBAJ9U2ct9SicvQBFxIXd72C3h/O/F+QCfQCsV
4PFOrm/YZys7udnfVV/v6oA=
=ckiH
-----END PGP SIGNATURE-----